Datasets:

WilliamWen

/

summarization_cata

like 0

The industrial revolution and the mass consumption of coal, petroleum and natural gas causes the depletion of conventional fossil-based resources. The continuous consumption of these fossil fuels causes energy shortage and global warming [1–3]. To mitigate the reliance on these resources, many countries have been planning to develop alternative fuels. Biofuel becomes one of the promising options for clean energy and sustainable sources [4,5]. Besides its biodegradability, zero sulphur, a minimal amount of carbon dioxide emission and non-toxic [6,7], it also has a similar characteristic of fossil fuel, with better cetane number, flash point and calorific value [8,9]. Many researchers have been involved in developing green diesel production that can be derived from edible, non-edible oils and fats efficiently, economic, and eco-friendly processes [10]. The deoxygenation reaction can be the potential pathway to enhance the physicochemical properties of these edible and non-edible oils close to the diesel fuel quality standards [2,3]. The final product obtained from the reaction comprises gas, coke, organic liquid product and water [11,12]. This reaction process has advantages in great flexibility in the feedstock selection, characteristic of pyrolysis products associated with petroleum products, fast conversion gradient, and the probability to industrial scale-up for biofuel production [13,14].Nowadays, many heterogeneous solid catalysts have been examined for biofuel production via catalytic cracking such as ZrO2 [2], MgO/Activated carbon [15], Mg-HZSM-5 [16], Ni-Mg/ZSM-5 [17], biochar [18], etc. Many selective solid catalysts become the best option for catalytic cracking of vegetable oils to improve the yield and reduce the cost of liquid fuels owing to the characteristics of these catalysts which are recyclable, regenerable and eco-friendly [1,2]. Zeolites are considered promising catalysts due to their greater cracking ability, unique porous structure, the presence of higher acidity and tremendous inherent stability [16,19]. However, numerous researchers are moved to ordered mesoporous materials due to their advantages such as high thermal stability, ease of surface modification, tolerable pore size, huge surface area and inert nature [20]. Moreover, the improvement on the physicochemical properties of the catalysts obtaining stability during a long period of time and outstanding catalytic conversion has been focused on. These might involve the combination of different transitions metals, the use of basic promoters of supports with oxygen storage capacity, and the optimization of preparation pathways of the catalysts.In biofuel production, supported heterogeneous catalysts are well known for their high activity and reusability. SiO2 [21], Al2O3 [22], TiO2 [23], SnO2 [24], zeolites [25] and other excellent supporting materials provide greater surface area, superior and long stability. Metal oxide catalysts have emerged as essential in most refining and petrochemical processes, as well as in the production of speciality chemicals and, more recently, in the improvement of environmental issues, particularly for depollution by increasing reaction selectivity to avoid unusable by-products [26]. Alkali or alkaline metal-supported solid catalysts are best known as solid base catalysts. Based on Hafriz et al. [27], the base catalyst is expected to obtain cracking oils with low acid values and good quality of oil in terms of good cold flow properties. Potassium will act as a chemical promoter which may increase the activity of the catalyst by an order of magnitude [28]. Besides that, potassium suppresses the production of methane and improves effective hydrocarbon selectivity [29]. A higher level of alkali oxide such as potassium oxide also has an adverse effect on hydrolytic stability when exposed to high humidity and temperature [30].Industrial players are concerned about catalyst deactivation since the cost for new catalyst and process shutdown require billions of dollars per year. The loss activity of the catalyst can be predicted for most processes and yet the drastic consequences must be avoided. Therefore, to ensure a good conversion process, the deactivation issues like the extent rate and reactivation must be determined. The deactivation is happened due to the formation of coke and blocking the micropore opening that is caused by the generation of polycyclic aromatic hydrocarbons and hence, lowers the catalytic activity [17]. A vital parameter to recover the catalytic activity is the capacity of the catalyst to be regenerated. Cleaning the coke deposit from the surface and the pores of the catalyst are some of the regeneration courses involved without destroying or modifying the structure of the catalyst [31]. From the economical aspect, the cost of regeneration is lower than by obtaining a fresh catalyst. From the environmental point of view, it is a spent catalyst that can form toxic metal compounds in the environment. Hence, it is an environmentally friendly option by regenerating the catalyst as an alternative to disposing of it as solid waste [12,32].The comparative study on the performance of K2O/SiO2 and dolomite catalysts, as well as reusability and regenerability study of the deoxygenation using K2O/SiO2 catalyst on deoxygenation reaction, has not yet been reported. Herein, the restrictions of the above mentioned have sparked the investigation on the comparative study, reusability and regenerability study of the K2O/SiO2 catalyst on deoxygenation of WCO. The final product from each cycle was analysed by using GC–MS for chemical composition study.The potassium oxide supported silicon dioxide (K2O/SiO2 or denoted as KSi) catalyst with purity > 99 % was generously provided by Pakar Management Technology (M) Sdn. Bhd. Dolomite was purchased from Northern Dolomite Sdn. Bhd. (Perlis, Malaysia) [33] and was calcined at 900 ℃ for 4 h [34]. The detailed characterization of calcined dolomite catalyst could be referred to Hafriz et al. [27,34]. Nitrogen gas (N2) was supplied from Biogas Sdn. Bhd. The standard for gas chromatography (GC) analysis namely liquid product alkane and alkene standard solution (C8-C20) with an internal standard of 1-bromohexane with purity (98%) were purchased from Sigma Aldrich. n-hexane for GC analysis (Merck) with purity > 98 % was utilized. The feedstock of waste palm cooking oil (WCO) with 82.8 % oxygenated compound and 17.2 % hydrocarbon compound (obtained from GC–MS analysis) was collected from a cafeteria in Serdang, Selangor. The WCO was preheated at 100 ℃ (30 min) to remove moisture content. However, as reported by Abdulkareem et al. [35], the moisture content in WCO will increase the decomposition of triglycerides into Free Fatty Acid (FFAs) and directly promote the oxygenation reaction under examination of the WCO deoxygenation under a series of FFAs (0–20 %) and water content (0.5–20 wt%). The main composition of WCO using GC–MS analysis has been shown in Fig. 1 . As tabulated in Fig. 1, the main groups of compounds present in WCO were carboxylic acid (59.13 %) and ester (14.58 %). There were three dominant peaks observed in WCO composition which is oleic acid (26.31 %), palmitic acid (20.64 %) and stearic acid glycidyl ester (9.12 %). As reported by Hafriz et al. [36], carboxylic acids were maintained even after deep frying as the boiling point of oleic acid is 358.85 ℃ and for palmitic acid is 351 ℃. These carboxylic acid and ester content are the major indicators for green diesel produced in terms of quality properties via catalytic deoxygenation.In order to identify the crystallography and dispersion states of the synthesized catalyst, an X-ray diffraction (XRD) analysis was performed. The XRD analysis was conducted using a Shimadzu diffractometer (model XRD-6000). A Brunauer-Emmett-Teller (BET) method was used to evaluate the specific surface area and pore distribution of the synthesized catalysts. The analysis was performed using Thermo-Finnigan Sorpmatic 1990 series with an N2 adsorption/desorption analyzer in a vacuum chamber at −196 °C. The samples were degassed overnight at 150 ℃ to remove moisture and foreign gases from the catalyst surfaces. Temperature-programmed desorption with CO2 as probe molecules has been used to investigate the basicity of the synthesized catalysts. The CO2-TPD analysis was carried out by using Thermo Finnigan TPD/R/O 1100 instrument equipped with a thermal conductivity detector (TCD). In the pretreatment step, 0.05 g of sample was pre-treated under N2 gas flow for 30 min at 250 ℃, followed by CO2 gas at ambient temperature for 1 h to allow adsorption of CO2 onto the surfaces. The excess CO2 was subsequently flushed with N2 gas at a flow rate of 20 ml/min for 30 min. The desorption of CO2 from the basic sites of the catalyst was detected by TCD under helium gas flow (30 ml/min) from 50 ℃ to 900 ℃ for 30 min. The area under the graph (CO2 desorption peak) provided will determine the basicity of the synthesized catalyst. The morphological characteristics and elemental analysis of the catalysts were investigated by using Scanning Electron Microscopy with energy-dispersive X-ray spectroscopy (SEM/EDX). The SEM images were observed through SEM LEO 1455 VP electron microscope with acceleration voltages of 30 kV. For the sample preparation, the catalyst powder was dispersed on an aluminium sample holder and glued by using double-sided tape. Then, it was coated with a thin layer of gold which is a type of conducting material using BIO-RAS Sputter Coater. Micrographs were recorded at various magnifications. The compositional analysis of the solid base catalyst was carried out by EDX that gives the flexibility of image control. The metal elements peaks were detected in the EDX spectra and the atomic percentages of the constituent elements presence in the fresh catalyst and regenerated catalyst were determined in the solid base catalyst.The deoxygenation reactions were performed in a fractionated cracking system as illustrated in Fig. 2 . 150 g of WCO was preheated before being poured into the round bottom flask. 5 wt% catalysts were added to the reactor flask. The reaction was carried out under nitrogen (N2) flow with a flow rate of 150 cm3/min to eliminate the air inside the system. The sample was heated for 30 min until the temperature reaches 390 ± 5 °C with a heating rate of 100 °C/min. The liquid product was collected in the collecting flask. The catalytic performance of KSi as a deoxygenation catalyst was compared with thermal deoxygenation (without catalyst) and calcined dolomite in terms of conversion of WCO, the yield of pyrolysis oil as well as the selectivity of fuel range. Then the reusability and regeneration studies of KSi catalyst was also performed. The catalyst was reused for 5 runs of reactions without any treatment. Regenerability of catalyst took place after the completion of the reusability cycles of five runs. The catalyst was collected at the end of the cycle and regenerated to remove the coke deposited on the catalyst and activate the catalyst before being reused back for five consecutive runs by calcination in the furnace. After one complete reusability cycle (five consecutive runs), the remaining reactants at the bottom of the flask (coke + catalyst) were collected in the crucible and heated in the furnace at a temperature of 700 ℃, a heating rate of 10 ℃/min for 4 h with nitrogen flows continuously. After that, the catalyst was heated again at a temperature of 1000 ℃ with a temperature rate of 10 ℃/min for 4 h at open airflow, the catalyst was denoted as KSi-RG1. The reaction was continued with 5 runs of reactions by using KSi-RG1. The spent catalyst was again regenerated at the same conditions and denoted as KSi-RG2. The reaction was further continued for another 5 runs of reactions by using KSi-RG2. The product for each experiment was collected once the system cooled down at room temperature and analysed by using GC–MS. The conversion of WCO was measured by using Equation (1) [16]: (1) % C o n v e r s i o n = m a s s o f W C O - m a s s o f c o k e massofWCO x 100 % The liquid products were quantitatively analysed by Shimadzu GC-14B equipped with ZB-5MS column (30 m length × 0.25 mm inner diameter × 0.25 μm film thickness) in a split mode. The oven temperature was set to hold at 40 °C for 3 min, the ramping rate at 7 °C/min to reach 300 °C and holding temperature at 300 °C for 5 min. The injector temperature was set at 250 °C and He gas flow rate of 3.0 ml/min. The liquid product was dissolved with n-hexane. A different class of compounds especially heavy hydrocarbon (C21-C24) and oxygenated compounds obtained were recognized using the National Institute of Standards and Testing (NIST) library [37]. The GC–MS results were measured using a peak area normalization method based on peak area percentages of the identified components. The hydrocarbon yield (Y) was calculated using Equation (2) [38–39]. (2) Y = ∑ a i + ∑ a j ∑ a k Where ai  = Area of alkene (C8-C20), aj  = Area of alkane (C8-C20), ak  = Area of the product.The yield of chemical groups in the oxygenated compound (Z) was determined using Equation (3) [38]. (3) Z product = a o ∑ a t x 100 % Where Zproduct  = Yield of organic compound (%), ao  = Area of the desired organic compound, and at  = Total area of organic compounds.The percentage removal of the oxygenated compound was calculated by using Equation (4) [34,36]. (4) Σ A r e a o f o x y g e n a t e d c o m p o u n d ( W C O ) - Σ A r e a o f o x y g e n a t e d c o m p o u n d ( P O ) Σ A r e a o f o x y g e n a t e d c o m p o u n d o f W C O x 100 % As shown in Fig. 3 , the different physical appearance of solid base catalyst of KSi in fresh, spent, and regenerated catalyst after first reusability cycles were observed. The fresh catalyst is shown in Fig. 3a was uncalcined as it can be directly used for the deoxygenation reaction of WCO. As mentioned in Table 1 , the diameter pore size of the fresh catalyst was 15.93 nm which classified it as a mesoporous catalyst with high base properties. As shown in Fig. 3a solid base catalyst which was before use in the deoxygenation of WCO has white colour. As reported by Ooi et al. [40], silicon dioxide (SiO2) support catalyst is a refractory oxide with a white and colourless appearance. However, after the reaction took place, the colour of the solid base catalyst was changed into black as shown in Fig. 3b. A black coke produced during deoxygenation reaction was deposited on the surface of the catalyst causing them to be a deactivated catalyst. Thus, catalytic activity was reduced and conversion of WCO could not occur on the surface of the active site after 5 times reusability without any treatment. However, after regenerating at 1000 °C in open airflow, the colour was changed into grey as shown in Fig. 3c. Based on Bayramoğlu et al. [41], calcination at high temperatures is a more appropriate and simple method to regenerate the activity of catalysts with no waste solvent or solid waste product formation. The deposited coke, organic compound and adsorbed impurities on the spent catalyst were expected to be removed or diminished. This was directly restoring the activity of the catalyst for the next regeneration. Table 1 shows the physicochemical properties of the base solid catalyst of K2O/SiO2 or KSi and calcined dolomite (CD). The structure of the base solid catalyst could be observed based on the XRD pattern in terms of particles arrangement especially in internal structure. As shown in Table 1, the KSi catalyst was the amorphous structured catalyst with lacked an ordered internal structure of particles and were randomly arranged which is seen in the SEM image result (Fig. 4 a. The crystallize size of K2O was present at 2θ = 28.27° with JCPDS File:00–025-0626. This showed that SiO2 was able to interact well with potassium by enhancing the crystallinity and stability of the synthesized catalyst. While calcined dolomite catalyst was the crystalline catalyst with the intense peak of CaO (54.39°) and MgO (62.39°). The particles of the crystalline catalyst were in an ordered structure of circular shape particles and a repeating pattern. The surface area of the CD catalyst was bigger than the KSi catalyst as mentioned in Table 1. This would provide a more active site for the substrate molecules to bind and undergo a chemical reaction onto a CD catalyst. Through the pore diameter mentioned in Table 1, it showed that KSi was a mesoporous catalyst (2–50 nm) and CD was a macroporous catalyst (>50 nm). Based on Hafriz et al. [42] and Asyikin-Mijan et al. [37] the mesoporous catalyst was an efficient catalyst to promote better decarboxylation (DCO2), decarbonylation (DCO) or hydrodeoxygenation (HDO) pathways. The basicity strength of the catalyst was measured by desorption of carbon dioxide in TPD-CO2 analysis. The CD catalyst desorbed 2872.73 μ atom/g of CO2 at a temperature of 722 ℃ as compared to the KSi catalyst, which desorbed 303.39 μ atom/g of CO2 at the temperature of 656 ℃. The CO2 desorption peak at temperature > 500 ℃ showed a relatively high basicity strength in TPD-CO2 analysis. The high basicity properties of catalysts should improve the quality of green fuel produced by lowering the oxygen content through a deoxygenation reaction [34].Scanning electron microscopy (SEM) analysis was performed to examine the morphology of the solid base catalyst (KSi) after 2 times of regeneration process with 15 times reusability and CD catalyst as shown in Fig. 4. Fresh KSi catalyst has gross and non-uniform particle structure as compared to calcined dolomite catalyst having a smooth spherical particle structure with the wide surface of catalyst as shown in Fig. 4 (a & b). The CD catalyst has an obvious wide pore diameter, in line with XRD analysis which classified it as a macroporous catalyst. However, the macroporous structure catalyst should tend to undergo surface poisoning or pore filling of carbon in coke accumulation due to this big porosity of catalyst. The SEM-EDX result of fresh catalyst, regenerated KSi catalyst after first and second reusability cycle was shown in Fig. 4 (a,c,d). It could be observed clearly in SEM images that the morphology of the fresh catalyst is different from the regenerated catalyst after the first and second reusability cycles. In a fresh catalyst, there are many tiny aggregated particles on the surface of the fresh catalyst than in a regenerated catalyst. The calcination at a high temperature not only influenced the morphology and structure of the regenerated catalysts, but it also effects on the porosity of the catalyst which became bigger as observed clearly on 1st regenerated (KSi-RG1) catalyst surface ( Fig. 4c). Coke formation and K2O deactivation after 5 times reusability cycle should occur on the catalyst surface resulted in a reduction in the performance of catalyst by lowering the yield of liquid product and affecting the selectivity of oil produced (fuel range). Based on Hafriz et al. [36] , coke accumulation might be occurred by two reaction pathways: polymerization of aromatic hydrocarbon (Equation 5) or condensation of WCO (Equation 6). While K2O deactivation could have resulted from the reaction of the K2O phase with H2O (Equation 7 and 8) or CO2 (Equation 9) to form potassium carbonate (K2CO3) which is H2O was generated from decarbonylation reaction and CO2 has been produced from decarboxylation reaction. The calcination at 1000 ℃ should burn off the coke and unwanted impurities on the catalyst surface and recovered the K2O phase formation under calcination at 1000 ℃.Coke accumulation: (5) Polymerization: Cn-Hn (Aromatics) → Coke (6) Condensation: WCO → Coke K2O deactivation: (7) K2O + H2O → 2KOH (8) KOH + CO2 → K2CO3 + H2O (9) K2O + CO2 → K2CO3 Besides that, it also could be seen that the pore structure and pore shape of the regenerated catalyst were reduced even though the overall structure was almost identical. Thus, it slightly reduced the catalytic activity as the pore structure and pore shape decreased. These might be due to the destruction of brittle basicity catalyst structure morphology after 5 times reusability at the high calcination temperature. According to Fatimah et al. [43], the texture and surface design of catalyst is important to ensure catalytic activity.The spectrum of x-ray energy at each position on the samples versus counts is evaluated to determine the relative elemental concentration for each element presence present in the fresh, regenerated solid base catalyst (KSi) and calcined dolomite (CD) catalysts as shown in Table 2 . Table 2 shows that the main element present in KSi was oxygen and silicon. It was in line with metal oxide composition analysis, which showed that SiO2 was the main composition of KSi which act as backbone or support catalyst. Based on Ooi et al. [40], SiO2 support is stable and has been used as good support for metal loaded catalyst in the catalytic reaction. With the advantages in low owing of coke deposition to the very weak acidity present in SiO2 and suitability to be used in hydrogen-free or low hydrogen pressure system, SiO2 support catalyst is more focused in these studies. While the element potassium, K presence showed that potassium oxide, K2O was successfully doped onto SiO2 surface and generated multiphase oxide catalyst with high base properties. K2O was expected to create a more active site, which plays a vital role in the deoxygenation of WCO to produce green diesel. In CD catalysts as shown in Table 2, CaO and MgO were the main composition presence after CaMg(CO3)2 was decomposed at calcination around 900 ℃ [27]. Based on Lin et al. [44] and Asikin-Mijan et al. [37], CaO can be acted as a deoxygenation catalyst that could absorb more CO2 either in liquid or solid phase in order to remove oxygen compound via the decarboxylation-decarbonylation mechanism. While MgO in CD catalyst can enhance the strength of dolomite particle structure and at the same time, it plays the important role in increasing the oil quality by reducing oxygen levels [36,45]. As compared to fresh and regenerated KSi catalysts, the impurities element could be observed after 1st and 2nd regeneration of KSi-RG1 and KSi-RG2 catalysts due to the reusability. The elemental analysis in Table 2 revealed that in regenerated catalyst there are foreign elements (Fe, Mg and S) deposited on the surface of the catalyst. This shows that the loss of activity might be due to certain elements deposited on the solid base catalyst during repeated use in the deoxygenation process. Nevertheless, solid base catalyst still could be considered as an effective regenerate catalyst for the cracking of waste cooking oil.Based on Fig. 5 , the conversion of WCO achieved the highest in the deoxygenation reaction using KSi (62.3 %) as a catalyst followed by thermal deoxygenation (without catalyst) (40.0 %) and CD catalyst (38.3 %). The highest conversion of WCO using KSi catalyst was due to lower accumulation of coke after reaction by measured using mass balance. The main composition of KSi catalyst which is SiO2 has been proven to play a bigger role in lowering the coke deposition due to the weak acidity surface present in SiO2. Besides that, many researchers have admitted that SiO2 as a supported catalyst is suitable for use in an H2-free reaction system. These advantages were also contributed to yielding the highest pyrolysis oil (51.8 %) when using KSi as a catalyst. The coke formation was found to be the highest (61.7 %) when using CD as a catalyst which might be due to macroporous structure catalyst resulted in increasing the possibility of coke deposition onto CD catalyst surface. Coke deposition caused active site coverage and pore-mouth blockage making the core pore network inaccessible to reactants (WCO). This will slightly reduce the catalytic activity of CD catalyst in deoxygenation reaction. Meanwhile, in thermal deoxygenation of WCO (without catalyst), the yield of pyrolysis oil was very low with only 1.9 %. Thus, this finding showed that the yield of pyrolysis oil from catalytic deoxygenation of WCO (with the presence of a catalyst) was enhanced as compared to thermal deoxygenation (without the presence of a catalyst). As reported by Faten et al. [46] and Hafriz et al. [27] , catalytic deoxygenation was faster and more selective than thermal deoxygenation which allows the reaction to happen under mild reaction conditions and hence maximizing the production of pyrolysis oil. As reported by Dewajani et al. [47], in catalytic cracking reactions, the oxygenated compounds which come from thermal cracking diffuse into the pores of catalysts and react with protons in the active site through several reaction pathways such as dehydration, decarboxylation, decarbonylation, and oligomerization. This will result in a low yield of gas products. Furthermore, the thermal deoxygenation of WCO was favoured to produce the highest gas yield as compared to catalytic deoxygenation in line with the result shown in Fig. 5. In order to increase the yield of pyrolysis oil, the catalyst presence is more favoured in the deoxygenation reaction due to the high surface area of the catalyst leading to increased active sites. The active site would provide more adsorption interaction of the reactant molecules into the surface of the catalyst. As reported by Horacek et al. [48], a high surface area of catalyst will improve the diffusion of reagent resulting in a higher deoxygenation degree at the same time it inhibits the breaking of the C–C bonds. In addition, Ooi et al. [40] found out that the use of catalysts especially metal oxide catalysts with good chemical stability and high activity is the main support for deoxygenation reaction to occur.The amount of hydrocarbon and oxygenated compound presence in pyrolysis oil generated via deoxygenation of WCO using different catalysts are shown in Fig. 6 . The higher yield of hydrocarbon compound generated in pyrolysis oil could be observed using calcined dolomite (CD) catalyst (88.3 %) followed by thermal deoxygenation (without catalyst) (82.2 %) and KSi catalyst (52.5 %). Calcined dolomite catalyst yielded high hydrocarbon compound was due to the presence of CaO and MgO, which can remove oxygen compounds by absorbing more CO2 and enhancing the strength of dolomite particle structure in deoxygenation reaction, respectively. Thus, the use of calcined dolomite was a favour to the formation of hydrocarbon which implied higher deoxygenation activity as in line with higher percentage removal of oxygenated compound (85.9 %) followed by in thermal deoxygenation (78.5 %) and KSi catalyst (42.6 %) as shown in Fig. 5. Chiam & Tye [49] and Kwon et al. [50] reported that the removal of oxygen contents inside used palm oil is required to improve biofuels quality: increase energy density, reduce viscosity, and stabilize biofuels. The undesired oxygen content especially insoluble impairs engine performance such as fouling of injector, plugging of the fuel filter, sticking of the ring, and formation of deposit in the engine. As reported by Sani et al. [51], at high temperatures, decomposition of oil to hydrocarbon is favoured without the presence of a catalyst which is in line with the result of hydrocarbon produced via thermal deoxygenation in Fig. 6. This is an indication that temperature plays a vital role in the production of hydrocarbon from waste cooking oil. However, in the deoxygenation of WCO using KSi catalyst, the hydrocarbon compound produced was lower due to the small pore diameter of the catalyst which inhibits the reaction pathway of deoxygenation to occur. As reported by Hafriz et al. [36], a catalyst with a larger pore structure promotes additional deoxygenation pathways including hydrodeoxygenation (HDO), decarbonylation (DCO), and decarboxylation (DCO2) in the deoxygenation of WCO. However, the largest porosity structure of catalyst (macroporous) contributed to the high deposition of coke formation in line with the result in Fig. 5 for calcined dolomite (coke; 61.7 %) and the mesoporous structure of catalyst would be the best catalyst for deoxygenation reaction pathway to occur.The composition profile of pyrolysis oil with different catalysts used was summarised as shown in Fig. 7 . The figure shows that the main composition of hydrocarbon generated in pyrolysis oil was alkenes and alkanes with the same trend for thermal deoxygenation and different catalyst used. This proves that decarbonylation-decarboxylation of deoxygenation reaction was the main reaction pathway involved. Based on Hafriz et al. [36] and Faten et al. [46] , the decarbonylation reaction would remove the carbonyl group in WCO to produce alkenes by releasing CO and H2O (Equation 10) and the decarboxylation reaction would eliminate the carbonyl group by releasing CO2 to produce alkanes (Equation 11).Decarbonylation reaction: (10) WCO (triglyceride) → Alkenes + CO + H2O Decarboxylation reaction: (11) WCO (triglyceride) → Alkanes + CO2 As shown in Fig. 7, oxygenated intermediates compounds such as aldehyde, ketones, alcohol, and carboxylic acid were also detected, and this result is in agreement with the work found out by Azman et al. [39]. The carboxylic acid was found to be higher in pyrolysis oil generated using KSi catalyst and it could be due to the low basicity properties of KSi catalyst as compared to CD catalyst. The presence of high carboxylic acid in pyrolysis oil could increase the acid number of oil and it would affect cold-flow properties such as cold filter plugging point and freezing point as well as reduce the heating value of the fuel at the same time [52]. Figure 8 shows the proposed reaction pathways for deoxygenation of WCO (comprised of oleic acid, palmitic acid and stearic acid glycidyl ester) using KSi catalyst. The role of the catalyst will efficiently remove oxygen molecules while decrease carbon loss to retain the quality of the final fuel product which is the ideal catalytic deoxygenation pathway. The deoxygenation reaction pathways were established based on the distribution of hydrocarbon product using GC–MS analysis with the majority of compound present in pyrolysis oil consisting of pentadecane, C15H32 − 7.44 %, heptadecane, C17H34 − 6.66 %, heptadecane, C17H36 − 2.48 % and the rest were C8-C14 around 24.59 %. It showed that the KSi catalyst will provide a mild extent of deoxygenation efficiency and high selectivity towards the formation of C15 and C17 products. As shown in Fig. 8, the presence of catalyst will facilitate the oxygen removal pathway via the formation of stearic acid (C18:0) from the hydrogenation of oleic acid by the addition of in-situ hydrogen (from water gas shift reaction) and hydrolysis of the carboxylic ester. Carboxylic acid will undergo oxygen removal via deoxygenation reaction in two pathways which are decarboxylation and decarbonylation reaction. In decarboxylation of stearic acid, heptadecane with C17:0 will be formed by releasing CO2. Meanwhile, in the decarbonylation of stearic acid, n-heptadecene with C17:1 will be generated by releasing CO and producing H2O as by-products. Heptadecene could react with in-situ H2 in hydrogenation reaction to form thermodynamically stable heptadecane. This can be observed by intensified peak appearing in chromatography for pyrolysis oil using KSi catalyst. Besides that, palmitic acid presence in WCO composition will simultaneously undergo this deoxygenation reaction. The carbonylation of palmitic acid will produce pentadecene (C15:1) and H2O as a by-product by releasing CO. Pentadecane with C15:0 could be generated via hydrogenation reaction of pentadecane by addition of in-situ H2. Besides that, pentadecane could be formed through the decarboxylation of palmitic acid by releasing CO2. Furthermore, light components (C8-C14) with 24.59 % were also detected in the GC–MS analysis due to mild cracking (C–C cleavage) to the intermediate product n-heptadecene and pentadecene.The composition of green fuel was grouped by the carbon number of the gasoline, kerosene, and diesel fraction as similar to petroleum products and is presented in Fig. 9 . The selectivity of hydrocarbon products was influenced by catalyst/without catalyst used. As shown in Fig. 9, the selectivity of hydrocarbon products generated using CD catalyst was towards gasoline carbon number which composes of light hydrocarbon with carbon number range C8-C12. The gasoline selectivity was due to the high basicity properties of CD catalyst, and this was in agreement with the finding by Asikin-Mijan et al. [37], Ca-based catalysed reaction with high basicity properties rendered higher selectivity towards gasoline products (C8-C12). While using thermal deoxygenation and KSi catalyst, the selectivity of hydrocarbon product generated was towards diesel range (C13-C24). The kerosene selectivity towards hydrocarbon fraction was composed of a mixture of alkanes, cycloalkanes, and aromatic compounds.The conversion of WCO is presented in Fig. 10 . The percentage conversion was measured as a function of coke retaining amount. The trend for a conversion percentage of WCO was KSi > KSi-RG1 > KSi-RG2. The conversion of WCO using KSi slightly decreased from 62.3 % in the 1st run to 41.8 % in the 5th run. This result was well correlated due to the loss of deoxygenation activity and deposition of coke formation on the active site of the catalyst. This was in line with the principle mentioned by Guisnet and Magnoux [53]; coke may affect catalyst activity in two ways: through active site coverage (poisoning) and pore blockages (active sites rendered inaccessible to reactants). After the first regeneration, the conversion of WCO in the first run was 58.6 %, a slight decrement in the conversion as compared to the 1st run using KSi catalyst. The conversion of WCO was continuously decreased as the KSi-RG1 was reused until the 5th run. The conversion obtained in the 1st run using KSi-RG2 was 53.0 %, lower as compared to the 1st run using KSi-RG1. The reduction in conversion of WCO using KSi-RG1 and KSi-RG2 was possibly due to the formation of coke that had not been totally removed from the surface of the catalyst in the regeneration process leading to the reduction of the active sites of the catalyst. As reported by Shao et al. [54], restoring the activity of a permanently deactivated catalyst with standard regeneration procedures is difficult. However, catalyst deactivation due to coke deposition is usually reversible, and the coke can easily be removed by oxidation with air (O2). Due to this, lesser conversion of WCO was produced using the regenerated catalyst as could be observed in Fig. 10. Figure 11 shows the product distribution in terms of liquid and estimation gas measured based on mass balance via deoxygenation of WCO. The liquid product obtained from the deoxygenation process had two layers which are an aqueous phase and an oil phase. The aqueous phase contained primarily water and a few organic compounds whereas the oil phase contained primarily organic compounds such as acids, esters, phenols, aldehydes, alcohols, ketones, ethers and hydrocarbons [3]. The trend of liquid yield according to the reusability of KSi catalyst was 1st run > 2nd run > 3rd run > 4th run > 5th run for the first consecutive cycle. Massive loss in liquid product yield after the first cycle of KSi was possibly due to pore blockage resulted in unable to provide enough space for catalytic activity. The pores were blocked by deposition of coke, organic compound and adsorbed impurities which could be observed through the physical appearance of regenerated KSi catalyst change from white to grey colour (Fig. 3) and the coke accumulation coated on morphology structure of regenerated KSi catalyst as shown in Fig. 4 (c & d). In addition, the presence of other elements and metal oxides can be observed after regeneration as shown in Table 2, proving that the porosity of regenerated KSi catalyst has been blocked. Furthermore, the yield showed a slight decrement throughout the five consecutive runs owing to sufficient active site on the surface of the catalyst. The yield of liquid product was boosted after the first (46.0 %) and the second regeneration (40.3 %). The pattern for cycle performance was quite similar after each regeneration. The thermal regeneration step can treat the deactivated KSi catalyst under high temperature by restoring the catalytic activity of regenerated KSi catalyst as shown in pyrolysis oil yielded. Figure 12 represents the value of hydrocarbon and oxygenated compounds in the liquid product. Hydrocarbon dominated over oxygenated compounds for each run and regeneration cycle. The highest yield of hydrocarbon (74.0 %) and the lowest oxygenated compound (26.1 %) were achieved during the 3rd run when using fresh KSi. However, the yield of hydrocarbon gave slightly improved after the first regeneration and achieved a high yield (74.9 %) during the 2nd run of KSi-RG1. On the contrary, the yield of hydrocarbon after the second regeneration (KSi-RG2) was reduced as compared to the hydrocarbon yield produced using KSi and KSi-RG1. The reduction in hydrocarbon yield could happen due to voids in the pores of the catalyst being partially filled with coke formation which contributes to the serious deactivation caused by difficult diffusion of reactants and products. The highest yield of hydrocarbon (67.5 %) and the lowest oxygenated compound (32.0 %) were attained during the 3rd run of KSi-RG2. This result was in agreement Li et al. [55] with that the composition and properties of hydrocracking oils obtained in this work were similar to the cracking and catalytic cracking oils obtained from vegetable oils in the presence of a basic catalyst. The minimum amount of oxygenated compound in a diesel engine, would lead to a good performance in power output and decrease fuel consumption [56]. Figure 13 shows the chemical constituents in the liquid product detected by GC–MS analysis for KSi, KSi-RG1, and KSi-RG2. As shown in Fig. 13a, the main chemical group found were alkanes (49.31 %), carboxylic acid (35.70 %), alkenes (27.69 %), alcohol (10.12 %), and cycloalkane (7.42 %) with the occurrence of smaller concentration of diene, cycloalkane, aromatic, ketone, ester and, etc. GC–MS analysis showed that oxygenated compounds that appear in the oil were mostly long-chain of carboxylic acid and ester. The significant amount of carboxylic acids in the liquid product was related to the high amount of fatty acids content in the waste cooking oil [14]. Nevertheless, the chemical group determined for KSi-RG1, and KSi-RG2 catalyst showed a similar trend. These findings are comparable with other researchers [57], where there were a high number of alkane groups in the pyrolysis product which used CaO as a catalyst for deoxygenation of WCO.As shown in Fig. 14 , the main component in the liquid product was in the diesel range, followed by the kerosene and gasoline carbon range. The main component of the products was always hydrocarbon, non-polar oxygenates and organic acids [58]. The diesel produced using fresh KSi catalyst was in the range of 75.82 % to 81.02%. After the first regeneration, the highest biodiesel yield was attained (86.72 %) during the 5th run of the KSi-RG1 catalyst. Meanwhile, the highest diesel produced in the second regeneration was 84.5% during the 3rd run using KSi-RG2 catalyst. On the other hand, the yield of kerosene in the liquid product was also obtained as high as 89.05% in the 4th run using KSi, 75.49% in the 4th run using KSi-RG1 and 81.88% in the 5th run of KSi-RG2. Moreover, the highest yield of gasoline 24.0% was acquired during the 3rd run using KSi, 26.5% during the 4th run using KSi-RG1, and 26.7% during the 5th run using KSi-RG2 catalyst. From this finding, the selectivity of the diesel was improved after the first regeneration i.e., during the second cycle, possibly owing to the retaining coke that can modify the base property and pore dimension of the KSi catalyst. This finding was in agreement with Shao et al. [59],the retaining coke decorated the catalyst pore, which changed the product distribution and product selectivity of hydrocarbons in catalytic pyrolysis of biomass-derivates. Table 3 shows the comparison studies in reusability and regenerability of various catalysts in the deoxygenation of WCO. The reusability and regenerability of catalysts are important steps as they can cut the cost of large-scale green diesel production. As reported by Shitao et al. [60], with the lengthening of the regeneration cycle of SBA-15@MgO@Zn catalyst to be used in deoxygenation of WCO, the production of liquid biofuels decreased while the yield of coke increased. This could be because, during the numerous catalytic cracking and regeneration cycles, the active sites of the SBA-15@MgO@Zn catalyst were gradually depleted. However, the average yield of hydrocarbon after 3 runs and 3 times of regeneration was still good which is around 68.5 %. As compared to a study conducted by Pacheco et al. [61], the yield of hydrocarbon products was low around 58.4 % but the selectivity of green diesel was higher which is 90 % by using thermal regeneration for Pd/SBA‑15 catalyst. The authors found out that, the incomplete elimination of the oxygen can be explained by different factors that include catalyst deactivation due to impurities present in the WCO and the formation of carbonaceous deposits during the reaction also has a deactivating effect. It was parallel with finding based on this research through observation of foreign elements (Fe, Mg and S) and coke deposited on the surface of the regenerated KSi-RG1 and KSi-RG2 after thermal regeneration. However, KSi (K2O/SiO2) has shown good stability through 15 runs with 2 times of regeneration by giving 65.8 % in an average yield of hydrocarbon product with 79.4 % in average of green diesel selectivity. In the reusability and regeneration study of activated carbon (AC) based catalyst conducted by Alsultan et al. [38] and Wan et al. [62] as shown in Table 3, the spent catalyst has been reactivated by using the hexane washing technique. The regenerated AC catalyst was washed and reused for the next reaction cycle under the set-up reaction conditions. As reported by Wan et al. [62] through TGA analysis, the spent NiLa/AC catalyst exhibited the presence of hard coke around 26 wt%, which decomposed between 330 ℃ to 750 ℃; this showed that the catalyst surface was covered by coke formation after the deoxygenation reaction. This finding agreed with EDX analysis, which showed a remarkable increment in the carbon content in the NiLa/AC spent catalyst after 6 deoxygenation runs, hence, it strongly implies that the spent NiLa/AC catalyst was masked by the coke. However, based on Alsultan et al. [38], there is another factor that affects the deactivation of catalyst which is metal leaching. The leaching of the Ca2+ and La3+ ions from active sites of CaO-La2O3/AC catalyst into the reaction medium might cause a decrease in the catalytic activity throughout the 6th run. Through ICP-AES analysis results, the researchers found that the metal content in the liquid product from 1st to 6th runs catalyst was found to gradually increase from 0.27 to 0.70 ppm and 0.0 to 0.10 ppm for Ca2+ and La3+, respectively. The dissolution of Ca2+ and La3+ metal species in the liquid product during deoxygenation reaction simultaneously led to the reduction of deoxygenation reaction by decreasing the catalytic activity of CaO-La2O3/AC catalyst.This comparative study demonstrated that K2O/SiO2 (KSi) can be used as a catalyst in the deoxygenation of WCO producing a high yield of pyrolysis oil (51.8 %); however, a low amount of hydrocarbon compound (52.5 %) was obtained as compared to thermal deoxygenation and calcined dolomite catalyst. The dispersion of a high amount of K2O into SiO2 as a supported catalyst was recommended in a future study to improve the basic properties of synthesized K2O/SiO2 catalyst by eliminating more oxygen content in pyrolysis oil generated. The catalytic pyrolysis of WCO using KSi catalyst rendered higher selectivity towards diesel products (C13-C24) as contrasted to that of kerosene (mixture of alkanes, cycloalkanes and aromatic compounds). Reusability study indicated that KSi catalyst showed tremendous thermal stability and marvellous reactivity after 15 times of reusability and 2 times regeneration with the average yield of hydrocarbon and diesel yield of 32.49 % and 79.36 %, respectively. The gradual reduction shown in deoxygenation activity was mostly due to the coke depositions on the surface of the KSi catalyst. R.S.R.M. Hafriz: Conceptualization, Data curation, Methodology, Visualization, Writing – review & editing. I. Nor Shafizah: Conceptualization, Data curation, Methodology, Visualization, Writing – review & editing. N.A. Arifin: Validation, Writing – review & editing. A.H. Maisarah: Data curation, Formal analysis, Investigation, Methodology, Writing – original draft. A. Salmiaton: Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Supervision, Validation, Writing – review & editing. A.H. Shamsuddin: Project administration, Resource.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.The authors acknowledge the financial support from the Ministry of Higher Education of Malaysia for Fundamental Research Grant Scheme (FRGS/11/TK/UPM/02) and AAIBE Chair of Renewable Energy Grant No. 201801 KETTHA for support this research publication.

In this work, the comparative study on the performance of K2O/SiO2 and calcined dolomite catalysts was conducted via deoxygenation of waste cooking oil (WCO). KSi catalyst has the potential as a deoxygenation catalyst due to mesoporous structure catalyst with high base properties which enhance oxygen removal. The result found that K2O/SiO2 catalyst could be used in the deoxygenation of WCO generating a high yield of pyrolysis oil as compared to thermal deoxygenation and calcined dolomite catalyst. Besides that, the reusability and regeneration of the K2O/SiO2 catalyst were evaluated in the deoxygenation reaction using WCO as a feedstock. Five consecutive runs of reusability test and three successive cycles with two regenerations were performed. The reusability and followed by regeneration tests were conducted at conditions: 30 min of reaction time, 390 ± 5 °C reaction temperature and 150 cm3/min of N2 flow rate. The liquid products obtained from each cycle were analyzed by GC–MS. The deoxygenation of WCO using K2O/SiO2 catalyst rendered higher selectivity towards diesel products (C13-C24). The K2O/SiO2 catalyst presented a good performance in reusability and regenerability test with only ∼19.3–22.4% dropped in pyrolysis oil yield after 5 consecutive runs and only ∼11.20–13.23% drop in diesel yield after regeneration. The results showed that K2O/SiO2 catalyst has tremendous stability in the deoxygenation of WCO into green diesel and could be the alternative deoxygenation base catalyst for the deoxygenation process.

The research of biomass feedstock-derived chemicals being converted into fuels has received much interest in recent years [1–5]. It contributes to the establishing green and sustainable development in carbon–neutral technology, which aids in alleviation of energy shortages and environmental pollution. Among these chemicals, renewable levulinic acid (LA) is considered a prominent platform chemical for biofuels and chemical synthesis [6]. LA is formed during the acid-catalyzed hydrolysis of lignocellulosic biomass. It is composed of a saturated keto and acid functional group that is selectively converted to valeric acid (VA) and γ-valerolactone (GVL) [7–9]. These two chemicals are used as intermediates in the synthesis of valeric biofuels. In fine chemical synthesis, GVL could be used as a green solvent, food additive, and also precursor in the synthesis of gasoline fuels [10,11]. LA is converted into a variety of products, the most notable of which is the hydrogenation of LA to VA synthesis, which has received lot a of attention [12]. VA is identified as one of the most prominent chemical intermediates in fine chemical industries because of its outstanding physicochemical properties such as low toxicity and unique fuel characteristics. It is also widely used in cosmetics production, perfumes, ester-type lubricants in aviation turbine oils, vinyl stabilizers, refrigerants, fine chemicals, and pharmaceuticals [13–16]. Furthermore, VA is easily converted into 5-nonane, which can be used as a solvent in the paint and resin industries as well as a precursor in biofuel synthesis in the vapour-phase of traditional ketonization technology [16–18]. Valeric esters synthesized from VA, which is regarded as a promising constituent in advanced biofuels [19,20]. As a result of the wide range of applications for VA and its derivatives, a cost-performance competitive manufacturing process in biorefinery industries is required. Currently, in industry, VA is produced by combining syngas and 1-butene in an oxo process, followed by air oxidation of the aldehyde. VA markets are currently under intense pressure. This is primarily because Rhodium (Rh) based catalysts are expensive, homogeneous and soluble in the reaction mixture. The recovery and separation of the catalyst from reaction residues is the main disadvantage. In addition, side-products are forming, the majority of which are branched products, alcohols, and alkanes [21].LA and/or GVL to VA synthesis involves several reaction steps and formation of intermediates [12]. In the past few years, the challenges of different steps have been merged into a single-step and bifunctional catalysts have shown great interest in this reaction. It is known that multiple tandem reactions of LA to VA synthesis require metallic centres with a reducing property and active acid site [22–25]. Several recent research studies have recently focused on the noble and non-noble metal supported catalysts investigated and various reaction conditions in LA to VA synthesis (Table S1) [12]. Zhou et al. reported that palladium on carbon (Pd/C) and hafnium trifluoromethanesulfonate (Hf(OTf)4) bifunctional catalysts tested at 150 °C and 5.0 MPa of H2, it is showed 92% of VA selectivity with 100% LA conversion [26]. Furthermore, noble-metal supported catalysts are highly active for hydrogenation reactions [12,22,23]. Still, their cost is high, they have harsh reaction conditions, they deactivate quickly and they suffer from inactivity due to coke formation and deposition on their surface, which prevents their technical implementation in near future bio-refineries [22–28].Similarly, alternative catalysts discovered that low-cost non-noble bifunctional catalysts performed well in LA hydrogenation [24,25]. The long-term security of experiments necessitates the use of a fixed bed continuous flow reactor. Liu et al. conducted the conversion of GVL to valeric esters synthesis over Cu/ZrO2 catalyst. The catalyst showed GVL conversion of 85.4% and pentylvalerate selectivity of 98% obtained at 230 °C and 1.5 MPa H2 [27]. Afterward, Chan-Thaw et al. examined the Cu/SiO2-ZrO2 catalyst for the synthesis of valeric esters from GVL. The catalyst showed 59% selectivity of ethyl valerate, and 69% of GVL conversion was obtained at 250 °C, 1 MPa H2 [29]. Jiang et al. demonstrated that liquid-phase hydrogenation of GVL into ethyl valerate over Ni/La-Y in ethanol solvent at 200 °C under 30 bar H2 conditions [30]. These GVL-to-VA conversions were good; but,theyfrequently required severalstepsorstarted withaGVLresource.To simplify the procedure, converting LA into VA in a single-pot without isolating the GVL intermediate would be preferable.Karanwal et al. demonstrated that liquid-phase conversion of LA to VA synthesis by Nb-Cu/Zr on a silica catalyst at mild reaction conditions at 150 °C and 3 MPa H2 for 4 h in an aqueous medium, produced 99.8% of VA [24]. However, the drawbacks of liquid-phase hydrogenation of LA in terms of high-pressure demand, waste-discharge and leaching leads to both a decline in activity and pollution of the product purification. These are all anticipated to be alleviated by seeking the vapour-phase hydrogenation of LA under ambient and/or moderate reaction conditions. Even so, there aren’t many reports on single-pot vapour-phase synthesis of VA from LA in a fixed bed continuous flow-reactor. Vapour-phase catalytic hydrogenation process is much simpler, more efficient and environmentally benign compared to the liquid-phase hydrogenation process [12].Due to its well-ordered mesoporous structure, wide surface area, thermal stability, and pore volume, Santa Barbara Amorphous-15 (SBA-15) supports metals and provides the necessary acidic sites for the conversion of LA. Weaker acid sites of SBA-15 prevent catalyst deactivation by coke formation, which may extend the catalyst lifetime and support acidity, as well as both Lewis and Brønsted acidity promotes LA adsorption and VA production while also controlling metal leaching and carbonaceous species on the catalyst surface which are more important for developing a prominent, stable bifunctional catalysts [12]. In this work, we investigated the LA to VA synthesis in a single step with the design of non-noble metal-supported catalysts with different metal-doped mesoporous SBA-15 catalysts (M/mesoSBA-15; M = Zr, Nb, and Ti), which are mainly focused on Zr/mesoSBA-15, Nb/mesoSBA-15, and Ti/mesoSBA-15 catalysts with tunable Brønsted/Lewis acidity. Based on the above results, we proposed a plausible reaction pathway for LA conversion to VA synthesis, the following steps are required [12,20]. These steps are follows: i) hydro-cyclization (intramolecular dehydration) of LA to GVL over a hydrogenation catalyst and acid sites via α-angilica lactone intermediate; ii) opening of the GVL ring on acid sites while GVL formed to pentenoic acid; and (iii) further reduction to produce VA, these as shown in Scheme 1 . The catalysts were characterized in more detail to identify their excellent catalytic performance. The physicochemical parameters of the catalysts were analyzed by X-ray diffraction (XRD), transmission electron microscopy (TEM), UV–vis diffuse reflectance spectroscopy (UV-DRS), temperature programmed desorptionof ammonia (NH3-TPD), and brunner-emmett-teller(BET) surface area, pyridine adsorption followed by FT-IR (Py-FTIR) analysis, etc. Finally, the enhanced catalytic activity studies have been tuned by the acidity and surface-active sites of the catalysts.Ammonium niobite (V) oxalate.hydrate was purchased from Alfa Aesar; levulinic acid, tetraethyl orthosilicate (TEOS), HCl, zirconium (IV) nitrate, titanium (IV) nitrate, cetyltrimethylammonium bromide (CTAB), Pluronic P123 were purchased from the sigma–aldrich company. All chemicals are directly used as precursors.The SBA-15 materials were synthesized by a hydrothermal method from the previously reported publications [31,32]. In the first step, 2 g of P123 was dissolved in a well-mixed solution of 15 mL of H2O and 45 mL of 2 M HCl with continuous stirring, then 0.2 g of CTAB and 5.9 g of TEOS were added. The ratio would be 1 TEOS: 0.02 CTAB: 3.1 HCl: 115 H2O: 0.012: metal ratio of 15 wt% Polymer. The resultant solution was transferred into an autoclave and put in an oven at 100 °C for 24 h. Finally, the resulting precipitate was washed with deionized water and an ethanol solution. The obtained powder was calcined at 550 °C for 5 h.Furthermore, different metal-doped Nb/mesoSBA-15, Ti/mesoSBA-15, and Zr/mesoSBA-15 (M/Si ratio of 15 wt%) samples were synthesized by simple wetness impregnation method. The solid was then oven-dried (100 °C, 12 h) and calcined (500 °C, 5 h).The XRD experiments were conducted on a Rigaku miniflex diffractometer with nickel filtered Cu Kα radiation with 40 kV and 20 mA from 2θ = 2 to 65°, the scanning rate is 2°min−1.The composition of the samples was performed on inductively coupled plasma-atomic emission spectrometry (ICP-AES) analysis by Perkin-Elmer Optima 3300 DV equipment. Prior to the analysis, the catalysts were dissolved in HCl, HF, H3PO4, and HNO3; these were well mixed for 2 h in a microwave oven. The subsequent solutions were diluted with demineralized water.The SBA-15 and various metal-doped SBA-15 materials textural properties were conducted by N2-adsorption–desorption isotherms at 77 K, in a Micromeritics ASAP 2020 system. First the catalysts were evacuated at 200 °C for 3 h. The BET values are calculated by using the BET equation.A JEOL 2010 instrument was used for TEM analysis and conducting at 200 kV. The powder catalysts were dissolved in an EtOH solution through sonication for 10 min, then diffused on copper grids. The sample holder was put into the microscope column. The UV–vis analysis was performed on a GBC UV–Visible Cintra instrument.The acidity of the catalysts was investigated by NH3-TPD analysis using the 2920 Micromeritics device. The catalysts were pre-treated with pure He gas for 50 mL at 200 °C for 1 h and cooled to a lower temperature. The catalyst was in-situ reduced by 5% H2-Ar 40 mL at 250 °C/2h and then treated with pure He gas for 50 mL at the same temperature for 30 min. After that the NH3 absorption was carried out over 10% NH3-He gas for 75 mL at 80 °C/1h, then purged at 120 °C/2h for the physisorption of NH3. The TPD run was conducted from 120 to 750 °C and the outlet amount of NH3 was analyzed by thermal conductive detector (TCD). Py-FTIR analysis was performed at 250 °C for 2 h under air flow, then the reduction was conducted at the same temperature. Later, the pyridine adsorption was conducted at 110 °C, subsequently, the samples were cool down to room temperature. These samples were mixed with KBr and ground, and pressed into tablets. The spectra recorded by GC-FT-IR Nicolet 670 and KBr have taken the background spectrum.The vapor phase hydrogenation of LA to VA synthesis was tested in a continuous fixed bed with a stainless-steel tubular reactor (SSTR) at 265 °C with 0.2 MPa of H2 pressure. About 0.4 g of the catalyst was placed in the reactor. At 300 °C, the sample was first reduced with 50 mL of H2 flow. The reactant was pumped along with H2 flow at the desired reaction temperature. The catalytic performance of the catalyst was conducted at various reaction conditions. Hourly, the products are collected using an ice-water trap, and they are analyzed using GC with an DB-wax column. Before the GC analysis, the samples were diluted with methanol and analyzed by GC–MS with an HP-1MS column. The amount of total carbon content was analyzed using the CHNS technique. The balance is about 98%. The below equation obtains the conversion of LA and selectivity. C o n v e r s i o n % = L A m o l e s i n - m o l e s o f L A o u t L A m o l e s i n × 100 S e l e c t i v i t y % = m o l e s o f o n e p r o d u c t m o l e s o f a l l p r o d u c t s × 100 The chemical composition of the different metal-doped SBA-15 catalysts was investigated by ICP-AES analysis and presented in Table 1 . The chemical composition of the theoretical and experimental results is very close to ∼ 15 wt%. However, the actual doping of the metal content is lower than expected.The N2-physisorption isotherms and pore size distribution of SBA-15 of various metal-doped SBA-15 catalysts are shown in Fig. 1 a & b, and the textural parameters of the catalysts are presented in Table 1. All of the catalysts exhibited type IV isotherms with H1 hysteresis, which is assigned to the mesoporous nature owing to cylindrical pores and also a uniform pore size according to IUPAC classification [33]. These results revealed that all samples contained the hexagonal arrangement with a mesoporous structure even after doping the metals into the SBA-15. Parent SBA-15 showed a surface area value of 830 m2g−1 and a pore diameter of 6.3 nm, whereas the surface area of Zr, Ti, and Nb-doped mesoporous SBA-15 catalysts was decreased when compared to SBA-15. This is mainly due to increases in pore wall thickness and a decrease in pore volume, and also metal species could exist outside the skeleton, which led to a decrease in pore diameter, pore volume, and BET surface area values. These results are consistent with previous publications and also confirmed by XRD analysis [34,35]. Among those catalysts, the Zr/mesoSBA-15 catalyst exhibited a higher BET surface area and a 0.82 pore volume, while the Ti/mesoSBA-15 catalyst showed a lower surface area of 711 m2.g-1and a pore volume of about 1.03. The above results indicate that high surface area will be beneficial for the higher activity of LA to VA synthesis.The parent SBA-15 and metal-doped SBA-15 materials of XRD patterns are presented in Fig. 1c. The parent SBA-15 exhibited the peaks at 0.8°, 1.59o, and 1.8° of 2θ values, which are assigned to the (100), (110), and (200) reflections, and the corresponding d spacing is (100) = 11 nm. These findings might be confirmed by the fact that SBA-15 showed a highly ordered 2D-hexagonal symmetry of the mesoporous nature of SBA-15 (p6mm), which can be assigned to the excellent long-range order formed within the SBA-15 [31,36]. Fig. 1c shows three characteristic peaks of hexagonal arrangements observed, but the intensity was decreased with the doping of Zr, Nb, and Ti. These results can also be co-related with the BET surface values, with the doping of metal promoters, the BET surface values decreased. These are presented in Table 1. The Ti/mesoSBA-15 catalyst only exhibited the plane (100) with lower intensity, this can be assigned to the support structure being lost, which can influence the BET surface area [37].The parent SBA-15 and those metal-doped SBA-15 materials of XRD patterns are presented in Fig. S1. The catalysts displayed the broad peak at 2θ = 20-30°, were assigned to the amorphous silica. No diffraction peaks were observed from the metallic oxide species, indicating that the metallic phases had better dispersion on the external and internal surfaces of SBA-15, whose size should be below the XRD limits [38]. The mesoSBA-15 exhibited the peak at 2θ = ∼22°, ascribed to the amorphous-silica nature [39].TEM images evaluated the morphology of catalysts and the distribution of particle size of various metal-doped SBA-15 materials are presented in Fig. 2 . The metal-doped SBA-15 samples were uniformly distributed on the mesoporous SBA-15. The average particle size of the catalyst was found to be 2.3 to 12 nm. According to the statistical results on more than 250 nanoparticles, the particle size was found to be 2.3 ± 1.0 nm for the Zr/mesoSBA-15 catalyst. The modification with Nb and Ti to form an inappropriate structure led to a slight increase in the particle size to 4 to 11 nm for the Nb/mesoSBA-15 and Ti/mesoSBA-15 catalyst (Fig. 2). Based on the above results reveal that well dispersion with the smaller particle size in the Zr/mesoSBA-15 catalyst is favourable for the higher catalytic activity.This spectra is useful for the identification of isolated transitional metal ions and clustered transition oxides with local coordination environments and electronic states. Here different metal-doped SBA-15 catalysts were investigated. The SBA-15 like silica materials showed that the absorption bands were found to be 200–400 nm in Fig. 3 a and the adsorption bands gave more information about the metal atoms of the surrounding environment in the SBA-15. The band at 200–240 nm is attributed to ligand–metal charge transfer (L-M CT) transition that occurred. This band gave direct evidence of the metal-doped SBA-15 framework silica. The catalysts showed a sharp peak at 200 nm. The Ti/mesoSBA-15 catalysts sample exhibited two bands at high intensity, the band at 186–230 nm, is ascribed to M−O charge transfers in the TiO2 ion and the lower-intensity second band between 250 and 310 nm was assigned to the Ti atoms doped into the SBA-15 framework and occupied in the tetrahedral position. Here the bulk phase of isolated TiO2 might be negligible [40,41]. The Zr/mesoSBA-15 samples exhibited two bands at 206 and 318 nm. The 206 nm band is ascribed to the O2-Zr4+ (L-M CT) in tetrahedral coordination inside the SBA-15 [42,43]. The 318 band is assigned to the O-M CT cation in the octahedral coordination sphere. In the case of Nb/mesoSBA-15, it is also exhibited two types; a small intensity absorption band at 210–240 nm observed, which is an O2–-Nb+4 CT band and a second absorption band was observed at 240–380 nm, this band indicates that the Nb species is in the form of tetrahedral coordination site [44–46]. Finally, UV-DRS spectra concluded that the metals are successfully doped into the mesoporous structure of SBA-15.The total amount of acidic sites on the catalyst's surface was determined using NH3-TPD analysis. The small size of the NH3 molecule has a strong basic nature and is stable at higher temperatures. NH3 is a probe molecule in the NH3-TPD analysis. The hydrogenation reaction is an acid catalyzed reaction, and it will proceed through the active and stable acid sites of the catalysts. The surface acidity plays a prominent role in this hydrogenation activity and selectivity. The amount of total acidity was determined and is presented in Table 2 . From Fig. 3b, the catalysts exhibited three desorption peaks, at 100–200 °C the characteristic nature of weak acid strength, and the peak in the range of 220–400 °C is attributed to the moderate acidic sites, whereas the peak at ≥ 400 °C is ascribed to the strong acidic sites [47–49]. The total amount of acidity was increased drastically after metal-doped on the SBA-15 support, which might be the replacement of protons by Zr (IV) ions, Ti (IV), and Nb (IV). Therefore, the metal ions interact with the acidic sites along with the SBA-15. Among those catalysts, the Zr/mesoSBA-15 catalyst exhibited a higher amount of acidic strength, these acidic sites are beneficial for the higher activity of LA to VA. As follows the order of the total acidity of the catalysts: Zr/mesoSBA-15 > Ti/mesoSBA-15 > Nb/mesoSBA-15 > mesoSBA-15. The above results demonstrated that the acidic sites of materials were primarily attributed to the nature of metal-doped in mesoSBA-15 catalysts.From the NH3-TPD we could not be distinguished the Lewis and Brønsted acid sites. Py-FTIR of SBA-15 and different metal-doped SBA-15 catalysts are presented in Fig. 3c. The Lewis and Brønsted acid sites could be discovered with the help of this analysis. From the previous literature, pyridine molecule adsorbed via hydrogen-bonding interactions with surface OH groups and protons (H) might be combined with the Lewis (L) and Brønsted (B) acidic surface sites. The bands observed at 1450 and 1610 cm−1 are assigned to L acidic sites [38,49]. The bands observed at 1540–1548 cm−1 , are assigned to the B acidic sites [50]. The band at 1490–1500 cm−1, is ascribed to the combination of both B + L acidic sites. We observed that metal doping to SBA-15 increased the L acidic sites while slightly decreasing the B acidic sites. The metal (0) acts as an electron acceptor in reducing conditions, this can be beneficial for the L acidic site generation, including all the samples, the Zr/mesoSBA-15 catalyst showed the higher amount of L acidic sites and a lower amount of B acidic sites when compared to other catalysts and these results are presented in Table 3 .We have examined the vapor phase synthesis of VA from LA over different metal-doped SBA-15 catalysts (Table 4 ). The parent SBA-15 exhibited very low selectivity of VA about 0.9% but it showed GVL with 18% selectivity, this is mainly due to the SBA-15 containing slightly Brønsted and Lewis acidic sites. The Ti-doped SBA-15 showed 49% selectivity of VA and the Nb/SBA-15 sample exhibited 38% selectivity of VA. In contrast, the Zr-doped mesoporous SBA-15 catalyst showed 68% selectivity of VA. The best results can be obtained from the Zr-doped SBA-15 this might be because of the high diffusivity of Zr species on the support surface and the accessibility of the strong acidic sites, which are beneficial for the higher selectivity of VA [22]. Recently, several groups reported improvements to zirconia-supported catalysts, which can improve the acidity and accessibility of active sites on the catalyst surface, enhancing the activity in biofuel synthesis [51,52]. The characterization results revealed that the increase of VA selectivity with the doping of Zr species into SBA-15 can be improved by the formation of Lewis acid sites confirmed by Py-FTIR and NH3-TPD studies.In this reaction, the temperature will have a crucial impact on product selectivity’s. The influence of temperature was studied over LA to VA synthesis on the Zr/mesoSBA-15 catalyst from 175 to 295 °C was studied and the obtained results are displayed in Fig. 4 a. It illustrates that the conversion of LA increased from 28% to 98% as the temperature rose from 175 °C to 295 °C, this is mainly ascribed to the increase in the number of active sites and these sites activate the reaction at higher temperatures. Whereas VA selectivity was increased to 68% at 275 °C when a further rise in temperature, the VA selectivity would be slightly decreased. It demonstrated that the reaction temperature above 275 °C does not improve the selectivity toward VA. Hence, the best reaction temperature for this reaction is 275 °C.The influence of WHSV on the performance of the Zr-doped SBA-15 catalyst for LA to VA synthesis is shown in Fig. S2. The pure LA solution, with a flow rate of 0.5 to 3.0 mL/h with WHSV = 1.425 to 8.55 h−1. With the increase of feed flow rate, the conversion of LA and VA selectivity decreased from 91% to 74% and 68% to 52%; the selectivity of GVL, alkyl levulinates (AL), and 2-methyltetrahydrofuran (MTHF) could be enhanced. The lower activity at higher WHSV is mainly attributed to the insufficient residence time in the reactor on the catalyst surface. The above results revealed that higher selectivity of VA was obtained at lower WHSV. This is primarily owing to the longer residence time on the catalyst surface. The calculation of WHSV is done by using the following formula. W e i g h t h o u r l y s p a c e v e l o c i t y ( W H S V ) = M a s s o f f l o w S a m p l e w e i g h t ( g ) Whereas,. M a s s o f f l o w = F e e d f l o w r a t e d e n s i t y o f f e e d f l o w The influence of the catalyst weight on LA to VA synthesis over the Zr/mesoSBA-15 catalyst is shown in Fig. 4b. The weight of the catalyst ranged from 0.2 to 0.6 g, and the conversion of LA increased from 55% to 91%. As the amount of catalyst increased, the selectivity of VA reaches 68% at 0.6 g of catalyst weight. However, a further increase of catalyst amount (0.6 to 1 g) leads to a slight drop in VA selectivity. This might be due to the lower space velocity of active sites compared to reactants molecules. The optimal catalyst weight to perform the LA hydrogenation to VA synthesis is 0.6 g of catalyst.Zr/mesoSBA-15 catalyst was conducted for 52 h of TOS, and it is shown in Fig. 4c. An increase of reaction time from 1 h to 11 h, the catalyst showed 91% conversion of LA and the 68% selectivity toward VA. However, the LA conversion and the VA selectivity could be stable upto 17 h of reaction time and then decreased with TOS. These results reveal that the deactivation is mainly due to the blocking of the active acidic sites by coke formation on the catalyst surface, these can be proved by the CHNS technique as shown in Table 4.In summary, the different metal-doped mesoporous SBA-15 catalysts were designed successfully and then evaluated in a continuous fixed-bed vapor phase reaction of LA into VA synthesis. The different reaction parameters of the effect of temperature, WHSV, and catalysts weights were studied. The well distributed Zr species on SBA-15 will strengthen the strong acidic sites. Under the optimized reaction conditions, the Zr/mesoSBA-15 catalyst showed the highest 68% selectivity toward VA at 91% conversion of LA. The control of the acidity of strong acidic sites will be beneficial for the LA hydrogenation and GVL ring-opening to form VA. The catalyst also showed stable catalytic activity with a TOS of 52 h. VA selectivity has decreased mainly due to the coke formation over the catalyst's active sites. The results of XRD, TEM, NH3-TPD, and Py-FTIR characterizations are well evidenced for better catalytic activity. Ramyakrishna Pothu: Conceptualization, Methodology, Investigation, Supervision, Writing – original draft. Harisekhar Mitta: Data curation. Rajender Boddula: Conceptualization, Methodology, Investigation, Supervision, Writing – original draft. Putrakumar Balla: Data curation. Raveendra Gundeboyina: Data curation. Vijayanand Perugopu: Data curation. Jianmin Ma: Conceptualization, Methodology, Investigation, Supervision, Writing – original draft.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.Ramyakrishna Pothu (Award number: 2017SLJ018367) acknowledges the China Scholarship Council (CSC), China, for the financial support.Supplementary data to this article can be found online at https://doi.org/10.1016/j.mset.2022.09.006.The following are the Supplementary data to this article: Supplementary data 1

Chemoselective hydrogenation of biomass platform molecules into value-added chemicals and fuels is essential for the exploitation of biomass, and SBA-15 based metal catalysts with hydrogenation centers and acid sites seem promising in this regard. Valeric acid (VA) is the most important platform molecule for valeric biofuels and value-added chemicals production. The main issue with using such bifunctional catalysts for biomass conversion is maintaining the catalyst's stability in the liquid phase under harsh conditions. In-addition, direct one-pot selective hydrogenation of levulinic acid (LA) into VA synthesis is challenging due to its complex reaction conditions involved. Herein, we design a bifunctional mesoporous catalysts (SBA-15 mesoporous material doped with various metals Nb, Ti, and Zr) investigated for this reaction under the vapour phase. Different instrumental approaches were used to examine the structure, phase composition, morphology, and surface elemental analyses of catalysts as-prepared. Among those catalysts, Zr-doped mesoporous SBA-15 catalyst showed the 91% conversion of LA and the 68% selectivity toward VA and promising stability in a 52 h time on-stream run. Metal dispersion inside the SBA-15 and their surface acidity (sufficient number of acid sites and surface-active metal oxide species) and higher surface area are beneficial for the selectivity of VA. This work offers a highly-efficient bifunctional catalyst for selective hydrogenation of biomass feedstocks.

With the depletion and increasing environmental impacts of the traditional fuels, such as coal and petroleum products, the emerging global challenge in both energy and environment fields has prompted intensive research on renewable energy-conversion and energy-storage systems, such as fuel cells, electrolyzers, and supercapacitors, as well as various batteries. 1 Electrocatalysts for oxygen reduction reaction (ORR), hydrogen evolution reaction (HER), oxygen evolution reaction (OER), carbon dioxide reduction reaction (CO2RR), and nitrogen reduction reaction (NRR) are at the core of some of these energy-conversion and energy-storage systems. In the past decades, various materials, including noble metals, transition metals, and metal-free carbons, have been explored as electrocatalysts, aiming to achieve the high activity, durability, and selectivity for the reactions mentioned above. In general, catalysts are categorized into homogeneous and heterogeneous catalysts. Ninety percent of the current chemical industry processes use heterogeneous catalysts; 2 however, their low atom utilization efficiency brings crucial disadvantages in terms of activity and cost. 3 In comparison, homogeneous molecular catalysts offer uniform active sites, high atom utilization efficiency, and convenient structural tunability, but they have relatively low durability and recyclability. Therefore, it is desirable to develop novel catalysts that can combine the merits of both homogeneous and heterogeneous catalysts.Among various novel catalysts, single-atom catalysts (SACs) have aroused great interest due to their intriguing features, such as high atom utilization efficiency, 4–6 low-coordination environments of single-atom centers, 7 unique quantum size effects, 8 and tunable metal-support interactions. 9 For example, the atomic dispersion of metal atoms in SACs brings the maximum atom utilization efficiency for metals, and their quantum size effects create discrete energy-level distribution and distinctive HOMO-LUMO gaps. 10 Besides, strong interactions between active metal centers and adjacent coordinating atoms may enhance the catalytic activity, selectivity, and durability of metal centers. 11 Zhang and co-workers reported the first SAC in 2011 by anchoring Pt atoms on the surface of iron oxide nanocrystallites (denoted as Pt1/FeO x ), which exhibits excellent stability and high catalytic activity for CO oxidation. 12 Their density functional theory (DFT) calculations and experimental data reveal that the partially vacant 5d orbitals of the positively charged, high-valent Pt atoms in Pt1/FeO x decrease the CO adsorption energy and the activation barrier of CO oxidation.Subsequently, different substrates have been used in SACs fabrication, including metal oxides, metal hydroxides, perovskites, zeolites, metal-organic frameworks (MOFs), and carbon materials. Among them, carbon materials attract researchers' intense attention because of their excellent conductivity and porous structures, which are beneficial for the electron/mass transfer of the reaction intermediates. Also, the structural diversity and designability of carbon materials (e.g., amorphous carbon, graphite, and diamond) enable the study of their precise structure-performance correlations, which is critical for both the in-depth understanding of the catalytic mechanism and the design of high-performance catalysts. In the last few years, many carbon-supported SACs (CS-SACs) have been reported, including Fe, Co, Ni, Ru, Ir, Au, Rh, Pd, and Ag supported on carbon nanotubes, graphene, carbon nanofibers, porous carbon, carbon nanosheets, and many other carbon nanomaterials. They have shown decent catalytic activities for many energy-conversion reactions including ORR, HER, OER, CO2RR, or NRR. 13–16 Some CS-SACs also show bifunctional or multi-functional catalytic activities, such as ORR and OER in metal-air batteries, HER, and OER in water electrolyzers. 17–20 Tremendous advances have been achieved, making it possible for CS-SACs to overtake traditional metal particles-based catalysts in the race to the renewable energy marketplace. Apart from the applications in electrocatalysis, the recent breakthroughs of CS-SACs also enable them to catalyze various vital reactions in energy-storage devices, such as supercapacitors, rechargeable lithium batteries, and sodium-sulfur batteries. A timely review of the rapidly growing field is highly desirable. This article aims to provide a concise and critical updated overview of recent progress by summarizing important work reported within the past 3 years on CS-SACs for essential reactions involved in energy conversion and storage and to present critical issues governing the fundamental understanding of reaction mechanisms and design strategies of CS-SACs. Through such a comprehensive review, we hope to increase the knowledge of CS-SACs significantly and attract a broad range of scientific communities to their future developments.Various CS-SACs have been reported in the last few years, and they have been used as electrocatalysts in different energy-conversion and energy-storage applications. However, the discussions of their advantages and disadvantages are scattered throughout the existing literature. Here, we provide a summary of their unique benefits and shortcomings related to material designs, superior catalytic activities, and practical applications.In general, CS-SACs are composed of carbon frameworks and metal dopants, which offer numerous opportunities to fabricate a variety of functionalized catalysts for satisfying the requirements in catalysis and energy-conversion devices. Carbon nanomaterials themselves formed by strong covalent bonding between carbon atoms possess the unique physicochemical properties, including controllable dimensions, ease of accessibility, high surface area, excellent conductivity, controllable porosity, and abundant defects. Take the structural controllability as an example: the wide variety of nanoscale physical dimensions, such as zero-dimensional (0D) graphene quantum dots, one-dimensional (1D) carbon nanotubes, two-dimensional (2D) graphene, and three-dimensional (3D) nano-diamond, provides an ideal platform to design high-performance catalysts. Also, their adjustable chemical compositions that may contain single, double, or ternary metal species, or no metal species, can be used for activity regulation. The high surface area, hierarchical pore structure, and excellent conductivity of carbon materials are also the essential preconditions for mass/electron transfer and accessibility between reactants and active sites, which are the incomparable superiorities for pure metal oxides, and hydroxides or perovskites. Moreover, the potential strong metal-support interactions between well-dispersed single metal atoms and carbon substrates can not only limit the aggregation of metal atoms but also tailor the geometric structures and electronic configurations of active catalytic sites. The activities of different coordination configurations of central atoms to a specific electrochemical reaction are of importance to the catalyst design, which may even change the reaction pathways via different electron transfer numbers. 21 For example, Yao et al. unveiled the nature of M-N-C (M = Mn, Fe, Co, Ni, and Cu) as catalytic centers that can change the ORR pathways spanning from 1e to 4e transfer processes. 22 DFT calculations show that the electronic structures of atomic Co can be finely tuned by bonding to diverse types of transition metals in the form of M-N4 motifs, resulting in different target reaction pathways in the ORR. N species can play two significant roles in boosting the intrinsic activities of Co-SACs while N coordinated with Co can manipulate the reactivity by modification of electronic distribution. 23 In another study, high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) and X-ray absorption spectroscopy (XAS) have revealed that strong metal-S interactions on S-doped carbon substrates can effectively suppress the aggregation of metal atoms, resulting in the formation of single metal atom catalytic sites. 24 , 25 Besides, CS-SACs with homogeneously dispersed catalytic active sites and tunable physicochemical properties are the ideal candidates for the improvement of catalytic activity for many vital reactions (e.g., ORR, HER, and CO2RR) and related mechanistic understanding. For example, a series of transition-metal catalysts (i.e., Co, Cu, and Fe) with single atomic sites embedded in hollow N-doped carbon were synthesized to catalyze ORR. 26 In particular, the Co-based catalyst showed a catalytic activity similar to that of Pt/C catalyst in an acidic electrolyte. The experimental and computational analyses revealed that the superior catalytic activity for ORR could be ascribed to the dramatically enhanced hydrogenation of OH∗ species at single Co atom sites. 27 Besides, a series of single-atom transition metals (i.e., Co, Ni, and W) supported on N-doped graphene recently were fabricated as electrocatalysts for HER, exhibiting a low overpotential (E j  = 10 mA cm−2) and long-term durability. 28 Their DFT calculations and experimental data demonstrate a correlation between catalytic activity and the electronic structures of single atoms. The energy states of active valence d z 2 orbitals and the corresponding antibonding states play a decisive role in the catalytic activity for HER. In particular, the Co catalyst has antibonding state orbitals neither empty nor fully filled, which offer optimum adsorption strength toward H adsorption than other catalysts. 28 Thanks to the tunability of active sites in CS-SACs, a series of Ni/Fe/bimetallic CS-SACs were successfully prepared by various synthetic strategies for some important electrochemical reactions. 29 , 30 For instance, the Ni porphyrin-based covalent triazine framework containing atomically dispersed NiN4 centers was recently fabricated by a pyrolysis process and further used for electrocatalytic CO2 reduction. The as-prepared Ni-based CS-SACs exhibit superior activities for CO2RR with high faradic efficiency (FE) of 97% for converting CO2 to CO at the potential of −0.9 V versus reversible hydrogen electrode (RHE). 29 Apart from catalytic activities, CS-SACs have the potential to offer lower cost and better stability over many noble metal-based catalysts for practical applications. 31 CS-SACs would be more cost-effective because they provide the maximal atom utilization efficiency (100%) and potentially low manufacturing cost. For example, many recently developed CS-SACs have already shown superior catalytic performances to Pt/C in the fuel cells using alkaline electrolytes, 32 , 33 and some CS-SACs even outperform the state-of-the-art non-precious metal catalysts in acidic polymer electrolyte membrane fuel cells (PEMFCs). 34 , 35 Therefore, compared with conventional nanoparticles (NPs) or bulk catalysts, CS-SACs are expected to become a more efficient alternative for boosting the practical applications of advanced conversion reactions and renewable energy devices.Although CS-SACs have the potential to play essential roles in many energy-conversion and energy-storage applications, there are some critical challenges associated with their synthesis and stabilization. First, carbon-based substrates in M-N-C electrocatalysts would be electro-oxidized to CO2 and CO under high electrical potentials (e.g., >0.9 VRHE) during fuel cell tests, which results in carbon surface destruction and ruins active catalytic sites, such as FeN x C y species. 20 , 36 , 37 Second, a significant issue for the synthesis of CS-SACs is the difficulty in stabilizing isolated metal atoms on carbon supports without compromising their catalytic activities, especially at high-temperature conditions or under harsh reaction conditions. This can be ascribed to the higher mobility of individual metal atoms. They are more likely to aggregate into large particles because their surface energies are higher than corresponding metal clusters and NPs. To efficiently stabilize isolated metal atoms, the metal mass loading of current CS-SACs is kept low to minimize the agglomeration of metal atoms. Third, due to the diversity and complexity of their structures and components, it is challenging to pinpoint exact atom arrangements in CS-SACs and identify the origin of catalytic activity (see more discussion in the final section of this review). Therefore, various rational design strategies are needed for further development of CS-SACs.CS-SACs can offer many desirable benefits, as already described. However, due to the high specific surface energy of single atoms in CS-SACs, individual atoms may easily migrate and agglomerate into larger clusters. A key focus of the synthesis of CS-SACs is to achieve highly dispersed single atoms with high stability and density. In this section, the synthetic strategies of CS-SACs will be critically reviewed and classified into two types based on the integration mode of components, namely bottom-up and top-down methods.Several bottom-up methods have been used to synthesize CS-SACs, such as atomic layer deposition (ALD), wet chemistry synthesis, and ball milling. 38 , 39 They usually first create surface defects or heteroatom-induced coordination sites in carbon substrates, followed by absorption and reduction of metal precursors to obtain isolated single metal atoms. The modification of pristine carbon surfaces prevents the fast aggregation of metal precursors and metal atoms, which plays a vital role in CS-SAC preparation. Surface defects can change electronic structures of carbon surfaces, creating trapping centers for metal atoms. Furthermore, rationally designed coordinate sites using N, O, or S atoms can precisely anchor metal precursors via chelation.ALD allows precisely controlled deposition of diverse structures on carbon substrates, which enables well-dispersed single atoms by the self-limiting reaction mechanism. It is a typical bottom-up technique with excellent controllability in synthesizing CS-SACs. For example, functional groups on the graphene surface can react with a Pt precursor ligand, and the reaction ends spontaneously once all surface functional groups are occupied, resulting in a monoatomic Pt layer (Figure 1 A). 5 , 40 Furthermore, the mass loading of Pt and the size of Pt clusters can be controlled by altering ALD cycles.Subsequently, many efforts have been devoted to demonstrating that surface functional groups on carbon substrates are critical in obtaining high-performance CS-SACs. 41 Typically, pristine graphene will be oxidized in strong acid, followed by a thermal deoxygenation process to create anchoring sites or defects. For example, the single Pd atom catalysts were fabricated by using ALD and used for selective hydrogenation of 1,3-butadiene. 5 Pd metal precursors can be tightly anchored on surface phenol groups, forming -O-Pd-hafc surface species (Figures 1B and 1C), after which they are transformed into -O-Pd surface species (Figure 1D). A transmission electron microscopy (TEM) image (Figure 1E) shows that abundant single Pd atoms are well dispersed on the carbon substrate. Although many studies have demonstrated the excellent controllability of ALD in synthesizing CS-SACs, its employment in commercial catalyst production is often limited by its high cost and limited scalability. 42 In comparison with ALD, wet chemistry synthesis methods (including co-precipitation and impregnation methods) are currently regarded as the most promising also widely adopted preparation routes because of their low cost, ease of operation, and excellent potential for mass production. They are also capable of providing high mass loading and excellent dispersity of single atoms through precisely controlling the nucleation and growth rate of metal species. In general, wet chemistry methods include three steps: adsorption of metal precursors, calcination, and activation.Two most common strategies used for the adsorption of metal precursors are the coordination strategy via heteroatoms (e.g., N, O, and S) in carbon substrates and an atom confinement strategy by designing structurally complex and highly defective carbon substrates. 43–45 For example, N-doped porous carbons were used as the substrates for synthesizing Au-based carbon-supported catalysts (denoted as AuSAs-NDPCs) by a coordination strategy (Figure 2 A). Compared with the N-free carbon, the loading amount and dispersity of single Au atoms are improved significantly. X-ray photoelectron spectroscopy (XPS) results (Figure 2B) show that the electron density of N atoms in AuSAs-NDPCs is much lower, indicating the existence of strong interactions between Au and N. 44 Alternatively, a DFT calculation predicts that single atoms prefer to adsorb on the confined interfaces between reduced graphene oxide (rGO) and activated carbon (Figures 2E and 2F). 39 Inspired by the prediction, the atom confinement strategy can be used to synthesize a Pd-based SAC. Pd atoms are confined in a double-shelled hollow carbon substrate with rGO as the inner shell and amorphous carbon (AC) as the outer shell. TEM images show that abundant Pd single atoms with an average size of approximately 3–4 Å are well dispersed on the boundary of rGO@AC (Figures 2C and 2D). 39 However, due to the limited boundary area in rGO, the mass loading of Pd single atoms is low at 0.290 wt %.Other than the above methods, several other approaches have also been used to increase the mass loading of single atoms. For example, Zhang et al. 47 proposed a facile and inexpensive defect trapping strategy to synthesize a highly stable, atomically dispersed Ni catalyst on defective graphene (DG) (A-Ni@DG) with a high Ni loading (1.24 wt %) by an incipient wetness impregnation method with subsequent acid leaching treatment. HAADF-STEM images demonstrate that the atomic Ni species (aNi) are uniformly trapped in the defects of graphene to form an aNi@defect catalyst. More recently, a universal three-step defect trapping approach was developed for preparing various Co-N4-x C x (x = 0–4) catalysts with excellent controllability. 23 These studies demonstrate a promising defect trapping approach for producing highly active and stable CS-SACs with well-controlled coordination.Besides, the impregnation-adsorption method was developed to synthesize the Pt-based CS-SAC (denoted as Pt1/hCNC) with a higher Pt mass loading up to 2.92 wt %. 46 The high mass loading is attributed to both the physical confinement of [PtCl6]2− precursors with a size of ∼0.5 nm in micropores (∼0.6 nm) and strong interactions between [PtCl6]2− anions and N dopants. The adsorption energy of [PtCl6]2− on N-doped hCNC is 4.6 or 4.9 eV, indicating the formation of stable [C x (NH)2]2+[PtCl6]2− ion pairs via the strong electrostatic interaction (Figure 2G). This strategy has also been extended to other precious metals, such as Pt, Au, and Ir. More recently, a photochemical solid-phase reduction strategy was used to synthesize well-isolated Pt atoms on N-doped porous carbon (denoted as Pt1/NPC). 48 In brief, the PtCl6 2− ions are directly reduced under UV light irradiation and then preferentially deposited on NPC without additional treatments. Pt atoms are well dispersed on the carbon surface without clusters or NPs with a mass loading of 3.8 wt %. Apart from the metal loading amount, the efficiency of this synthetic method is another concern. In this regard, microwave heating as an ultrafast method has an advantage for the synthesis of CS-SACs. By the microwave heating reduction of GO, a series of monodispersed atomic transition metals (for example, Co, Ni, and Cu) can be obtained within 2 s 49 Moreover, it not only reduces the reaction time but also largely suppresses side reactions, dramatically improving the synthesis efficiency and catalyst yield.Ball milling can cut and reconstruct chemical bonds of materials/molecules as a powerful technique to develop CS-SACs. The moving balls apply their kinetic energy to the materials, leading to single metal atoms embedded in the surface of carbon substrates. Recently, an Fe-based SAC was synthesized by ball-milling the mixture of iron phthalocyanine (FePc) and graphene nanosheets. 50 During the milling process, the external macrocyclic component of the FePc precursor could be destroyed and thus form FeN4 centers. Subsequently, the FeN4 centers are trapped on defect sites of graphene nanosheets while adjacent C atoms are reconstructed. The maximum mass loading of Fe is around ∼2.7 wt % without agglomeration. The method has been successfully extended to synthesize other single metal atom catalysts supported on graphene, including Mn, Co, Ni, and Cu. 51 Top-down methods offer several advantages, including simplification in the synthesis procedures, low cost, and environmental friendliness, in comparison with the bottom-up methods discussed above. They can effectively control different types of metal atoms and their loading rates by regulating synthesis parameters, such as temperature, concentration of metal precursors, and gas flow rate. Moreover, abundant defects and unsaturated sites can be created once bulk structures are downsized into clusters or even individual atoms, which is essential for improving the catalytic performances of CS-SACs. We discuss the top-down methods in three subsections: pyrolysis of organic precursors, solid-state reactions, and electrochemical activation.Pyrolysis of organic precursors has long been considered as an efficient synthesis method for developing CS-SACs owing to its simplicity, high metal mass loading, and decent scalability. Also, it can be readily carried out across different laboratories without sophisticated instrumentation. The precursors include polymers, MOFs, and many other organic compounds containing metal species. Among them, various polymers are the most widely used precursors to form carbon substrates for CS-SACs by high-temperature pyrolysis because of the following reasons: (1) a variety of dopants, such as O, N, and S, can be precisely introduced into the framework of polymers, which lead to specific heteroatom dopants in the carbon substrates for metal ion adsorption; (2) the structures of carbon substrates, such as pore size, surface area, and bulk geometry, can be easily controlled by the polymer structures, solvents, and pyrolysis conditions; (3) polymers can also create carbon substrates with distinct atomic arrangements, which facilitates the study of structure-activity relationships in CS-SACs. For example, Zhao et al. 52 polymerized dicyandiamide and Ni(II) acetylacetonate to synthesize a Ni-based SAC (denoted as NiSA-N-CNT) supported on tubular carbon structures for CO2RR, which has an ultrahigh Ni mass loading of 20.3 wt %. Figure 3 A shows that the single Ni atoms coordinate with four nearest N atoms. The ratio of the Ni precursor and dicyandiamide can significantly affect the morphology of the obtained samples. The formation of tubular substrates is ascribed to the stress-introduced rolling of Ni-containing graphitic carbon nitride layers under high temperatures and electron beam irradiation conditions. In another study, Cheng et al. 53 polymerized hemin porcine (HP) with Fe(III) acetylacetonate to synthesize an Fe-based SAC supported on graphene-like 2D carbon nanosheets for fuel cells in an acidic electrolyte. The HP precursors are assembled into 2D carbon nanosheets, resulting in a high surface area of 670.8 m2 g−1. To increase the defect concentration in the substrate, the dual (S and N)-doped polymers have been designed and used for the Fe-based SAC synthesis. The dual-doped polymers can create more anchoring sites and optimize the interactions between carbon substrates and metal atoms, leading to outstanding catalytic performances. 33 , 54 Other than using polymer precursors, MOFs, with controllable metal sites, periodic structural units, and adjustable pores, are regarded as ideal templates to produce carbon substrates by a one-step pyrolysis. The unique structural characteristics of MOFs provide advantages over simple mixtures of metal ions and organic precursors. First, MOFs can generate carbon substrates, heteroatoms, and metal atoms simultaneously by a one-step pyrolysis. Organic ligands and metal atoms in MOFs play different roles in resulting CS-SACs. Second, organic linkers in MOFs can serve as coordination sites to adsorb targeted metal ions. Alternatively, MOFs can encapsulate metal precursors into their porous framework by spatial confinement. 58 Both coordination sites and pore structures of MOFs can be tailored to increase the density of active catalytic sites. In addition, steric repulsion forces related to pore structures of MOFs can also be tuned to influence the adsorption of reaction intermediates. 59 For example, bimetallic (Co and Zn) MOFs were used to synthesize a Co-based CS-SAC supported on N-doped carbon (denoted as CoSA/N-C). 55 Zn2+ sites in MOFs could be considered as “diluting sites” to prevent the agglomeration of Co2+ during pyrolysis (Figure 3B). X-ray absorption fine structure (XAFS) results show that there is no Co–Co bond, suggesting that there are only isolated Co atoms. Similar but different, Fe-TCPP (TCPP = tetrakis(4-carboxyphenyl) porphyrin) with a 3D network was used to synthesize an Fe-based SAC supported N-doped carbon for ORR. H2-TCPP ligands effectively prolong the distance between Fe3+ sites, thereby suppressing the agglomeration of Fe atoms during pyrolysis. Also, 3D network structures can be partially preserved, which are beneficial for the mass transfer of O species during ORR (Figure 3C). 56 Furthermore, various fillers and soft and hard templates have been used in pyrolysis to optimize the pore structures of carbon substrates. For example, Zhu et al. 60 pyrolyzed MIL-101-NH2 encapsulated with the fillers (dicyandiamide and FeCl3) to synthesize an Fe-based SAC supported on hierarchical carbon substrate for ORR. The hierarchically porous architecture is achieved owing to the internal stress and stains during the thermal decomposition of the fillers. One the other hand, Yang et al. pyrolyzed ZIF-8 to synthesize a Zn-based SAC supported on hollow N-doped porous carbon with a high mass loading of 11.3 wt %. A hollow structure can be obtained by using polystyrene (PS)/SiO2 particles as a hard template (Figure 3D). 57 , 61 CS-SACs can also be synthesized by high-temperature solid-state reactions, which are simple, generic, and potentially suitable for large-scale production. However, the rapid nucleation and growth of solid-state products during solution-phase synthesis are regarded as the main challenge for the formation of single atoms. To improve the yield of single atoms, Wu and co-workers proposed an in situ thermal atomization approach to convert carbon-supported Ni NPs into surface-bounded single Ni atoms at 900°C in Ar atmosphere (Figure 4 A). 62 Ni NPs not only acted as a metal precursor but also served as a catalyst to break C–C bonds on the carbon surface. In situ environmental TEM observation showed that Ni NPs first agglomerated and diffused into carbon matrix as the pyrolytic temperature increases. Thereafter, Ni NPs were eroded at a higher temperature (∼900°C), and vaporized single Ni atoms were anchored on carbon surfaces by forming Ni-N coordination sites at N-rich defects. In another study, bulk Cu powder and ZIF-8 were heated in a tube furnace under NH3 atmosphere to synthesize a Cu-based SAC supported on N-doped porous carbon (Figures 4B and 4C). It was proposed that NH3 molecules served as a “transporter” to haul out Cu atoms to form volatile Cu(NH3) x . Afterward, single Cu atoms were bonded to volatilized Zn nodes in ZIF-8. Furthermore, SACs based on other metals, such as Co and Ni, have also been synthesized by similar solid-state reactions. 63 As an effective top-down synthesis method for CS-SACs, electrochemical activation plays an essential role because of its simplicity, low cost, and environmental friendliness. Creating defects and vacancies on a carbon substrate surface is beneficial to capture moving transition-metal atoms. Moreover, CS-SACs fabricated by this method provide some advantages for in situ studies of correlations between structures of performances of catalysts. For example, Yao et al. synthesized atomically isolated Ni atoms embedded in graphitized carbon (denoted as A-Ni-C) for efficient HER via the electrochemical activation method. 64 The as-prepared Ni-MOF as a precursor was carbonized at 700°C in N2 to obtain Ni@C. After an acid (HCl) leaching treatment, the electrocatalyst was activated by constant potential and cyclic voltammetry treatment. Subsequently, in situ formed single Ni atoms were homogeneously dispersed on the graphitized carbon substrate. Besides monometal CS-SACs, bimetallic Co-Pt C/N based single-atom catalysts (denoted as A-CoPt-NC) were also synthesized by a facile electrochemical activation strategy. 65 , 66 Specifically, Co cores were removed from stable Co/C core-shell structures, producing N-doped defective carbons for anchoring atomic metal species. HAADF-STEM and X-ray absorption near-edge structure (XANES) analysis showed that isolated Co/Pt atoms are trapped in a vacancy-type defect in the shell of carbon capsules, thereby forming atomic Co-Pt-N-C coordination structures as active centers. Although the metal loading amount in the resultant A-CoPt-NC catalyst is low (∼1.72 wt % and ∼0.16 wt % for Co and Pt, respectively), the catalyst displays excellent activity and robust stability for ORR.Due to their excellent catalytic performances, CS-SACs have been explored in various critical energy-conversion and energy-storage applications. Many studies have been devoted to this research area in the last 3 years. Here, we summarize representative studies in three subsections. The first subsection focuses on the applications related to several essential reactions for energy conversion, including ORR, HER, OER, CO2RR, and NRR. Studies on bi-, tri-, even multi-functional catalysts for multiple reactions simultaneously are also included. The second and third subsections concentrate on the emerging applications of CS-SACs in supercapacitors and rechargeable batteries, respectively.The cathodic ORR is at the heart of many energy-conversion devices, including fuel cells and metal-air batteries. 67 , 68 However, the large overpotential required by ORR severely hinders the efficiency and practical application of these devices. Currently, carbon-supported Pt NPs are the commonly used catalyst with the best catalytic activity for ORR. The scarcity and high cost of Pt make it challenging for the wide adaption of Pt/C catalysts. Moreover, the poor stability of Pt catalysts caused by methanol crossover or CO poisoning severely affects their service life. Considerable efforts have been devoted to developing CS-SACs with low cost and high catalytic activities for ORR.A common strategy is to stabilize isolated atomic metal species on porous carbon substrates to obtain high catalytic activities for ORR. For instance, Li et al. synthesized an Fe-based SAC supported on carbon nanospheres. Specifically, 3D Fe and N co-doped hollow carbon nanospheres were formed by polymerizing aniline and pyrrole in the presence of carbon nanotubes (denoted as CNT-Fe/NHCNS) (Figures 5A–5C). 69 This porous structure inhibited the agglomeration of Fe, leading to the formation of abundant atomic Fe-N x sites. The CNT-Fe/NHCNS (with a mass loading of 0.2 mg cm−2) displayed a half-wave potential (E 1/2) of ∼0.84 V and a high limiting current density of ∼5.40 mA cm−2 in an O2-saturated 0.1 M HClO4 electrolyte. Another Fe-based SAC with similar structures was constructed by dispersing Fe atoms on N-doped carbon nanospheres (denoted as Fe-N-C HNSs). Atomically dispersed Fe atoms were anchored on carbon nanospheres produced from a biomaterial (histidine) with SiO2 NPs as templates. 70 The Fe-N-C HNSs exhibited an E 1/2 of 0.87 V and a limiting current density of 5.80 mA cm−2 in an O2-saturated 0.1 M KOH electrolyte.Besides their superior activities in alkaline electrolytes, CS-SACs also demonstrate high ORR performances in acidic electrolytes. 35 , 71 For example, Liu et al. synthesized a Pt-based SAC by maximizing the utilization efficiency of Pt on a defective carbon substrate (denoted as Pt1.1/BPdefect). 34 Computational calculations combining with experimental data revealed that single Pt atoms were anchored by four C atoms neighboring C divacancies. The formed Pt-C4 moieties are usually considered as the main active catalytic centers for ORR. As a result, the limiting current density (5.50 mA cm−2) and power density (520 mW cm−2) of an acidic fuel cell using Pt1.1/BPdefect in its cathode is comparable with fuel cells using commercial Pt catalysts (Figure 5D). Moreover, the atom utilization in Pt1.1/BPdefect is as high as 0.09 gPt kW−1, which can reduce the cost of catalysts in fuel cells. 34 In another study, an Fe-based SAC supported on N-doped carbon was produced by pyrolyzing hollow ZIF-8 with ferric acetylacetonate and g-C3N4. 71 A high-density Fe(II)-N4-H2O moiety (4.5 × 1013 sites cm−2) is anchored on the porous carbon. Of note, the catalyst has the E 1/2 of 0.780 V and 0.845 V in 0.1 M HClO4 and 0.1 M KOH electrolytes, respectively (Figures 5E and 5F). When it was used in the cathode of an H2/O2 PEMFC, the device delivered a current density of 400 mA cm−2 at 0.7 V or 133 mA cm−2 at 0.8 V, as well as a maximum power density of 628 mW cm−2. Most recently, Li et al. used a secondary-atom-assisted method to synthesize an Fe-based SAC supported on 1D porous N-doped carbon nanowires (denoted as Fe-NCNWs). 72 Due to its unique geometric structure and high Fe mass loading, Fe-NCNWs yielded an E 1/2 of 0.91 V and average kinetic current density (J K) of 6.0 mA cm−2 at 0.9 V in an alkaline electrolyte, and a satisfactory E 1/2 of 0.82 V and average J K of 8.0 mA cm−2 at 0.8 V in an acidic electrolyte, respectively. Furthermore, Fe-NCNWs also displayed superior long-term stability and methanol toleration in both alkaline and acidic electrolytes.HER, together with OER, determine H2 production by electrocatalytic water splitting using energy generated from sustainable sources. 73 Although Pt-based catalysts are one of the best catalysts for HER, the scarcity and high cost of Pt call for novel catalysts. Furthermore, the utilization of Pt in conventional Pt-based catalysts is very low. 43 A large number of CS-SACs, containing metal atoms, such as Co, Fe, Ru, Pt, W, and Mo, have recently been reported to address the above issues. 74–76 Moreover, some CS-SACs offer additional benefits, such as high electrical conductivity, tunable porous structures, and long-term stability in both acidic and alkaline electrolytes.Liu et al. 74 synthesized a Pt-based SAC for HER in an acidic electrolyte by anchoring Pt atoms on onion-like carbon nanospheres (denoted Pt1/OLC). Pt1/OLC with 0.27 wt % displayed a low overpotential (η 10) of ∼38 mV at the current density of 10 mA cm−2. Theoretical calculations suggested that a tip-enhanced local electric field at Pt sites on a curved carbon surface would boost the reaction kinetics of HER. In another study, a Pt-based SAC was synthesized by anchoring single Pt atoms on aniline-stacked graphene (denoted as Pt SASs/AG) using a facile microwave reduction method. 32 Extended X-ray absorption fine structure (EXAFS) results revealed that d-electron structures of Pt atoms are optimized by coordinating Pt with N in aniline molecules. Impressively, the as-fabricated Pt SASs/AG catalyst delivered a η 10 of 12 mV and a mass current density of 22,400 AgPt −1 under 50 mV. It should be stressed that the utilization of Pt in Pt SASs/AG is 46 times higher than that in commercial 20 wt % Pt/C catalyst. Furthermore, computational calculations were used to explain the correlation between catalytic activities and atomic structures of Pt SASs/AG. The considered models included Pt(111), single Pt atom adsorbed on graphene (Ptad/G), and Pt SASs/AG (Figures 6A–6C). Their corresponding partial densities of states (PDOSs) of 5d orbitals of Pt atoms are quite different (Figures 6D–6F). Pt atoms in Pt SASs/AG can retain atomic orbital characteristics of isolated Pt atoms via their adjacent aniline molecules. The Gibbs free energy (ΔG H∗) diagram (Figure 6G) indicates that the ΔG H∗ of Pt SASs/AG is −0.127 eV, which is comparable with that of Pt(111) facet (−0.121 eV). Thus, it was proposed that d-electron structures of Pt atoms and the hydrogen adsorption energy are optimized by the coordination between atomically isolated Pt and N in aniline molecules, leading to the improved catalytic activities for HER.Beyond CS-SACs supported on porous carbon substrates, CS-SACs with core-shell structures have also attracted significant interest for HER. Recently, Zhang et al. reported a Pt-based SAC by isolating Pt atoms in N-doped porous carbon core-shell structures (denoted as Pt@PCM). 77 Significantly, Pt@PCM exhibited low η 10 of 105 and 139 mV in 0.5 M H2SO4 and 1.0 M KOH electrolytes, respectively (Figures 7A–7C). It also showed long-term durability in both acidic and alkaline electrolytes. EXAFS results and DFT calculations suggested that the active catalytic sites are lattice-confined Pt centers and activated C/N atoms adjacent to Pt, which should be contributable to the active origin of the superior performances.Compared with electrochemical HER in acidic media, it is more challenging to catalyze HER in alkaline electrolytes due to additional requirements on water adsorption and activation. 78 Several theoretical studies have suggested that HER in acidic electrolytes is related to H adsorption (H ad) at active catalytic centers, whereas in alkaline electrolytes it is governed by the delicate balance among three descriptors: (1) the H ad on the catalyst surface, (2) the prevention from hydroxyl adsorption (OHad) referred as the poisoning of active sites, and (3) the energy required to dissociate water molecules. 79–81 In general, the electron transfer kinetics of HER on Pt surfaces in alkaline electrolytes are approximately two orders of magnitude lower than those in acidic electrolytes. 82 Recently, Lu et al. synthesized a Ru-based CS-SAC by co-doping Ru and N on carbon nanowires. 75 The catalyst showed η 10 of 12 mV in a 1 M KOH electrolyte and 47 mV in a 0.1 M KOH electrolyte. The theoretical calculations revealed that RuC x N y moieties in the catalyst have a very low hydrogen binding energy, which lowers the kinetic barrier for water dissociation. Most recently, a W-based CS-SAC (denoted as W-SAC) was synthesized by pyrolyzing an MOF (Figures 7D and 7E). 58 The W-SAC showed a η 10 of 85 mV and a small Tafel slope of 53 mV dec−1 in 0.1 M KOH electrolyte. HAADF-STEM and XAFS analysis results suggest that the W1N1C3 moiety (Figure 7F) is the favored local structure of W species. Furthermore, DFT calculations indicated that the W1N1C3 moiety can significantly reduce H adsorption energy, thereby lowering the overpotential of HER.OER is a four-electron-proton coupled reaction, which requires high energy to overcome its kinetic barrier. 83 In the past decades, various novel catalysts in forms of clusters, NPs, and nanosheets have been explored to reduce the overpotential of OER. However, developing high-performance OER catalysts with low cost, high stability, and high activities remains difficult. CS-SACs with the ultimate metal utilization efficiency may deliver better catalytic performances for OER. 17 Several CS-SACs have been applied to OER.Recently, a Ni-based CS-SAC (denoted as HCM@Ni-N) was synthesized by pyrolyzing a core-shell structure composed of silica nanospheres coated with resorcinol formaldehyde and methylimidazole-Ni (MI) (Figure 8 A). 84 Theoretical calculations indicated that Ni and N atoms can greatly reduce the reaction energy barrier for OER and accelerate its catalytic kinetics (Figures 8B and 8C). Consequently, HCM@Ni-N exhibited a η 10 of 304 mV, which is lower than that of RuO2 (393 mV), HCM@Ni (without N) (489 mV), and HCM@N (without Ni) (502 mV) (Figure 8D). Besides, the corresponding Tafel slope (Figure 8E) of HCM@Ni-N (76 mV dec−1) is considerably smaller than that of HCM@Ni (187 mV dec−1) and HCM@N (203 mV dec−1). For pristine HCM (hollow carbon matrix) and HCM@N, the conversion of OOH∗ to O2∗ is the rate-determining step (RDS) with the largest reaction free energy of 4.14 and 3.6 eV, respectively. The effective polarization between isolated Ni and coordinated N atoms exerts a significant impact on the RDS in the catalytic process. For HCM@Ni, a lower limiting barrier of 2.36 eV is obtained with the formation of O2∗ as the RDS. Moreover, the RDS for HCM@Ni-N is completely changed to the oxidation of OH∗ to O∗ with the smallest limiting barrier of 1.83 eV, indicating the much easier dissociation of H2O on the HCM@Ni-N and resulting in the promoted reaction kinetics and modified reaction mechanism. Specifically, the unsaturated Ni atoms should be the dominant active centers, and the electronic coupling between Ni and surrounding N atoms can induce electronic redistribution and the lower electron density near the E Fermi, leading to a significant enhancement of OER activity. More recently, a Ni-based CS-SAC (denoted as A-Ni@DG) was synthesized by incipient wetness impregnation of atomically dispersed Ni (1.24 wt %) on DG. 47 A-Ni@DG showed a η 10 of 270 mV for OER in a 1.0 M KOH electrolyte, which is lower than that of Ir/C (320 mV), DG (340 mV), and Ni@DG (310 mV). Furthermore, A-Ni@DG exhibited a Tafel slope of 47 mV dec−1, indicating excellent OER kinetics. Theoretical calculations unveiled that the excellent catalytic activities originate from the tuned local electronic structures of the atomic Ni.To systematically study the relationship among catalytic activity and the interaction of metal centers and carbon, Chen et al. constructed two types of N on the surface of graphene (graphitic N and pyridine-like N). 85 DFT calculation results suggested that pristine graphene would exhibit negligible catalytic activities for OER. However, the incorporation of graphitic and pyridine-like N in graphene can modulate the electronic structure of supported single Co atoms. The calculated free energy change diagrams imply that pyridine-like N can effectively induce local positive charges on the supported isolated Co atoms, which is beneficial for the adsorption of O-containing species. However, the local charge density of single Co atoms coordinated with more N atoms are likely to become too positive, which is unfavorable for conversing O intermediates. Hence, the authors concluded that moderate positive charge density on single Co atoms is desirable for efficient adsorption and transformation of O intermediates, which may be realized via suitable structure modification of the graphene surface.The excessive CO2 in the atmosphere from the intensive consumption of fossil fuels may cause significant changes in our ecosystems, for example, ocean acidification and global warming. 86 , 87 Consequently, exploring methods to convert CO2 to fuels or other value-added chemicals is important for addressing our environmental and energy challenges. 87–89 Among different methods, electrochemical CO2RR is promising because it usually can take place at room temperature and under ambient pressure, and it can produce a variety of useful gas and liquid substances, such as CO, CH3OH, CH4, and HCOOH. For example, CO can be directly used in gas-to-liquid conversion reactions to produce methanol by hydrogenation or generate liquid hydrocarbon fuels by the Fischer-Tropsch process. Nevertheless, CO2RR suffers from sluggish kinetics because of the low local concentration of CO2 and the low density of active sites on catalyst surfaces. In the past few years, various materials have been explored as catalysts for CO2RR, including metals, metal oxides, chalcogenides, and molecular metal compounds, as well as carbon nanomaterials. 90–92 Recent studies show that CS-SACs outperform many catalysts based on metal NPs. 93–95 For example, Zheng et al. 96 reported a Ni-based SAC (denoted as Ni-NCB) by depositing Ni on low-cost N-doped carbon black and applied it for CO2RR toward CO. A TEM image (Figure 9 A) shows that Ni-NCB has onion-like, defective graphene layers, serving as substrates for single Ni atoms. Ni-NCB also displayed an excellent performance for CO2RR when tested in a traditional H cell under 0.55 V overpotential in a 0.5 M KHCO3 aqueous electrolyte. A plateau of faradic efficiency toward CO (FECO) above 95% was observed over a broad potential range from −0.6 to −0.84 V versus RHE (Figure 9B). Ni-NCB delivered a current density above 100 mA cm−2 with a nearly 100% selectivity to CO, as well as excellent stability (Figures 9C and 9D). Ni-NCB outperformed several previously reported noble metal-based catalysts. 97 , 98 When Ni-NCB was integrated into a 10 × 10 cm−2 modular cell, the CO evolution current can reach 8.3 A with a high FECO of 98.4% and a CO production rate of 3.34 L h−1 per unit cell. The superior catalytic performance was attributed to the high Ni mass loading, the maximum utilization efficiency of Ni atoms, and the gas diffusion layer.In another study, atomically dispersed FeN5 single-atom sites were embedded in N-doped graphene (denoted as FeN5) and further used for electrocatalytic reduction of CO2. 99 FeN5 exhibited a high FE toward CO at 97.0% under an overpotential of 0.35 V (Figure 10 A), which is superior to many catalysts, including N-doped carbon nanomaterials 100 , 101 and CS-SACs containing Fe, Co, Ni, or Zn. 93 , 102 , 103 , 94 FeN5 also possessed excellent stability with FECO above 97% under −0.46 V versus RHE over 24 h (Figure 10B). DFT calculation results (Figure 10C) demonstrate that the free energy change of a key step (CO2 → ∗COOH) is only 0.77 eV on FeN5, which is much lower than that on FeN4 at 1.35 eV. Moreover, it is proposed that the axial pyrrolic N ligand in FeN5 can diminish the electron density of Fe 3d orbitals and reduce the Fe-CO π back-donation, thus achieving rapid desorption of CO (Figures 10D–10G).Apart from CO product, a Cu-based CS-SAC (denoted as CuSAs/TCNFs) was synthesized by anchoring Cu atoms on carbon nanofibers for electrocatalyzing CO2 into methanol. 95 Brunauer-Emmett-Teller measurement showed that the as-prepared CuSAs/TCNFs possesses a high specific surface area of 618 m2 g−1 with uniform pores of ∼100 nm, benefiting the mass transfer during electrocatalysis. Impressively, the C1 production selectivity of the as-prepared catalysts was nearly 100% under an overpotential of −0.9 V, including CH3OH (44%) and CO (56%), respectively. DFT calculations reveal ed that the ∗CO intermediates adsorbed on the active sites prefer to be converted into CH3OH rather than being released from the catalyst surface as CO due to the relatively high binding energy between single Cu atoms and ∗CO. 95 Furthermore, the C–C coupling (dimerization) pathway of ∗CO intermediates was substantially obstructed because of the synergetic effect of porous carbon structures and atomically distributed Cu active sites, resulting in the high selectivity to CH3OH. Therefore, it is critical to better understand the correlation between the material design and catalytic selectivity toward different products of CS-SACs, which offers an excellent platform for developing highly efficient catalysts with improved CO2RR performances (Table 1 ).Ammonia (NH3) is an essential precursor for the synthesis of fertilizers and other biological compounds. It has also been considered as an emerging fuel. 107 The current Haber-Bosch process used in NH3 synthesis is energy intensive and consumes 1–3% of the world's annual energy usage. Furthermore, the thermodynamically limited conversion is only ∼15%. The Haber-Bosch process also requires H2, which is currently produced by the steam reforming of natural gas with substantial CO2 emission (∼2 tonCO2 tonNH3 −1). 108 Therefore, it is highly desirable to develop more green and sustainable synthesis methods for NH3. Different catalysts have been explored for NH3 synthesis, including biological nitrogenases, photocatalysts, thermal catalysts, and electrocatalysts. Among them, NRR by electrocatalysts has attracted significant interest 109 because NH3 can be produced from H2O and atmospheric N2 without H2 at room temperature. 110 In comparison with ORR, HER, and OER, it is much more challenging to split N2 with the incredibly sturdy N≡N bond into free N radicals. Moreover, the adsorption of N2 on catalyst surfaces is often unsatisfactory, which adversely affects the formation of reaction intermediates, limiting the selectivity and yield of NH3. 111 Although many metal-based catalysts have been studied for NRR, both the NH3 yield and FE are still far from the requirements of practical applications. It should be noted that the bonding between most metals and N2 is too weak to enable efficient N2 adsorption and activation, which is often considered as the rate-limiting step for NRR. Besides, d-orbital electrons in transition metals are more favorable for the formation of metal H bonds, resulting in the adverse HER at 0 V, lower than that of NRR at 0.057 V versus standard hydrogen electrode (SHE), which compromises the FE for NRR. CS-SACs have some unique advantages to serve as electrocatalysts for NRR: (1) abundant exposed catalytic active sites providing high catalytic activities, especially for the adsorption of N2; (2) porous and conductive carbon substrates facilitating rapid mass transport and electron transfer; and (3) optimized surface properties with tunable hydrophobicity and hydrophilicity, enabling efficient three-phase contacts among solid catalysts, liquid electrolytes, and gaseous reactants.Recently, an Fe-based CS-SAC (denoted as FeSA-N-C) was synthesized by modulating polypyrrole-iron coordination complexes, yielding atomically dispersed Fe atoms on N-doped graphene-like structures (Figures 11A and 11B). 112 FeSA-N-C and N-doped carbon (N-C) were studied as catalysts for NRR in an N2-saturated 0.1 M KOH solution under different applied potentials. FeSA-N-C is more active for NRR, with a more positive onset potential of 0.193 V versus RHE compared with that of N-C (Figure 11C). The FE of FeSA-N-C is up to 56.55%, with an NH3 yield of 7.48 μg h−1 mg−1 at 0 V versus RHE (Figure 11D). In contrast, N-C shows a much lower FE of 9.34% with extremely low NH3 yield (Figures 11E and 11F). DFT calculation results suggested that FeSA-N-C can effectively boost the access of N2 with a low energy barrier of 2.38 kJ mol−1. The localized high concentration of N2 around Fe sites can facilitate N2 adsorption with a low binding Gibbs free energy of −0.28 eV.Apart from Fe-based CS-SACs, several other CS-SACs also show excellent performances for NRR. Qin et al. 44 reported an Au-based CS-SAC (denoted as AuSAs-NDPCs) by decorating single Au atoms on the surface of N-doped porous carbon for NRR. AuSAs-NDPCs showed a stable NH3 yield of 2.32 μg h−1 mg−1 at −0.2 V. Besides noble metals, a Mo-based CS-SAC (denoted as SA-Mo/NPC) was synthesized by anchoring Mo atoms on N-doped porous carbon. Due to the high density of MoN x active sites on hierarchically porous carbon frameworks, SA-Mo/NPC delivers an NH3 yield of 34.0 ± 3.6 μg h−1mg−1 with a high FE of 14.6% ± 1.6% in a 0.1 M KOH electrolyte at room temperature. 113 Moreover, no obvious current drop was observed after a long-term test of 50,000 s. Geng et al. synthesized a Ru-based CS-SAC (denoted as Ru SAs/N-C) by pyrolyzing a Ru-containing derivative of zeolitic imidazolate frameworks (ZIF-8). 114 Besides well-dispersed single Ru atoms, Raman spectra demonstrate the existence of defective structures in the N-doped carbon substrate. XANES and EXAFS results revealed that N atoms are well coordinated with Ru atoms. Due to its unique structure, Ru SAs/N-C exhibited an NH3 yield of 120.9 μg h−1mg−1 in an N2-saturated 0.05 M H2SO4 electrolyte, which is 1–2 orders higher than that of traditional metal-based electrocatalysts 115 , 116 or metal-free carbon catalysts. 117 , 118 Some critical test methods, such as the isotopic (15N2) labeling experiment, are necessary to unambiguously demonstrate that the NH3 obtained in NRR experiments is substantially produced by electrochemical reduction of N2 rather than from other exogenous sources. 119–121 The isotopic labeling experiment can show a distinguishable chemical shift of doublet coupling in 1H nuclear magnetic resonance spectra, which can be attributed to 15N in 15NH4+, thus confirming the origin of the obtained NH3.Bifunctional electrocatalysts for ORR and OER, HER and ORR, or HER and ORR are desirable for rechargeable metal-air batteries, water electrolyzers, and regenerative fuel cells. They are capable of catalyzing two reactions; thus, it may help to simplify devices that would otherwise be required to accommodate two different types of catalysts.In general, rechargeable Zn-air batteries require catalysts in their cathodes for both ORR and OER during discharging and charging, respectively. To this end, a bimetallic Co-Ni-based CS-SAC (denoted as CoNi-NPs/NC) supported on N-doped hollow carbon nanocubes (NC) was synthesized (Figures 12A and 12B). 17 As shown in Figure 12D, CoNi-SAs/NC displays a high catalytic activity for ORR with an onset potential of 0.88 V, an E 1/2 of 0.76 V, and a large limiting current density of 4.95 mA cm−2 in an O2-saturated 0.1 M KOH electrolyte, outperforming N-doped hollow carbon nanocubes, CoNi NPs supported on NC (CoNi-NPs/NC). Also, CoNi-NPs/NC exhibited a superior activity for OER with a η 10 of 340 mV and a Tafel slope of 58.7 mV dec−1 (Figures 12E and 12F). The rechargeable Zn-air battery was assembled using CoNi-NPs/NC in its cathode for both ORR and OER. Worthy of note, this battery exhibited excellent stability over 40 discharge-charge cycles (Figure 12G). Multiple Zn-air batteries were connected to power many red light-emitting diodes (LEDs) (Figure 12C).Both ORR and HER are required for water electrolyzers. One example for both ORR and HER is a Co-based CS-SAC (denoted as CoSAs/PTFs) synthesized using porous porphyrinic triazine-based frameworks. 18 The combination of HAADF-STEM image and XAS analysis confirmed that isolated single Co atoms in the Co-N4 moiety are homogeneously embedded in hierarchically porous substrates. 18 As a result, CoSAs/PTFs showed an excellent catalytic activity for ORR with a high E 1/2 of 0.808 V and a large limiting current density of 6.14 mA cm−2 in an O2-saturated 0.1 M KOH electrolyte. For HER, CoSAs/PTFs displayed a η 10 of 94 mV and a Tafel slope of 50 mV dec−1. DFT calculations unveiled that the electrocatalytic activity is related to the synergistic interplay between hierarchically porous carbon substrates and Co-N4.Pt-based catalysts usually show high activities for ORR and HER; however, their activities for OER are relatively unsatisfactory. 122 , 123 Several metal oxides, such as RuO2 and IrO2, are efficient catalysts for OER, whereas their activities for HER are poor. 124 , 125 Some researchers have explored tri- or multi-functional electrocatalysts, which can catalyze multiple reactions. 126–128 For example, Co-based CS-SACs are catalytically active for ORR, OER, and HER in separate studies; 113 , 129 , 130 while Ni-based CS-SACs are capable of catalyzing OER, HER, and CO2RR in other studies. 64 , 86 , 131 This research suggests that CS-SACs may serve as tri- or multi-functional electrocatalysts.A recent demonstration involved an Fe-based CS-SAC (denoted as Fe-N4 SAs/NPC) synthesized by a polymerization-pyrolysis-evaporation method, which contains atomically dispersed Fe-N4 active sites embedded in N-doped porous carbon. 132 Fe-N4 SAs/NPC exhibits excellent activities for ORR, OER, and HER (Figures 13A–13C). Its onset potential and E 1/2 for ORR were ∼0.972 V and 0.885 V, respectively, and its η 10 for OER was 0.43 V, outperforming commercial 20 wt % Pt/C and RuO2 catalysts. Furthermore, Fe-N4 SAs/NPC showed a η 10 of 0.202 V and a Tafel slope of 123 mV dec−1 for HER, comparable with those of 20 wt % Pt/C. Importantly, Fe-N4 SAs/NPC was successfully applied in a water electrolyzer and a Zn-air battery (Figures 13D and 13E). The electrolyzer using electrodes containing Fe-N4 SAs/NPC required a low η 10 of 1.67 V (Figure 13F). Encouragingly, the Zn-air battery containing Fe-N4 SAs/NPC has a lower charge-discharge voltage gap (1.45 V at 50 mA cm−2) and a larger power density (232 mW cm−2) than those of the battery based on Pt/C and Ir/C catalysts (1.59 V at 50 mA cm−2 and 52.8 mW cm−2) (Figure 13G). DFT calculations attributed the trifunctional activity for ORR-OER-HER to the coupling effects between Fe-N4 active centers and porous carbon frameworks.Supercapacitors are electrochemical energy-storage devices, which can store electrical energy based on electrical double-layer capacitance (EDLC) or pseudocapacitance resulting from fast surface redox reactions or ion insertions on electrode surfaces. They can deliver higher power with long cycling life but lower energy density than conventional batteries. 133 , 134 The efficient adsorption/desorption of electrolyte ions on electrode surfaces of supercapacitors is critical for their energy storage. 135 Current supercapacitor electrodes are mostly based on porous carbon materials with large surface area to deliver EDLC. Emerging applications in consumer electronics, hybrid electric vehicles, and industrial electric utilities require supercapacitors with higher energy-storage density. 134 , 136 CS-SACs may be incorporated into carbon electrodes. It has been proposed that single-atom sites may catalyze some surface redox reactions or play other beneficial roles, which increase pseudocapacitance, leading to higher energy-storage density. 137–140 The following studies have explored this idea and demonstrate its feasibility to some extent.Yu et al. embedded multi-components, including Ni, P, N, and O, into a carbon substrate by one-step pyrolysis of a pre-designed MOF at different temperatures. The resulting hierarchical Ni/P/N/C composite was directly used as supercapacitor electrodes. 137 The composite contained Ni single atoms uniformly dispersed in microporous carbon substrates. The Ni/P/N/C composite pyrolyzed at 500°C has a high specific capacitance of 2879 F g−1 at the current density of 1 A g−1 (Figure 14 A). In another study, a Co-based CS-SAC (denoted as Co-POM/rGO) was successfully synthesized by depositing polyoxometalate (POM) on the surface of rGO aerogel at a mild temperature. 138 This material has a relatively small specific surface area of ∼173.3 m2 g−1, and the electrode made of Co-POM/rGO has a specific capacitance of 211.3 F g−1 at 0.5 A g−1 based on galvanostatic charge/discharge (GCD) measurement (Figure 14B). A solid-state asymmetric supercapacitor was assembled using Co-POM/rGO with an energy density of 37.6 Wh kg−1 at the power density of 500 W kg−1. The supercapacitor has reasonably good cycling stability with capacitance retention of 95.2% after 5,000 charge-discharge cycles at 2 A g−1 (Figure 14C). More recently, Shan et al. 140 incorporated K or Na single atoms on 2D g-C3N4 decorated with MnO2 (denoted as CNM). Figure 14D shows that K-CNM has a specific capacitance of 373.5 F g−1 at 0.2 A g−1, which is approximately 4 times higher than that of g-C3N4@MnO2 with K atoms, and also higher than that of Na-CNM at 294.7 F g−1. Also, K-CNM has some cycling stability with 93.7% capacitance retention after 1,000 charge-discharge cycles at 1 A g−1 (Figure 14E). The authors proposed that doping single metal atoms improved the electrical conductivity of electrodes and enhanced the mass transfer of electrolyte ions, contributing to the observed higher capacitance.Many new batteries are currently being explored due to the strong demand for more efficient energy-storage solutions. 141–143 Rechargeable metal-air batteries, sodium-sulfur (Na-S) batteries, and lithium-sulfur (Li-S) batteries have attracted significant interest because of their potentially low cost, high energy-storage capacity, and prolonged service life. 144–147 It has been proposed that adding CS-SACs in battery electrode materials may improve the energy density and rate performance of these emerging batteries. CS-SACs play essential roles in Zn-air batteries, as discussed earlier in the Bifunctional Electrocatalysis section. In this section, we focus on recent studies of the application of CS-SACs in Na-S and Li-S batteries. 148–152 Single Co atoms were distributed in hollow carbon nanospheres to serve as S host (denoted as S@Con-HC) in Na-S batteries. 148 Na-S batteries using S@Con-HC electrodes delivered a capacity of 220.3 mAh g−1 at 5 A g−1 and displayed an enhanced cycling stability with a capacity of 508 mAh g−1 after 600 cycles at 0.1 A g−1 (Figures 15A and 15B). Combining operando Raman spectroscopy analysis, XRD, and computational calculations, the authors proposed that S@Con-HC can improve the reactivity of S and alleviate the “shuttle effect.” 148 The adsorption energy of Na2S2 on Co sites is −7.85 eV whereas that of Na2S is −10.67 eV. The binding of Na2S2 and Na2S clusters is much stronger on Co sites than on pristine carbon surface, which alleviates the dissolution of S and impedes the shuttle effect. This work connects the fields of batteries and electrocatalysts, providing a new exploration direction to enhance the performance of Na-S batteries.Several other studies have explored the application of CS-SACs in Li-S batteries. Wang et al. 149 synthesized a nanostructured Li2S cathode composed of uniformly dispersed single Fe atoms anchored on a porous N-rich carbon substrate. The assembled Li-S battery demonstrated a high rate performance of 588 mAh g−1 at 12 C and a long cycling life with a capacity decay rate of 0.06% per cycle for 1,000 cycles at 5 C. A working mechanism was proposed and is presented in Figure 15C. The Fe-based CS-SAC reduces the energy barrier, leading to a low activation voltage of Li2S. In another study, single Ni atoms were confined in N-doped graphene (denoted as Ni@NG) as separators in Li-S batteries. 150 XANES results (Figure 15D) suggest that active Ni centers in Ni-N4 can chemically trap Li polysulfides (LiPS) and form strong S x 2−···Ni–N bonds. The charge transfer between LiPS and oxidized Ni sites provides low energy barriers, leading to fast LiPS conversions during charging/discharging. Li-S batteries assembled with Ni@NG displayed a stable cycling life with 0.06% capacity decay per cycle after 500 cycles. Most recently, Xie et al. synthesized a Co-based CS-SAC (denoted as SC-Co) as interlayers for Li-S batteries. 151 Homogeneous dispersion of Co atoms was obtained, which catalyzed redox reactions of S (Figure 15E). Inspiringly, Li-S batteries with the SC-Co interlayer exhibited a high initial capacity of 1,130 mAh gS −1 at 0.5 C (1 C = 1,672 mAh gS −1) and a final capacity of 837 mAh gS −1 after 300 cycles, corresponding to a capacity retention of 74.1% and a low fading rate of 0.086% per cycle (Figure 15F). Such batteries may find applications in electric vehicles.The development of efficient and cost-effective catalysts is of great importance for many energy-conversion and energy-storage applications. CS-SACs, as a new class of catalysts, have the potential to replace or complement current noble metal-based catalysts for a range of electrochemical reactions, including ORR, HER, OER, CO2RR, and NRR. They are also useful in improving the performance of electrochemical energy-storage devices such as supercapacitors and rechargeable batteries. Presently, CS-SACs are fabricated using both bottom-up and top-down methods, such as ALD, wet chemistry synthesis, high-energy ball milling, pyrolysis of organic precursors, and high-temperature solid-state reactions, as well as electrochemical activation. Considering the quality of resulting CS-SACs and scalability, current synthesis methods still face various challenges for practical applications. For the characterization of CS-SACs, high-resolution TEM (HRTEM) and XAS analysis involving EXAFS and XANES are usually required to identify geometric and electronic structures. In particular, XAS analysis has been extensively applied to determine the oxidation state of metal centers, bond length, short-range disorder, coordination number, and local geometry. Furthermore, computational methods have often been used to understand the mechanisms related to catalytic activity and stability. Although tremendous progress has been made in the past few years by the strategy that combines advanced synthetic approaches and characterizations, there remain many challenges to be addressed. We provide our views on several key challenges below.Currently it is still challenging to prepare CS-SACs with high metal mass loadings. For example, the pyrolysis of organic precursors is a high-temperature process, which often inevitably leads to the aggregation and sintering of metal atoms into nanoclusters or NPs if strong interactions between metal atoms and substrates are absent. High energy input is required in ball-milling-based methods, which also would cause the agglomeration of single atoms. This decreases the total number of catalytic sites and changes their electronic structures, thereby reducing the catalytic activities of CS-SACs. However, high-temperature treatment is often beneficial for the graphitizing of carbon materials to obtain better electrical conductivity in carbon substrates. Therefore, novel synthesis methods, which can minimize exposure to high temperatures or high energy inputs while producing carbon substrates with sufficient electrical conductivity, are desirable to obtain CS-SACs with high metal mass loadings.Certain advances in the rational design of high-performance CS-SACs have been made by developing various synergistic approaches in the past years. On the foundation of such design lies the materials' atomistic and electronic structure that dictates the intrinsic activity trends and related catalytic mechanism. However, the unambiguous identification of the atomistic structure of the CS-SACs represents a severe challenge for their activity origin study. Several characterization techniques, such as STEM, XRD, Mössbauer spectroscopy, and electron paramagnetic resonance spectroscopy, have been used to probe the atomistic or electronic structures of SACs. The inherent characteristics of CS-SACs, such as the non-crystallographic ordering of the metal atoms and the heterogeneity in structure and composition, suggest that sophisticated techniques should be employed. Recently, many publications have proved synchrotron XAS to be a well-suited and powerful technique for characterizing CS-SACs because it can provide valuable information about the coordination environment and the chemical state of the probed atom in an element-selective way. Moreover, XAS allows the study of the dynamic process of electrochemical reactions under operando conditions, which is critical for establishing a precise structure-activity relationship and understanding the catalytic mechanism on the three-phase interface. Nonetheless, unambiguously extracting the exact geometric and electronic structure by XAS remains neither trivial nor straightforward, and the complementary use of different techniques is often necessary. Therefore, methods employing multiple technologies combined, such as HRTEM-XAS-XPS, are highly feasible for genuinely identifying the geometric and electronic structure of CS-SACs.In general, chemical reactions would take place at active catalytic sites that contain unsaturated, distorted, or edge atoms. Understanding the correction between the structures of active sites and their catalytic properties is critical to developing better catalysts. The difficulty of gaining such understanding can be attributed to the structural ambiguity of many potential active sites in a catalyst as well as the poor applicability of transferring information gained from one catalyst to others. Thus, it is desirable to develop model CS-SACs, in which their active catalytic sites are well defined and applicable to different catalysts. Furthermore, an important task linked to this is the precise quantification or imaging of active sites in CS-SACs, especially in reaction conditions. Advanced characterization tools, such as XAS, TEM, atomic force microscopy, and various spectroscopies, are critical to resolving many puzzles.Stability or durability is another crucial factor in practical applications of CS-SACs. The degradation of active catalytic sites is a common phenomenon in many catalysts. For example, Pt NPs in catalysts used in fuel cells would be oxidized from Pt to Pt2+ and aggregate into larger particles due to Ostwald ripening. The dissolved Pt2+ ions can diffuse at the micrometer scale and cause the degradation of fuel cells over time. Thus, there is generally a trade-off between the activity and long-term stability of catalysts. Most of CS-SACs reported so far demonstrating superior activities were tested in research lab settings. There is still a lack of reliable experimental data related to their long-term stability in practical devices. It is essential first to obtain such data to evaluate the performance of CS-SACs in realistic application conditions. If they face a trade-off between activity and stability similar to that in other catalysts (most likely they do), novel methods are needed to increase their stability while retaining their high activities.Many current studies of CS-SACs have already used various computational studies to explain their observed catalytic properties. For example, theoretical calculations are quite successful in predicting atomic structures of CS-SACs, which have been verified by spectroscopies experimentally. The experimental results further stimulate the development of new theories to understand the reaction mechanism. The feedback loop between computational and experimental studies has driven the in-depth understanding and development of CS-SACs for various potential energy-storage and energy-conversion applications. However, many computational studies are based on ideal models with assumptions, simplification, and multiple tunable parameters. It is necessary to avoid the temptation of jumping into simple conclusions by artificially matching experimental data with computational results. We believe the successful feedback loop between theory and experiments will be essential to an in-depth understanding of the structure, mechanism, and kinetics of the catalytic centers. Also, it will guide the design and development of CS-SACs with a desirable activity for specific reactions crucial in energy conversion and storage, as well as for large-scale chemical synthesis, environmental monitoring, and energy devices.This work was supported by the Australian Research Council under the Future Fellowships scheme (FT160100107), Discovery Project (DP180102210), and ARC Discovery Early Career Researcher Award (DE200101669). S.Z. thanks financial support from the F.H. Loxton fellowship.Conceptualization, S.Z. and Y.C.; Writing – Original Draft, Yongchao Yang and Yuwei Yang; Writing – Review & Editing, Z.P., K.-H.W., C.T., H.W., L.W., A.M., C.Y., J.D., S.Z., and Y.C.; Supervision, S.Z. and Y.C.

Single-atom catalysts (SACs) have the advantages of both homogeneous and heterogeneous catalysts, which show promising application potentials in many renewable energy-conversion technologies and critical industrial processes. In particular, carbon-supported SACs (CS-SACs) are of great interest because of their maximal atom utilization (∼100%), unique physicochemical structure, and beneficial synergistic effects between active catalytic sites and carbon substrates. In this review, we offer a critical overview of the unique advantages of CS-SACs related to their material designs, catalytic activities, and potential application areas. The state-of-the-art design and synthesis of CS-SACs are described under the framework of bottom-up and top-down approaches. We also comprehensively summarize recent advances in developing CS-SACs for important electrochemical reactions, i.e., oxygen reduction reaction, hydrogen evolution reaction, oxygen evolution reaction, CO2 reduction reaction, nitrogen reduction reaction, serving as bi-/multi-functional electrocatalysts, and usages in supercapacitors and batteries. Lastly, the critical challenges and future opportunities in this emerging field are highlighted.

Because of the diverse magnetic domain configurations from superparamagnetism, single domain, vortex domain to multi-domain, magnetic functional materials with different frequency spectrum characteristics are widely used in the field of medical treatment, information record/storage, electronic device, and electromagnetic wave absorption [1–3]. Due to the lack of effective synthesis strategy and direct visible characterization methods, the responding behaviors of magnetic vortex domain at nanoscale are not clear when interacting with electromagnetic (EM) wave at gigahertz (GHz) frequency band.Recently, metal-organic frameworks (MOFs) derivatives have been widely popular in EM wave dissipation materials and functional devices because of their unique three-dimensional periodic structure and adjustable electromagnetic properties [4–6]. Firstly, size, morphology, and micro-nano structure of MOFs-derived composites can be maintained from targeted precursors, providing the confined component distribution. Secondly, MOFs-derived EM wave absorbers show unique magnetic-carbon interfaces design, electronic migration routes, and EM balance features, encouraging the dielectric dissipation. More importantly, the cation host can be converted into different magnetic components, including Fe/Co/Ni metal, FeNi/Ni1-x Co x alloy, Fe3O4 oxide, Fe3C nitride, and NiFe2O4 ferrite with different magnetic structure [7,8]. These magnetic particles offer huge space to regulate the magnetic domain, saturation magnetization, and magnetic respond behaviors, dominating the final consumption ability with incident EM energy. To fabricate the wide-frequency EM wave absorbers, many MOFs precursors and carbonized derivatives were designed such as Co-ZIF-67, Co-MOF-74, Fe-MOF-5, Fe-MIL-88, Co–Zn-ZIF-67 and so on [9–14]. Those EM wave absorbers exhibit magnetic-carbon synergy effect and strong absorption intensity. However, the MOFs-derived absorbers still face some challenges for modern practical application, including wide frequency absorption, low adding mass, and tunable absorption frequency. Meanwhile, the magnetic domain motion, magnetic coupling effect, and absorption mechanism is unclear. So, the further investigation is urgently needed, which still face huge challenge.Herein, a nanoscale magnetic vortex domain is firstly observed in the MOFs-derived Ni particles wrapped by graphitized carbon shell. Because of the high symmetry spheres boundary exposed by carbon shell, Nickel magnetic domain tends to be arranged to form the special magnetic vortex conformation according to the easy magnetization surface. Magnetic-carbon microsphere was successfully assembled by abundant of core-shell Ni@C units using Ni-MOFs as a precursor, which not only constructs balanced magnetic-dielectric distribution but also forms directional electronic migration routes. As EM wave absorbers, MOFs-derived Ni@C microspheres exhibit outstanding energy absorption ability and wide frequency respond regions. The minimum reflection loss (RLmin) values of Ni@C–H, Ni@C–S, and Ni@C–V microspheres can reach to −47.2 ​dB, −48.5 ​dB, and −54.6 ​dB, respectively. Benefited from the confined magnetic vortex motion under high-frequency EM field, Ni@C–V microsphere possess the widest efficient absorption bandwidth (EAB) to 5.0 ​GHz at only 2.0 ​mm and 25% mass adding. Simultaneously, both reversion motion and core polarity change of the magnetic vortex together contribute to the enhanced EM energy dissipation and expanded absorption frequency. MOFs-derived magnetic-dielectric microspheres with special magnetic domain structure project a new idea to fabricate excellent EM wave absorption candidates.The synthetic schematic of MOF-derived Ni@C microspheres is displayed in Fig. 1 . By adjusting the organic ligands, various Ni-MOFs precursors are firstly synthesized in the mixed EtOH/DMF/H2O solvent under undergoing a solvothermal reaction. Due to the different functional groups on the benzene ring structure, changeable coordination and periodic structure can be built between the organic ligand (H3BTC, TA and ATA) and metal Nickel (Ni2+) host, dominating the final size and morphology. In Fig. S1, it can be found the diversity microsphere morphology among Ni-MOFs-H, Ni-MOFs-S, and Ni-MOFs-V precursors. Secondly, obtained Ni-MOFs powders are further annealed in an Ar condition. Nickel ions are reduced to nickel metal particles/clusters acting as a catalyst. Finally, adjacent organic ligand is transformed into graphitized carbon shell wrapping the Ni core, which form a basic core-shell Ni@C unit. Finally, MOFs-derived Ni@C–H, Ni@C–S, and Ni@C–V microspheres are successfully fabricated with different Ni-MOFs precursors.As displayed in the XRD spectrum (Fig. 2 a), the typical diffraction peaks located at 2θ ​= ​44.5°, 519°, and 76.5° are belonged to the Ni metal (JCPDS# 83–4000) and the weak diffraction peak of 2θ ​= ​26° are contributed to the carbon component. To further the evaluate the degree of graphitization, Raman spectrum of MOFs-derived Ni@C microspheres are further obtained (Fig. 2b). Generally, the intensity ratio of D-band and G-band (I D/I G) is a key indicator, dominating the catalytic ability of reduced Ni NPs. The I D/I G ratio of MOFs-derived Ni@C microspheres keep similar value of 1.02, implying the equal capacity to promote the graphitization of organic ligands. Meanwhile, due to the presence of magnetic Ni NPs, MOFs-derived Ni@C microspheres exhibit unique magnetic responding capacity, reflecting by the hysteresis loop curves (Fig. 2c). The saturation magnetization (Ms) values of Ni@C–H, Ni@C–S, and Ni@C–V microspheres are 50.2 emu/g, 58.8 emu/g, and 53.1 emu/g, respectively. As results, MOFs-derived Ni@C microspheres display similar component, graphitization degree, and magnetic property.Due to the diversity of the coordination mode between the host nickel ion and organic ligand, MOFs-derived Ni@C microspheres exhibit different microstructures from the SEM and TEM images (Figs. 2 and 3 ). Using H3BTC as ligand, the size of Ni@C–H microspheres is 1.5–2.0 ​μm, and spindle-shaped nanoparticles randomly decorate on the surface of the rough microsphere (Fig. 2d–f, Fig. 3a1-a3). When the ligand is TA, obtained Ni@C–S microspheres show a smooth surface and a perfect spherical shape at 2.0–4.0 ​μm (Fig. 2g–i, Fig. 3b1–b3). Different with above-mentioned Ni@C composites, MOFs-derived Ni@C–V microsphere has a unique wrinkled surface and the particle size is 1.5–2.0 ​μm when the ligand is ATA (Fig. 2j-l, Fig. 2c1-c3). In the HRTEM images, highly graphitized carbon shells wrap the metallic nickel particle constructing unique core-shell structure (Fig. 3a4, 3b4, 3c4). Because of the ligand transformation and brooked MOFs frameworks, there are lots of porous and gaps in those Ni@C microspheres (Fig. 2d–l). Due to the confinement reduction effect at nanoscale, carbonized Ni-MOFs microspheres indicate the uniform element mapping distribution at micrometer scale (Fig. S2). As results, magnetic-carbon Ni@C composites with various nano-micro structure are successfully prepared using Ni-MOFs as templates.To evaluate the EM wave dissipation capacity, MOFs-derived Ni@C microspheres are characterized by the vector network analyzer (VNA) at 2–18 ​GHz. Generally, EM parameters are composed of complex permittivity (ε r ​= ​ε′-jε″) and complex permeability (μ r ​= ​μ′-jμ″) [15–17]. They can not only reflect the intrinsic storage properties (ε′, μ′), but also determine the final energy absorption ability (ε″, μ″) of EM wave conversion materials. As shown in Fig. 4 a–c, EM parameters of three MOFs-derived Ni@C microspheres all show classical frequency-dependency feature. Remarkably, the real part (ε′) value of the Ni@C–H, Ni@C–S, and Ni@C– Ni@C–V decrease from the initial 9.1 to 5.4, 11.5 to 8.1, and 12.6 to 7.4, respectively. As the frequency increases, the imaginary part (ε″) data of the complex permittivity show a similar declining trend. The ε″ values decrease from 4.1 to 2.4 for Ni@C–H, from 4.8 to 2.6 for Ni@C–S, and 7.0 to 2.8 for Ni@C–V, respectively. Due to the permeability dispersion characteristics, the real part (μ′) and imaginary part (μ″) values of MOFs-derived Ni@C microspheres are relatively similar without huge difference, which maintain at 1.15–1.0 and 0.1–0.02, respectively. Higher EM parameters values in the Ni@C–V microspheres mean the stronger energy absorption [18]. Compared with other Ni@C microspheres, MOFs-derived Ni@C–V microspheres hold the higher attenuation constant (Fig. S3) and loss tangent values (Fig. S4), also meaning the better EM wave absorption capacity.Based on the obtained EM parameters, the reflection loss (RL) values of MOFs-derived Ni@C powders are calculated as shown in Fig. 4d–f. Due to the well impedance matching (Z ​= ​Z in/Z 0) values (Fig. S5) and synergy EM wave absorption ability, MOFs-derived three Ni@C microsphere all display high-performance EM absorption. Changing the thickness from 1.5 ​mm to 4.0 ​mm, the RLmin values of the Ni@C–H are −7.8 ​dB, −32.4 ​dB, −47.2 ​dB, −31.6 ​dB, −30.2 ​dB, and −30.7 ​dB (Fig. 4g). When the thickness is 3.0 ​mm, Ni@C–S microsphere displays the strongest RL value of −48.5 ​dB at 8.6 ​GHz (Fig. 4h). For the Ni@C–V microspheres, the RLmin value can reach −54.6 ​dB at 10.6 ​GHz, and the efficient absorption (RL ​≤ ​−10 ​dB) regions up to 5.0 ​GHz at only 2.0 ​mm (Fig. 4i). In addition, the influence of ratio of Ni/C, the annealing temperature, and the filling ratio on the EM absorption of MOFs-derived Ni@C powder are also discussed in detail (Table S1, Table S2, Table S3, Fig. S6, Fig. S7, Fig. S8). It can be concluded that the Ni@C–V microspheres with special conditions (1.5 ​mmol Ni2+ adding mass, 600 ​°C annealing, 25% filling ratio) can exhibit outstanding EM energy conversion behavior. Surprisingly, MOFs-derived Ni@C–V microspheres exhibit outstanding EM wave conversion ability and tuning efficient absorption frequency (C-band, X-band, Ku-band), which dominate the huge potential as lightweight and efficient EM wave absorbers. Definitely, the EM wave dissipation ability and the absorption bandwidth of MOFs-derived Ni@C–V microspheres get enhanced significantly compared with the Ni@C–H and Ni@C–S microspheres, indicating that the magnetic vortex could provide more contributions. The associated EM wave absorption mechanisms are discussed as following aspects. i) 3D magnetic coupling network enhanced electromagnetic wave consumption. Due to the presence of metallic Ni NPs, MOFs-derived magnetic Ni@C microspheres exhibit unique magnetic domain structure and responding properties. Clearly, the metal content of Nickel in the Ni@C–H, Ni@C–S and Ni@C–V is 89.4%, 67.4%, and 77.1%, respectively (Fig. S9). With the support of electronic holography technology, the reconstructed phase hologram is used to display the intrinsic magnetic field line distribution of Ni@C microspheres (Fig. 5 b, e, 5h). It can provide strong evidence for understanding the magnetic loss mechanism toward EM wave energy [4,5,19–22]. Clearly, all the Ni@C samples can radiate out high-density magnetic field line surrounding the microspheres surface (Fig. 5a, d, 5g). Meanwhile, the space scope of emitted magnetic flux lines greatly exceeds the size of the Ni@C sphere itself. Zooming in the surface region of Ni@C–V microsphere (Fig. 5j and k), shape-dependence magnetic lines distribution is directly observed from the reconstructed holography images (Fig. 5l). The shared magnetic flux lines among those Ni@C microspheres dominates the enhanced magnetic coupling effect (Fig. 5c, f, 5i). The unique interaction can boost the magnetic responding intensity and expand the responding space [23–25]. As a result, constructing unique 3D magnetic coupling network improve the complex permeability and the magnetic consumption ability toward incident EM wave energy. ii) Heterojunction Ni–C interfaces and connected graphited carbon shells boosted dielectric absorption. Intrinsic dielectric properties of Ni@C microspheres are particularly important, which determines the dielectric absorption ability. According to the Debye theory and dielectric dissipation mechanism, the main contribution is represented by the conduction loss and interfacial polarization [26–29]. The relationship between the complex permittivity (ε′, ε″) and intrinsic conductivity (σ) can be explained as follow formulas: (1) ε ′ = ε ∞ + ε s − ε ∞ 1 + ω τ 2 (2) ε ″ = ε s + ε ∞ 1 + ( ω τ ) 2 ω τ + σ ω ε 0 where ε′ is the real part of complex permittivity, ε″ is the imaginary part of complex permittivity, ε s is the static permittivity, ε ∞ is relative dielectric permittivity at the high-frequency limit, ω is angular frequency, τ is polarization relaxation time and σ is conductivity, and ε 0 is vacuum dielectric permittivity. 3D magnetic coupling network enhanced electromagnetic wave consumption. Due to the presence of metallic Ni NPs, MOFs-derived magnetic Ni@C microspheres exhibit unique magnetic domain structure and responding properties. Clearly, the metal content of Nickel in the Ni@C–H, Ni@C–S and Ni@C–V is 89.4%, 67.4%, and 77.1%, respectively (Fig. S9). With the support of electronic holography technology, the reconstructed phase hologram is used to display the intrinsic magnetic field line distribution of Ni@C microspheres (Fig. 5 b, e, 5h). It can provide strong evidence for understanding the magnetic loss mechanism toward EM wave energy [4,5,19–22]. Clearly, all the Ni@C samples can radiate out high-density magnetic field line surrounding the microspheres surface (Fig. 5a, d, 5g). Meanwhile, the space scope of emitted magnetic flux lines greatly exceeds the size of the Ni@C sphere itself. Zooming in the surface region of Ni@C–V microsphere (Fig. 5j and k), shape-dependence magnetic lines distribution is directly observed from the reconstructed holography images (Fig. 5l). The shared magnetic flux lines among those Ni@C microspheres dominates the enhanced magnetic coupling effect (Fig. 5c, f, 5i). The unique interaction can boost the magnetic responding intensity and expand the responding space [23–25]. As a result, constructing unique 3D magnetic coupling network improve the complex permeability and the magnetic consumption ability toward incident EM wave energy. Heterojunction Ni–C interfaces and connected graphited carbon shells boosted dielectric absorption. Intrinsic dielectric properties of Ni@C microspheres are particularly important, which determines the dielectric absorption ability. According to the Debye theory and dielectric dissipation mechanism, the main contribution is represented by the conduction loss and interfacial polarization [26–29]. The relationship between the complex permittivity (ε′, ε″) and intrinsic conductivity (σ) can be explained as follow formulas:Based on the equations, the imaginary part (ε″) of complex permittivity is proportional to the electronic conductivity (σ), which means that the higher electronic conductivity is benefit to the enhanced dielectric dissipation ability [30–35]. Undergoing the carbothermal reduction process, carbon-containing organic ligand were finally converted into the connected carbon shells. Catalyzed by the inner nickel core, graphitized carbon shells connect to each other, forming an electron transport network. Under the action of high-frequency microwave field, the rapid electron migration ability improves the conduction loss in the Ni@C–V material by generating the microwave energy into Joule heat [36,37]. Meanwhile, MOFs-derived Ni@C microsphere is assembled by plentiful connected magnetic@carbon nanoparticles. Reduced Ni NPs are wrapped by the graphitized carbon shell, constructing heterojunction Ni–C interfaces and fast electronic migration routes.Visually, the information of charge density distribution surrounding at Ni–C interfaces is observed via high-resolution electron holography images (Fig. 6 a). The Ni–C interfaces are clearly distinguished by different featuring colors between nickel core and carbon shell (Fig. 6b and c). Focusing on the carbon layers, graphitized carbon tightly bridges magnetic Ni core, building rich heterejunction Ni–C interfaces regions (Fig. 6d). Because the metallic nickel core is wrapped by the graphitized carbon layer, a mutation of charge density distribution is generated at the Ni–C core-shell interfaces (Fig. 6e). Due to the difference in the electrical properties between Ni and carbon, the metallic Ni tends to accumulate more negative charges, while the position of the carbon layer gathers more positive charges (Fig. 6f). The similar situation also occurs in the two connected Ni@C–V units, indicating that the charge distribution can be effectively modulated in the heterejunction region (Fig. 6g and h). Therefore, MOFs-derived Ni@C–V microspheres provide the high-density interfacial polarization regions, contributing to the enhanced dielectric loss. iii) Confined magnetic vortex reversal expanded the efficient absorption frequency. By adjusting the organic ligand into ATA, MOFs-derived Ni@C–V microsphere are assembled by plentiful core-shelleyi@C unit. Compared with other reported MOFs-derived Ni@C EM wave absorption materials, special magnetic vortex structure was firstly observed in the Ni core at nanoscale derived from Ni-MOF-V precursor. Benefited from the chemistry environment and coordination structure from ATA ligand, host nickel ion (Ni2+) was reduced into magnetic NPs, which further promoted the graphitization of organic ligands. In turn, formed carbon shell on the surface of Ni core will limited the expand space with the growth of magnetic Ni NPs. With high symmetry spheres and boundary restriction of graphited carbon shell, confined magnetic vortex structure is generated in the Ni@C–V microspheres powders. However, the Ni@C–H, Ni@C–V and other reported Ni@C magnetic-carbon composites do not show the magnetic vortex structure. The other important factor is the size of magnetic NPs, which is necessary to provide enough space to generate magnetic vortex core and domains. According to EM wave absorption theory, the natural resonance of magnetic Ni@C–V microsphere becomes the main dissipation mechanism below 8 ​GHz. Increased ferromagnetic resonance behavior contributes to the magnetic loss at 10–18 ​GHz [38–42]. Surprisingly, unique magnetic vortex domain is firstly observed in the soft magnetic Ni NPs at nanoscale (Fig. 7a). Benefited from the easy magnetization axis and high symmetry of confined sphere spaces, magnetic Ni NPs with vortex domain structure are confined by the carbon shell. With the assistance of electron holography technology, the vortex domain is visually observed in magnetic Ni NPs (Fig. 7b). Surrounded by carbon shell, dispersed magnetic Ni NPs hold the whole magnetic domain structure with the vortex core region about 10 ​nm, and two connected magnetic Ni NPs display similar vortex domain, which marked by the purple circular shape (Fig. 7c). Confined magnetic vortex reversal expanded the efficient absorption frequency. By adjusting the organic ligand into ATA, MOFs-derived Ni@C–V microsphere are assembled by plentiful core-shelleyi@C unit. Compared with other reported MOFs-derived Ni@C EM wave absorption materials, special magnetic vortex structure was firstly observed in the Ni core at nanoscale derived from Ni-MOF-V precursor. Benefited from the chemistry environment and coordination structure from ATA ligand, host nickel ion (Ni2+) was reduced into magnetic NPs, which further promoted the graphitization of organic ligands. In turn, formed carbon shell on the surface of Ni core will limited the expand space with the growth of magnetic Ni NPs. With high symmetry spheres and boundary restriction of graphited carbon shell, confined magnetic vortex structure is generated in the Ni@C–V microspheres powders. However, the Ni@C–H, Ni@C–V and other reported Ni@C magnetic-carbon composites do not show the magnetic vortex structure. The other important factor is the size of magnetic NPs, which is necessary to provide enough space to generate magnetic vortex core and domains. According to EM wave absorption theory, the natural resonance of magnetic Ni@C–V microsphere becomes the main dissipation mechanism below 8 ​GHz. Increased ferromagnetic resonance behavior contributes to the magnetic loss at 10–18 ​GHz [38–42]. Surprisingly, unique magnetic vortex domain is firstly observed in the soft magnetic Ni NPs at nanoscale (Fig. 7a). Benefited from the easy magnetization axis and high symmetry of confined sphere spaces, magnetic Ni NPs with vortex domain structure are confined by the carbon shell. With the assistance of electron holography technology, the vortex domain is visually observed in magnetic Ni NPs (Fig. 7b). Surrounded by carbon shell, dispersed magnetic Ni NPs hold the whole magnetic domain structure with the vortex core region about 10 ​nm, and two connected magnetic Ni NPs display similar vortex domain, which marked by the purple circular shape (Fig. 7c).In order to determine the magnetic vortex domain conformation, the final magnetic structure is reconstructed with phase information images (Fig. 7d–f). Nanoscale Ni cores possess various magnetic vortex configuration with different spin (C ​= ​±1) and polarity (p ​= ​±1), respectively (Fig. 7g–i). The moment and respond behaviors in vortex domain of Ni particles can build a new EM wave dissipation mechanism. In the results of micromagnetic simulation (Fig. 8 ), when the magnetic Ni NPs are magnetized under a high-frequency magnetic field, initial magnetic vortex state quickly changes to other configurations (Fig. 8a-l, Movie S1, Supporting information). Limited by the geometry shape, the magnetized vortex state in spherical magnetic Ni core will change to the “C shape” state as the external magnetic field changes. Incident EM energy can be converted and dissipated in the domain evolution process from the energy lowest point to the unstable energy configuration [43]. Magnetic coupling effect and vortex-domain moment behaviors of magnetic Ni core provide an innovative insight and discussion about the magnetic absorption mechanism [44–46]. Compared with reported Ni-based MOFs-dervied EM absorption materials, MOFs-derived Ni@C–V exhibits outstanding EM energy conversion with the advantages of lightweight, low filling ratio, strong absorption and wide absorption frequency, especially in the X-band (Table S4). As results, constructing vortex domain structure in the MOFs-derived functional materials provides a new adjustment strategy to boost electromagnetic wave energy dissipation.Supplementary data related to this article can be found at https://doi.org/10.1016/j.apmate.2023.100111.The following is the supplementary data related to this article: Multimedia component 2 Multimedia component 2 In summary, a unique magnetic vortex domain structure is firstly observed in the MOFs-derived Ni NPs. Limited by the symmetry spheres and boundary condition, reduced Ni core is confined in the graphited carbon shell, constructing the basic electromagnetic wave dissipation unit. Benefited from the magnetic-dielectric synergy effect, MOFs-derived Ni@C–V microsphere exhibits outstanding EM absorption ability. The RLmin value of Ni@C–V microspheres can reach −54.6 ​dB at 2.5 ​mm. More important, the magnetic coupling network and vortex-domain reversal in the Ni@C–V microsphere together contribute to the expanded efficient absorption frequency up to 5.0 ​GHz at only 2.0 ​mm. Meanwhile, core-shell Ni@C–V units construct plentiful heterojunction Ni–C interfaces and build connected electronic migration routes, which encourages the interfacial polarization and conduction loss. With unique magnetic vortex domain structure and synergy dissipation mechanism, MOF-derived functional microspheres show great application prospects in electromagnetic wave absorption filed.All the chemicals used were of analytical grade without further purification and were purchased from Sinopharm Chemical Reagent Co, Ltd. Typically, 1.5 ​mmol nickel nitrate hexahydrate (0.436 ​g) and 1.5 ​g PVP K-30 are poured into the mixed solution with 10 ​mL distilled water (H2O), 10 ​mL ethanol (EtOH) and 10 ​mL ​N, N-dimethylformamide (DMF). The mixed solution is magnetic stirred for 10 ​min. Then, 0.15 ​g trimellitic acid (H3BTC), 0.15 ​g terephthalic acid (TA), and 0.15 ​g diaminoterephthalic acid (ATA) as the ligand are added into above-mentioned solution, respectively. After another magnetic stirred for 10 ​min, the three solutions are transferred to a 50 ​mL autoclave and keep at 150 ​°C for continuous heating for 12 ​h. The products are taken out and washed with ethanol and distilled water three times, respectively. The solid powders are placed in a vacuum drying oven at 60 ​°C for 12 ​h, and products are marked as Ni-MOFs-H, Ni-MOFs-S, and Ni-MOFs-V, respectively.Obtained Ni-MOFs precursors are annealed in a tube furnace. The calcination conditions are 600 ​°C for 5 ​h in an Ar atmosphere, and the heating rate is 2 ​°C/min. Due to the different ligands, the final black MOFs-derived product magnetic Ni@C powders are marked as Ni@C–H, Ni@C–S, and Ni@C–V, respectively.The crystal structure and components of as-synthesized Ni@C microspheres are characterized by the X-ray diffractometer (XRD, Bruker, D8-Advance, Germany). Raman data is obtained by the Ramoscope (inVia, Renishaw, United Kingdom). The magnetic prosperities and hysteresis loop of MOFs-derived Ni@C powders are tested by vibrating sample magnetometer (VSM, Quantum Design, United States). The morphology and microstructure of the Ni@C microspheres are examined by the scanning electron microscope (FESEM, S-4800, Japan) and transmission electron microscope (TEM, JEM-2100 ​F, Japan). Those Ni@C microspheres are prepared with paraffin matrix with mass ratios 1:4. Then, mixed sample are compacted into a coaxial ring of 7.00 ​mm outer diameter and 3.04 ​mm inner diameter. Related electromagnetic parameters were measured via a vector network analyzer (VNA, N5230C, Agilent, United States) in 2–18 ​GHz range. The reflection loss (RL) values were calculated by the following formula: (3) Z = | Z i n / Z 0 | = μ r / ε r tan h j 2 π f d c μ r ε r (4) R L = 20 log | ( Z i n − Z 0 ) / Z i n + Z 0 | where Z in is the normalized input impedance of absorber, Z 0 is the impedance of free space, ε r is the complex permittivity, μ r is the complex permeability, ƒ is the frequency, c is the light velocity, and d is the thickness of the absorber, respectively.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.Lei Wang and Mengqiu Huang contributed equally to this work. This work was supported by the National Natural Science Foundation of China (52231007, 51725101, 11727807, 52271167, 22088101), the Ministry of Science and Technology of China (973 Project Nos. 2021YFA1200600 and 2018YFA0209100), the Shanghai Excellent Academic Leaders Program (19XD1400400).The following are the Supplementary data to this article. Multimedia component 1 Multimedia component 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.apmate.2023.100111.

Magnetic domain structure plays an important role in regulating the electromagnetic properties, which dominates the magnetic response behaviors. Herein, unique magnetic vortex domain is firstly obtained in the Ni nanoparticles (NPs) reduced from the Ni-based metal-organic frameworks (MOFs) precursor. Due to both the high symmetry spheres and boundary restriction of graphited carbon shell, confined magnetic vortex structure is generated in the nanoscale Ni core during the annealing process. Meanwhile, MOFs-derived Ni@C assembly powders construct special magnetic flux distribution and electron migration routes. MOFs-derived Ni@C microspheres exhibit outstanding electromagnetic (EM) wave absorption performance. The minimum reflection loss value of Ni@C–V microspheres with vortex domain can reach −54.6 ​dB at only 2.5 ​mm thickness, and the efficient absorption bandwidth up to 5.0 ​GHz at only 2.0 ​mm. Significantly, configuration evolution of magnetic vortex driven by the orientation and reversion of polarity core boosts EM wave energy dissipation. Magnetic coupling effect among neighboring Ni@C microspheres significantly enhances the magnetic reaction intensity. Graphitized carbon matrix and heterojunction Ni–C interfaces further offer the conduction loss and interfacial polarization. As result, MOFs-derived Ni@C–V powders display unique magnetic vortex, electronic migration network, and high-performance EM wave energy dissipation.

Catalytic decomposition of methane (CDM) is recognized as a promising approach for the co-production of CO2-free H2 and high value-added carbon nanomaterials (CNMs) [1]. Conventionally, steam methane reforming (SMR) followed by the water-gas shift reaction is one of the most developed processes for large-scale hydrogen generation. Despite the optimizations, SMR is associated with high emissions of CO2 (ca. 12 t CO2/t H2), high capital and operating costs [2]. CDM has the advantage of producing H2 and high value-added carbon nanostructures in a single step without generating greenhouse gases (Reaction 1). In this regard, the CDM process becomes increasingly cost-competitive when public policies support free taxes or negative costs related to CO2 yield and the solid carbon has commercial value [3]. The carbon nanostructures formed in this process include mainly carbon nanotubes [4] or nanofibers [5] and in some cases few-layered graphene or graphite nanosheets [6]. (1) CH 4 ( g ) → C ( s ) + 2 H 2 ( g ) + Δ H 25 ° C 0 = 75.6 . k J / mol Catalysts typically used in the CDM reactions are based on Ni and Fe, with operating temperatures between 500 and 900 °C [7–9]. Although Ni-based catalysts are the most active and stable at temperatures between 500 and 700 °C, it rapidly deactivates with increasing temperature [10,11]. On the other hand, Fe-based catalysts are cheaper and require higher temperatures (700–900 °C) [8,12,13]. This latter range of temperatures provides a positive shift of the thermodynamic equilibrium of the CDM reaction, and thus higher methane conversion may potentially be obtained, as well as an improved structural order in the obtained graphitic nanomaterials [14].The main steps involved in the CDM reaction are: (1) methane cracking, (2) dissolution and diffusion of carbon through the metal particle, and (3) the supersaturation and subsequent precipitation of carbon for the formation of nanostructured carbon [5]. Although the stages of the formation and growth of carbon for Ni- and Fe-based catalysts are certainly related to common factors, there are some differences. The production of as-grown carbon by CDM for Ni catalysts occurs through facet mechanism [15], while for Fe catalysts it is through a complex system of different active phases composed of metallic Fe structures and Fe–C alloys, i.e., Fe3C, α-Fe, γ-Fe, and their alloys [16]. To the best of our knowledge, no studies have been concerned with the influence of each of these components on the results of catalytic reaction, which is an important aspect of the use of iron-based materials in the CDM reaction [17]. A possible reason for this remains in the difficulty to rationalize the factors that lead to the formation of iron phases and metaphase observed under different reaction conditions and catalysts.Catalyst deactivation by carbon encapsulation and sintering is the prime challenge found in the CDM process [18]. To promote a longer catalyst lifetime, different metal loadings [19], supports [20], synthesis methods [21], reactor configurations and conditions [22] have been studied. For example, Inaba et al. [23] investigated Fe-supported alumina catalyst at different temperatures, CH4 flow rates, and CO2 concentrations for the production of carbon nanotubes. CH4 conversion achieved 60% at temperatures higher than 700 °C. They stated that it was possible to increase stability by decreasing gas velocity. By adding CO2 in the feedstock, higher temperatures and longer catalytic lifetimes can be obtained. Besides this, the prereduction of iron oxides using H2 is not mandatory and their reduction can proceed during the CH4 stream at temperatures higher than 680 °C to provide a sufficiently high and stable conversion.Expensive catalysts and synthesis methods can compromise the viability of CDM [2]. Depending on the application of the carbon, it is necessary to purify the obtained carbon by removing the metal with acid treatment [4]. An alternative approach is to regenerate the spent catalyst from the CDM by oxidation to reuse the catalyst [24]. In both contexts, Fe-based catalyst is considered suitable for CDM because of its price. Table 1 presents a literature survey on Fe-based catalysts for CDM with their respective reaction conditions and main catalytic results. Table 1 shows that the studies related to CDM mostly use H2 in the catalyst activation stage; however, from an industrial standpoint, it is desirable to operate with CH4 in the reduction stage to minimize costs. Enakonda et al. [25] studied supported Fe–Al materials for CDM evaluating the reducing atmosphere with CH4 and H2. Interestingly, the catalytic activity using CH4 activation was higher than H2 activation (Table 1). The authors suggested that part of spinel FeAl2O4 was reduced by H2, which may result in the sintering of Fe0 and the lowering of surface area. In contrast, the effect of CH4 and H2 gases as a reducer agent on non-supported iron-based material is still poorly known, thus it was thoroughly investigated in this work.Recently, iron ores have been identified as a promising unconventional catalyst for the CDM reaction to minimize costs [18]. Here, this work explores the use of two different iron ores as catalysts in the production of hydrogen and nanostructured carbon materials via CDM. The iron ores (Tierga and Ilmenite) were chosen because of their low price, wide availability, non-toxicity and catalytic activity in other reactions involving CH4 [26,27]. Both were subjected to CDM reaction for the first time. Tierga iron ore is mainly composed of Fe2O3, and Ilmenite ore contains species of iron and titanium. After the selection of the most active ore (Tierga), several aspects were studied such as reducing atmosphere, reaction temperature and weight hourly space velocity (WHSV) to find an optimal reaction condition providing high catalytic activity and stability. The reduction with CH4 had a positive impact on the structure of Tierga and the yield of as-deposited nanocarbon mainly at more moderate temperatures. We observed that various types of carbon nanostructures such as graphite-like nanosheets and tubular carbon structures with a high degree of graphitization were obtained over Tierga.The iron ore that has iron oxide as its main component was named according to its place of origin, Tierga. The other ore containing iron and titanium was called Ilmenite. Tierga was supplied by PROMINDSA (Tierga, Spain) and the Ilmenite by Titania A/S (Sokndal, Norway). The materials were sieved to 200–300 μm, and then used as a catalyst for the CDM reaction without further treatment.The crystalline structures of the materials were characterized by X-ray diffraction using a diffractometer Bruker D8 Advance Series 2. The powder XRD patterns were further processed for quantitative and qualitative analysis by applying the Rietveld refinement method (see Supplementary materials). The existence of impurities was determined by inductively coupled plasma optical emission spectrometry (ICP-OES; Ametek Spectroblue). Temperature programmed reduction (TPR-H2) tests were performed using an AutoChem Analyzer II 2920. TPR-H2 profiles were acquired using 250 mg of fresh catalyst, under a hydrogen-argon mixture (10% H2) with a flow rate of 50 mL/min from room temperature to 950 °C using a heating rate of 10 °C/min. N2 physisorption experiments were analyzed in a Micromeritics Tristar apparatus. The adsorption and desorption of N2 were determined at −196 °C. Thermogravimetric analysis (TGA) was carried out in a NETZSCH TG 209 F1 Libra thermobalance coupled with the mass spectrometer (MS), OmniStar TM. The sample (ca. 30 mg) was heated from room temperature to 900 °C in a total flow rate of 50 mL/min of methane or hydrogen diluted in argon (10% CH4 or 10% H2) using a heat rate of 10 °C/min. Temperature programmed oxidation (TPO) profiles of the carbon were obtained in the same apparatus from room temperature to 900 °C using a heating rate of 10 °C/min, under an air/nitrogen flow rate of 50 mL/min (25:75 vol:vol). The microstructure of the samples was investigated by transmission electron microscopy (TEM, JEOL-2000 FXII). Raman spectra were measured in a Horiba Jobin-Yvon LabRAM HR800 UV spectrometer equipped with a charge-coupled detector. The degree of graphitization of carbon was measured using the Raman and XRD results. From the characteristic peaks of carbon from XRD data, it was possible to obtain the interplanar distance (d 002 ) between graphene layers of diffraction peak (002) using the Bragg equation. The graphitization index, g , was calculated using Equation (2) [54]. The layer thickness ( L c ) of carbon was calculated by Equation (3), where λ is the X-ray wavelength, B is the angular width of the (002) diffraction peak at half-maximum intensity (radians) and θ is the Bragg angle for reflection (002). The number of graphene layers ( n L ) was estimated using Equation (4). (2) g = 0.3440 − d 002 0.3440 − 0.3354 (3) L c = 0.89 λ B cos θ (4) n L = ( L c / d 002 ) + 1 The catalytic tests were performed in a fixed bed reactor at different pretreatment and reaction conditions. In a typical run, 600 mg of fresh catalyst was reduced from room temperature to 900 °C for 1 h under H2 or CH4 flow rate of 1.2 L/h. Then, the CDM reaction was carried out using a pure CH4 flow rate of 1.2 L/h, at 800, 850, or 900 °C for 3 h. The samples after the reaction were named according to the reducing atmosphere and the reaction temperature. For example, the sample Tierga reduced with H2 at 900 °C for 1 h was named Tierga-H2, and after CDM at 850 °C it was named Tierga-H2850. The composition of the exhausted gases was determined by gas chromatography (see Supplementary materials). The CH4 conversion [XCH4(%)] is given by Equation (5), where C H 2 = F H 2 / F T × 100 is referred to the percentages of the hydrogen content in the exhausted gases and F H 2 and F T are the H2 molar flow rate and total molar flow rate in the reactor output, respectively. The amount of carbon deposited on the catalyst (gc/gcat) was estimated using Equation (6), where M c is the carbon molar mass (12.0107 g/mol), V m is the CH4 molar volume (22.4 L/mol), QCH4 is the volumetric CH4 flow rate fed to the reactor (1.2 L/h), and t is the run time (h). (5) X C H 4 ( % ) = C H 2 200 − C H 2 × 100 (6) g c = M c V m ∫ 0 t Q C H 4 X C H 4 d t Tierga and Ilmenite presented a non-porous structure with surface area of 5.2 and 0.8 m2/g, respectively. The results of XRD and ICP can be seen in Table 2 , Tierga consisted mainly of iron (III) oxide (α-Fe2O3; hematite) and Ilmenite of pseudobrookite (Fe2TiO5).The reducibility of these iron ores was studied by TPR-H2, and its profile is shown in Fig. 1 . The main peaks were observed in the profile at 440, 690, and 860 °C for Tierga. TPR profiles observed in the literature for unsupported α-Fe2O3 materials were analogous to those observed for Tierga [32], which suggested the following global reduction mechanism: α-Fe2O3 → Fe3O4 → FeO → α-Fe [32]. In Fig. 1, the first peak close to 400 °C was related to the transformation of hematite to magnetite, Fe2O3 → Fe3O4. The existence of peaks above 570 °C in these conditions implied the occurrence of the intermediate FeO phase [32]. After that, the transformation of the Fe3O4 phase to metallic Fe between 500 and 900 °C occurred in a two-step magnetite reduction pathway, Fe3O4 → FeO → Fe [33].Four main peaks centered at 430, 620, 920 and 945 °C were observed for Ilmenite in Fig. 1. The peak at 430 °C corresponded to the transformation of α-Fe2O3 → Fe3O4, followed by the stepwise reduction process previously described. The other stages of the reduction of α-Fe2O3 were overlapped by the changes of Fe–Ti–O. The peak between 500 and 650 °C was attributed to the reduction of: Fe3O4 → FeO, Fe2TiO5 → FeTiO3 (Reaction 7) and Ilmenite-Fe3+ → Ilmenite-Fe2+ [34,35]. Peaks above 900 °C were ascribed to the reduction of FeO → Fe and Ilmenite-Fe2+ → Ilmenite-Fe0 [35]. The H2 consumption of Tierga was five times higher than Ilmenite, 314 and 56 cm3/g, respectively. (7) Fe2TiO5(s) + TiO2(s) + H2(g) ↔ 2FeTiO3(s) + H2O(v) The steps of in situ activation with H2 or CH4 are the same for iron oxide: α-Fe2O3 → Fe3O4 → FeO → α-Fe [25]. However, the in situ reduction of the catalyst with CH4 may differ from that with H2 in the formation of gaseous byproducts throughout the reduction process. By the reduction with CH4, the formation of traces of COx gases is motivated by the reaction between CH4 and the oxygen of the catalyst from the metal oxide and support [25]. Regarding the properties of the catalyst, it had been pointed out that the reducing agent can modify the type of as-grown carbon [36]. Given these aspects, it is expected that the catalyst undergoes different transformations in particle size, sintering and carbon deposit when reduced with CH4. The study of the reducing atmosphere effect was conducted with Tierga because it has a greater amount of active phase.The evaluation of byproduct formation during the reduction step of Tierga was carried out in a thermobalance using 30 mg and a flow rate of 50 mL/min containing 10% of CH4 or 10% of H2 in Ar. The gases evolved were analyzed by mass spectrometry. Fig. 2 shows the variation of each gas during the experiment, as well as the sample mass variation and temperature. The appearance of CO and CO2 mainly occurred at approximately 700 °C for both atmospheres (CH4 and H2) due to the decomposition of the dolomite phase that takes place at that temperature [37]. The profile of CO and CO2 occurred differently between the two pretreatments because an additional formation of COx gas was expected from the interaction between CH4 and catalyst between 600 and 900 °C as aforementioned. Simultaneously, the reduction with CH4 can lead to other interactions between CH4 and the byproducts formed during the reduction (H2O, CO2, CO, H2), i.e., the gas-water shift reaction, steam and dry reforming of CH4. This can be evidenced by the diverse water vapor profiles between CH4 and H2 reduction pretreatments. For Tierga pretreated with H2 (Fig. 2-b), the water vapor profile had maximum peaks at 490, 710, and 810 °C, similar to that observed in the TPR-H2 (Fig. 1). However, the water vapor profile for CH4 reduction pretreatment material had more discrete peaks with maximum peaks at 740 and 880 °C (Fig. 2-a), suggesting a CH4 reforming reaction along the reduction stage as also indicated in another study [38]. Additionally, it is worth mentioning that neither CO nor CO2 was found at 900 °C for Tierga-H2 or Tierga-CH4, in good agreement with the previously reported results [25].When the reduction with CH4 was carried out in a fixed bed reactor, the onset of the CH4 decomposition reaction and the formation of byproducts became more evident (Fig. 3 ). The CH4 decomposition started after 30 min of reduction at 900 °C as the CH4 conversion increased abruptly. The profiles of CO and CO2 were similar to those seen in the experiments using thermobalance (Fig. 2-a). The same experiment was not carried out with H2 at fixed bed because the analysis of gaseous byproducts during the reduction with H2 had already been verified in Fig. 2-b and the CDM reaction proceeds only by contacting methane. According to these reducibility tests, the reduction step for both atmospheres was established up to 900 °C after 1 h. The reduction stage for 1 h at 900 °C is called the activation step hereafter.XRD patterns for Tierga pretreated with H2 or CH4 at 900 °C for 1 h are shown in Fig. 4 . Both materials presented characteristic peaks of the α-Fe phase (ICSD 64998), SiO2 (ICSD 42498) and new peaks regarding CaO (ICSD 673084) and MgO (ICSD 88058) phases resulting from the decomposition of dolomite. The absence of iron oxides peaks indicated the complete reduction to α-Fe. Fe3C (ICSD 064689) and graphite (ICSD 76767) were also identified in Tierga-CH4, suggesting that the CDM reaction started during the reduction step with CH4. In fact, Zhou et al. [33] demonstrated that CDM starts with the formation of Fe3C and graphite simultaneously on the surface of α-Fe through the reaction between Fe and CH4 (Equation (8)). As soon as Fe3C is formed, it acts as a catalyst and promotes the methane decomposition into H2 and carbon [33]. The carbon diffuses into Fe3C to form supersaturated Fe3C1+x, which is unstable and immediately decomposes back to stoichiometric Fe3C and graphite carbon [33]. (8) 3Fe (s) + 2CH4 (g) ↔ Fe3C (s) + C (s) + 4H2 (g) The concentration and mean crystallite size of α-Fe depended on the pretreatment performed. The percentage of the α-Fe phase in Tierga-H2 and Tierga-CH4 catalysts were 85 and 73 wt%, respectively. The lower concentration of this active phase in Tierga-CH4 was explained by the transformation of this phase into Fe3C. Moreover, the mean crystallite size of α-Fe in the Tierga treated with H2 (80 nm) was bigger than the iron ore activated by CH4 (46 nm), which indicated the fragmentation of the α-Fe phase in Tierga-CH4 into smaller crystals by the adjacent formation of iron carbide and graphite as previously reported in other catalysts [39,40]. Another possibility involved the effect of Fe3C and carbon on the catalyst activity. From literature, the iron carbide and carbon can act as textural promoters and prevent the sintering of α-Fe particles, positively impacting in the catalytic activity, as already reported for Ni-based materials during CDM above 500 °C [41]. To confirm this hypothesis and to probe in more detail the ability of these structures to act as promoters in Tierga, different temperatures were used in the reaction (see Section Effect of temperature). Other works also report that the reduction with H2 is more severe and leads to larger crystallite sizes by sintering [42]. Once the catalyst is reduced along the activation stage, the reaction step proceeds.After pure H2 prereduction at 900 °C for 1 h, Tierga and Ilmenite were subjected to a reaction with pure CH4 at 800 °C. Fig. 5 shows the profiles of the samples in terms of H2 production (left y axis) and CH4 conversion (right y axis) during the reaction. Only H2 and CH4 gases were detected during the reaction. The catalytic activities of both declined with time on stream, and after 1 h it increased. A more detailed discussion of this behavior was made in Section Effect of temperature. Due to the lower Fe loading and higher reduction temperature of the Fe–Ti–O structures, Ilmenite exhibited worse catalytic activity than Tierga in terms of H2 concentration ranged from 15 to 18%, than those of Tierga with 46–47%. The CH4 conversion ranged from 8 to 10% for Ilmenite, and 30–32% for Tierga. Tierga produced a high carbon yield at 800 °C (0.82 gc/gcat and 1.6 gc/gFe), indicating that is a promising natural catalyst to be used in CDM and, consequently, it was conducted to further experiments.The influence of the space velocity at 850 °C, with CH4 activation, and WHSV ranging from 2 to 6 L/(gcat∙h) was evaluated and the corresponding H2 concentration and CH4 conversion evolutions over Tierga catalyst are shown in Fig. S1. With decreasing the space velocity, there was a gain in contact time and consequently, an increase in conversion. The H2 content profile slightly rose for a WHSV of 2 L/(gcat∙h). When increasing the WHSV to 4 and 6 L/(gcat∙h), the catalyst underwent a deactivation process after 1 h of reaction. The trend of the WHSV of 2 L/(gcat∙h) will be described in detail in the following section.The effect of the operating temperature on Tierga-CH4 and Tierga-H2 activities was evaluated at 800, 850 and 900 °C using WHSV = 2 L/(gcat∙h). H2 concentration and CH4 conversion changes are shown in Fig. 6 -a. A significant increase in the amount of produced H2 was obtained with rising temperature for both catalysts: Tierga-H2 and Tierga-CH4. According to literature, the amount of produced H2 by CDM increases as the temperature increases and the pressure falls [7]. High H2 concentration (70%) and no deactivation were observed for Tierga-H2 and Tierga-CH4 at 850 °C. At 800 °C stable conversion was observed for the catalyst treated with H2 during about 100 min, followed by slowly rose to 32%, while the CH4 conversion increased from 24 to 40% after 3 h of reaction for the catalyst treated only with CH4. At 900 °C, although it exhibited the highest initial catalytic activity, there was a slight deactivation after the first hour of reaction for both Tierga-CH4 and Tierga-H2 (ca. 10% H2 decay). The decrease in the catalytic activity of Tierga-CH4900 and Tierga-H2900 after 1 h can be primarily assigned to the encapsulation of the active phase.The early period of catalytic activity, immediately before the period of constant carbon growth, is commonly named the induction period. This step in CDM is usually associated with carbon migration and saturation in catalysts, and metal reconstruction [15]. In Figs. 5 and 6-a, the samples Ilmenite-H2800 and Tierga-H2 (at 800, 850 and 900 °C) showed an initial drop of H2 production and CH4 conversion between 10 and 50 min. This fall may be related to the period necessary for carbon supersaturation of α-Fe and Fe3C to take place. Such induction period tended to decrease with rising temperature over Tierga (Fig. 6-a). After carbon supersaturation, carbon precipitation occurs. Regarding Tierga-CH4, the active structures were already partially saturated and therefore had an increasing trend of catalytic activity. The low initial concentrations of Fe3C and graphite were not sufficient to make these catalysts act as a structural promoter at the beginning of the reaction. As there was an increase in the concentration of Fe3C and carbon, they could act as support and possibly explain the high stability at 800 and 850 °C. Fig. 6-b summarizes the amount of formed H2 and the conversion of CH4 after 3 h of reaction as a function of temperature. The final conversion of CH4 at 800 °C was about 35% for Tierga-H2 and Tierga-CH4. At higher temperatures, the conversion was close to 56% and it was independent on the treatment of Tierga. This result revealed that the initial fragmentation and previous saturation with carbon observed in the XRD pattern (Fig. 4) had a positive impact on the catalytic results mainly at 800 and 850 °C after 3 h of reaction. Such initial catalyst fragmentation with CH4 may have brought about the inhibition of agglomeration and sintering of iron-based materials. This disaggregation likely led to greater exposure of the active phase which resulted in higher catalytic activity for Tierga-CH4 catalyst, as other authors previously reported [43].The amount of deposited carbon from these CDM experiments is shown in Table 3 . Despite the slight difference between the results for the same temperature, the carbon formation was favored with CH4 as the reducing agent and with rising temperature. Comparing the carbon yield of Tierga with data taken from the literature (Table 1) is a non-trivial task owing to the diversity of experimental systems. In some cases, Tierga has superior performance than iron-based synthetic catalysts (e.g., 100% Fe2O3), which contribute to boosting the competitiveness of Tierga iron ore to reach a commercial level. On the other hand, Tierga material displayed inferior carbon yield than other ones possibly due to an absence of support and a small number of alkaline impurities such as potassium and sodium (Table 2) as previously reported [20]. The experimental conditions used in this work and the results obtained for Tierga without H2 pretreatment could be considered as a good advantage for industrial application. In addition, Tierga presents other advantages such as low-cost and high Fe loading. Fig. S2 shows the diffractograms of the Tierga catalysts after the reaction. The spent catalysts were composed mostly of α-Fe (ICSD 64998), γ-Fe (ICSD 185721), Fe3C (ICSD 064689) and graphite (ICSD 76767) phases in all samples except for Tierga-H2800 sample that did not have the pattern of γ-Fe. The as-deposited carbon presented d002 values between 0.3376 and 0.3366 nm and g p between 0.74 and 0.86, respectively (Table 3), i.e., parameters close to the perfect single crystal of graphite structure, which is 0.3354 nm and g p close to 1. The characterization of carbon by XRD indicated the formation of graphite-like materials with L c between 17 and 21 nm and a number of graphene layers ( n L ) between 52 and 64. Due to the low carbon formation over Ilmenite (0.3 gc/gcat), only Tierga catalysts were characterized after CDM.Most samples after the reaction were composed of the γ-Fe structure. This phase is less characterized experimentally due to its instability at temperatures below the boiling point (727 °C). γ-Fe can be an intermediate phase in the production of Fe3C and graphite at high temperatures [44]. The α-Fe (body-centered cubic system) and γ-Fe (face-centered cubic system) phases have a great affinity with carbon, which allows the dissolution of carbon atoms in the network of these metals, reaching a maximum of 0.022% wt. of C at 740 °C for α-Fe, and 2.14% wt. of C at 1150 °C for γ-Fe. The diffractogram of γ-Fe without carbon saturation found in the literature (ICSD 41506) had peaks at 2θ = 45.8, 53.4 and 78.9°; however, the peaks at 2θ = 43.8, 50.9, 74.9° presented in Fig. S2 can be attributed to γ-Fe saturated with carbon (ICSD 185721) [45,46]. This is because γ-Fe allows the insertion of carbon in the interstices of the crystalline network. The rearrangement decreases part of the associated metal-metal energy and changes the diffraction lines of γ-Fe metal to lower angles, as observed for Tierga. Similar results have been reported in earlier publications [45,46]. The carbon-saturated γ-Fe phase was observed only after reaction (Fig. S2), and not in the initial activation step (Fig. 4), which suggests that enough carbon was formed during the reaction to protect and stabilize this intermediate phase.Based on the XRD data, the iron-based phases were quantified by Rietveld refinement (Fig. 7 ). The amount of Fe3C decreased with increasing temperature, while iron species increased. The most striking variation in the final composition of the iron phase between the materials took place at 800 °C: Fe3C was the major product in Tierga-H2800, while α-Fe and γ-Fe become dominant in Tierga-CH4800. However, this difference between the catalysts gradually decreased up to 900 °C. These results indicated that the characteristics of the catalyst after diverse activation atmospheres led to distinct reaction mechanisms at moderate temperatures motivated by the generation of iron phases with distinct crystal systems and fractions.The most widely reported reaction mechanism is based on the transformation of α-Fe into Fe3C and graphite (Equation (8)). In contrast, previous studies have shown that the mechanism of carbon formation from α-Fe can vary according to the concentration [47] and crystallite size [45] of the α-Fe phase in reactions performed at the same temperature. Wirth et al. [47] revealed that depending on the concentration of α-Fe in a temperature range close to the eutectic temperature (700–800 °C), γ-Fe or Fe3C can be obtained, the latter would give rise to carbon. While for Takenaka et al. [45], the α-Fe structure was transformed into Fe3C or γ-Fe depending on the crystallite size of iron oxide. The supported Fe2O3 crystallites with smaller sizes were transformed into Fe3C, while larger ones were transformed into γ-Fe saturated with carbon atoms [45]. Based on these studies, it became evident that for Tierga with reaction taking place at 800 °C (close to the eutectic point), Fe3C nucleation was favored when the active phase of the catalyst was mainly composed of α-Fe with larger crystallite size, i.e., Tierga-H2 catalysts. Yet at 800 °C, the γ-Fe phase was preferably promoted by a system with a lower concentration of α-Fe and smaller average crystallite size (Tierga-CH4 catalysts). As the reaction temperature overpassed the eutectic point towards higher temperatures for other catalysts (Tierga-CH4850, Tierga-H2850, Tierga-CH4900 and Tierga-H2900), there was a higher tendency to promote the nucleation of γ-Fe [47]. Once α-Fe, γ-Fe, or Fe3C appeared, the carbon dissolution begins to happen and when it reaches the supersaturation of carbon in the metal and/or carbide, the precipitation and growth of carbon occur.Correlating the XRD results with the catalytic tests for Tierga-CH4800 and Tierga-H2800 it was possible to evaluate the effect of the different active phases (α-Fe, γ-Fe and Fe3C) on the conversion and H2 production. The concentration of 55% H2 (v/v) was obtained in CDM after 3 h for Tierga-CH4800. As Tierga-CH4800 was composed mainly of α- and γ-Fe at the end of the reaction, it seems to indicate that α- and γ-Fe phases were more effective catalysts than Fe3C. A possible explanation for these results may be that the carbide requires a higher amount of carbon for supersaturation than the metal, maximum of 6.67% wt. of C for Fe3C [17]. While α-Fe and γ-Fe require a lower amount of carbon for graphite precipitation to occur, usually less than 3% wt. of C [17]. In addition, the carbide bulk diffusion coefficient is lower than that of the metals γ-Fe and α-Fe, implying a higher difficulty in precipitating graphite using carbide [48]. Thus, the mitigation of carbide formation resulted in greater activity of the catalyst, and it was achieved by changing the activation atmosphere to CH4. As seen in the XRD results, the activation with CH4 led to the initial fragmentation of the α-Fe phase and inhibition of large amounts of Fe3C. Fig. 8 shows the TEM images of the as-grown carbon from Tierga with different pretreatments and CDM reaction conditions. TEM images confirmed as-deposited carbon in all spent Tierga in the form of carbon nanomaterials (CNMs) with a high degree of graphitization (d 002  = 3.35 Å), including multi-layered graphene, graphite nanosheets (GNSs) and carbon nanofilaments. The nanofilaments were multi-walled carbon nanotubes and chain-type carbon nanofibers.In all samples, the GNSs structures (marked with white dotted rectangles) appeared in higher quantities. Generally, they were transparent, rippled graphene/graphite layers, and disengaged from the metallic particles (Fig. 8-a, c). Tubular structures (marked with black dotted rectangles) were sparser and shorter, without (Fig. 8-f) and with (Fig. 8-g) encapsulated iron-based nanoparticles. The chain-type carbon nanofibers (Fig. 8-g) had multiple graphite walls around the metal, similar to those observed in previous works with Fe [23]. The metallic particles in the images were round and covered with a thin layer of graphite (Fig. 8-c). As the temperature increased, the agglomeration of the metal particles increased (Fig. 8-e) as well as the number of metal particles within the chains (Fig. 8-f).The nanocarbon structures such as GNSs and carbon nanofilaments observed in this work can be explained by the quasi-liquid state theory [16,20]. According to some studies [50,51], iron species in the quasi-liquid state combined with the absence of support can produce GNS. Iron-based species with low dispersion and large particles when in quasi-liquid state elongate and expand to form a thin film composed of metal and carbide metal [51]. This film is capable of allowing the dissolution, precipitation of carbon and growth of graphene or graphite sheets on its surface [51]. The formation and growth of short carbon nanotubes and chain-type carbon nanofibers observed in the TEM images may have happened analogously. We can infer that the segregation of the active phase during the reaction enabled the formation and growth of these carbons, as also noticed in some previous works [51,52]. The carbon was precipitated out from the smallest metallic iron particle supersaturated with carbon. The growth occurs in a cylindrical shape and extends to maintain the void inside the tube [7,20,33]. With its growth, the interface between the carbon and metal walls decreases and the insertion of the metallic particle into the tube or chain may occur during this process, thus forming the carbon nanotube or chain-type carbon nanofibers [7,20,33].Raman spectra of the as-deposited carbon nanostructures over the Tierga catalysts are presented in Fig. S3. In the Raman first-order spectra (1100 and 1700 cm−1) of the materials, it is possible to observe the characteristic peaks of disordered graphite including D, G, D′ at 1350, 1580 and 1620 cm−1 respectively. The second-order (2500–3300 cm−1) is the result of overtones and combinations of the bands in the first order, and for the studied materials, peaks were observed in approximately 2450, 2720 and 3240 cm−1, which are attributed to the first overtone of bands at 1220, 1350, 1620 cm−1, and the band 2950 cm−1 is a combination of band G and D. The 2D band (~2700 cm−1) is characteristic of structures with few and multiple layers of graphene and graphite. Analogous spectra are found in the literature for multilayer graphene and graphite [53].The integral intensity ratio I D /I G is widely used to express the degree of graphitization for the carbon, i.e. the lower I D /I G ratio, the higher crystalline order of the carbon species. The average parameters of the spectra are shown in Table 3. The I D /I G values of all samples were all below 1 (Table 3), which means that the carbon is ordered, with minor contributions from disordered particles.The low I D /I G value corroborates TEM images, showing that the carbon nanostructures were predominantly composed of multilayer graphene or graphite nanosheets and small quantities of nanofilaments. The results presented in this work agreed well with other studies in which low I D /I G was favored when the final product was multilayer graphene flakes, high temperatures, and flows of pure methane [36,54,55]. Compared to synthetic pure Fe2O3 reported in the literature [14], Tierga generated hybrid carbon with fewer defects and a higher amount of carbon. The materials showed I D /I G results close at the same reaction temperature (Table 3), however, the most significant difference was between the materials Tierga-CH4800 and Tierga-H2800. The I D /I G value was 0.19 for Tierga-CH4800 and 0.25 for Tierga-H2800, which means that the material Tierga-CH4800 was nanostructured with fewer defects than Tierga-H2800. These same materials showed the greatest difference in carbon yield, 1.05 gc/gcat for Tierga-CH4800 and 0.82 gc/gcat Tierga-H2800. This result is in concordance with previous studies [56] which suggested that one of the conditions for carbon growth is preventing disordered the carbon formation. The quality of produced carbon depended on the catalyst treatment and interestingly the results were better with the treatment of the catalyst with CH4 which is an advantage in the development of the CDM industrial process.Finally, TPO was performed to evaluate the thermal stability of the spent Tierga catalysts (Fig. 9 ). In all profiles, it is first observed that there was a slight gain in mass close to 450 °C, which may be related to the oxidation of the metallic iron and iron carbide phases located on the surface, followed by a sharp decay in mass between 600 and 630 °C. The higher the reaction temperature, the greater the displacement of the oxidation temperature to higher temperatures. This result is consistent with what was observed in Raman and TEM, which indicates a highly ordered crystal structures (500–700 °C) with the absence of amorphous carbon (~400 °C).Tierga and Ilmenite were confirmed as an active catalyst in the production of CO2-free H2 and carbon. However, Tierga showed significantly higher catalytic results than Ilmenite and was therefore further investigated. The methane conversion and hydrogen concentration over Tierga were 56% and 70%, respectively, after 3 h of reaction. Tierga reduced with CH4 demonstrated superior performance with greater activity and stability than Tierga pretreated with H2 at moderate temperatures. CH4 activation has contributed to the fragmentation of the active phase α-Fe which led to smaller crystallites preventing agglomeration and sintering. Such characteristics also promoted the formation of γ-Fe rather than Fe3C. The high stability of Tierga can be primarily associated with a high degree of graphitization. At 900 °C, there were no significant differences between the Tierga materials in terms of the conversion and reaction mechanism, however, the deactivation started after a certain time, which is related to the encapsulation of chain-like carbon nanofibers. XRD, TEM and Raman revealed the production of structures with nanosheets of graphite and carbon nanotube structures with a high degree of graphitization. WHSV and reaction temperature play a central role in the stability of this material as well, in which the optimal conditions were 2 L/(gcat∙h) and 850 °C. The use of iron ore as a natural and low-cost catalyst in the production of nanocarbon structures can contribute as an alternative to assessing the practical use of the CDM process.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.Brazilian funding to support this work was provided by CNPq [Process 141308/2018-4], FAPESP [Process 2018/01258-5], CAPES [Finance Code 001]. Spanish funding was provided by the European Regional Development Fund and the Spanish Economy and Competitiveness Ministry (MINECO) [ENE2017-83854-R]. Authors would like to acknowledge the use of “Servicio General de Apoyo a la Investigación-SAI, Universidad de Zaragoza”. The authors also thank PROMINDSA and Titania A/S for providing the iron ore used in this work.The following is the Supplementary data to this article: Multimedia component 1 Multimedia component 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.ijhydene.2021.08.065.

Tierga and Ilmenite Fe-based ores are studied for the first time in the catalytic decomposition of methane (CDM) for the production of carbon dioxide-free hydrogen and carbon nanomaterials. Tierga exhibits superior catalytic performance at 800 °C. The effect of the reaction temperature, space velocity and reducing atmosphere in the catalytic decomposition of methane is evaluated using Tierga. The highest stability and activity (70 vol% hydrogen concentration) is obtained at 850 °C using methane as a reducing agent. Reduction with methane causes the fragmentation of the iron active phase and inhibits the formation of iron carbide, improving its activity and stability in the CDM. Hybrid nanomaterials composed of graphite sheets and carbon nanotubes with a high degree of graphitization are obtained. Considering its catalytic activity, the carbon quality, and the low cost of the material, Tierga has a competitive performance against synthetic iron-catalysts for carbon dioxide-free hydrogen and solid carbon generation.

With the rapid development of social economy, the environmental pollution caused by the over-utilization of traditional energy is becoming more and more serious. In order to reduce the environment damage caused by fossil fuels, it is urgently needed to develop the green sustainable energy [1–4]. Due to the advantages of direct methanol fuel cell (DMFC) with low environmental pollution, high energy conversion efficiency, easy storage and transportation, it is expected to become an effective clean energy that can be widely used in electric vehicles or portable electronic devices [5–7]. However, the slow kinetics of methanol oxidation reaction (MOR) is the main challenge that hinders its further commercialization, and quite a few attempts have been made to solve this problem. At present, the most widely used catalyst for MOR in DMFC on the market is Pt-based catalysts [8–9], while there are still many problems to be solved. For example, they can be easily poisoned by intermediates such as CO produced during the MOR process, which reduces the efficiency of the catalyst [10–12]. Moreover, the high cost of Pt is another shortcoming which limits its further development [13–15]. It is very important to find low-cost, high catalytic performance and stable catalysts to replace precious metal catalysts [16]. It is reported that some transition metal and its oxide catalysts shows enhanced reaction kinetics and stronger anti-poisoning ability for MOR. At present, Ni/NiO [17–19] and Cu/CuO [20] catalysts are widely used in alkaline DMFC, which show excellent catalytic performance for the electrooxidation of methanol. For these catalysts, the oxygen-containing functional groups adsorbed on transition metal oxides (NiO or CuO) can effectively remove reaction intermediates, thereby improving anti-toxicity and obtaining higher stability [21–23]. In addition, compared to other expensive transition metals, Cu and Ni, as metal precursors, are inexpensive and have a wide range of sources.Another successful strategy to improve the performance of electrocatalysts is to load heterogeneous metal nanoparticles on the catalyst to form the composite [24–25]. One widely employed example is Au nanoparticles (AuNPs). The surface of AuNPs can be treated as a highly negative electron absorber to promote the oxidation of transition metal cations to higher oxidation states [26]. Anchoring AuNPs on the surface of nanomaterials hopefully facilitates electron transfer rate, thereby improving the electrochemical catalytic efficiency for MOR. AuNPs have been used to form the nanocomposite with metal oxides such as SnO2 [27], ZnNb2O6 [28], CeO2 [29] so as to improve the catalytic activity.In this work, via solvothermal method, nanoplates composed of copper and nickel oxides with different valence states were synthesized. AuNPs were absorbed onto the nanoplates by introducing them into the Au colloid. And the catalystic activity towards MOR was studied in potassium hydroxide electrolyte. XRD and XPS characterizations prove that the Cu and Ni in the nanoplates are multivalent state. TEM and EDS-mapping characterizations show that a large number of AuNPs are uniformly absorbed onto the ultra-thin multivalent Cu-Ni oxide nanoplate. Electrochemical tests show that the prepared electrocatalyst has high electrochemical activity and excellent stability towards the catalytic performance for MOR.Chloroauric acid tetrahydrate (HAuCl4·4H2O), nickel (II) nitrate hexahydrate (Ni(NO3)2·6H2O, 99%), potassium hydroxide (KOH, 99%), copper nitrate hexahydrate (Cu(NO3)2·6H2O 99%) and hexamethylenetetramine (C6H12N4, 99%) were all purchased from Aladdin (Shanghai, China). In the experiment, ultra-pure deionized water (18.2 MΩ) was used. All reagents are analytical reagent grade, no further purification is required.Energy dispersive X-ray spectroscopy (EDS) system (Oxford X-Max, UK), X-ray photoelectron spectroscopy (XPS, Thermo Fisher Scientific, Escalab 250Xi), transmission electron microscopy (TEM, Tecnai G2-20, American FEI company), X-Ray diffractometry (XRD, Rigakuultima iv, Japan). The XRD spectrums were recorded at 2θ values ranging from 30° to 80°.Dissolve 44.6 mg of nickel (II) nitrate hexahydrate, 53.8 mg of copper nitrate hexahydrate and 95.3 mg of hexamethylenetetramine in 68 mL of methanol to form a homogeneous dispersion. The resulting dispersion was then transferred into a 100 mL autoclave and heated to 120 °C for 9 h. After cooling, the sample was filtered out and washed three times with mixed solution of ethanol and water (volume ratio 1:1) by centrifugation under 8000 rpm for 10 min to remove residual impurities, and dried in a vacuum drying oven at 60 °C for 12 h. The obtained products were put into a muffle furnace, heated to 300 °C (5 °C per minutes) and kept at 300 °C for 4 h to prepare multi-valent Cu-Ni oxide nanoplates.Au colloids were prepared according to the literature [30]. For the AuNPs decoration, 40 mg of m-v oxide was dispersed in the Au colloid and stirred for 2 h. The products were then collected and washed in ultrapure water to remove impurities. AuNPs decorated multi-valent Cu-Ni oxide (AuNPs/m-v oxide) was finally dried at 60 °C for further use.For modification of the working electrode, the GCE (glassy carbon electrode) was firstly polished with alumina powder and cleaned before use. 4 mg of AuNPs/m-v oxide was dispersed in 1 mL of deionized water and ultrasonic for 10 min to get uniform dispersion. 3 µL of the dispersion was dropped onto the surface of the pre-treated GCE to get the AuNPs/m-v oxide/GCE. Finally, 5 µL Nafion solution (1%) was covered on the electrode surface for seal. For comparison, M−v oxide/GCE was also fabricated with the similar procedure by replacing AuNPs/m-v oxide with m-v oxide. All electrochemical measurements were performed on CHI 660D with a three-electrode electrochemical system. The MOR tests were carried out in 1 M potassium hydroxide with or without methanol. Platinum foil and calomel electrode (with saturated potassium chloride) were served as counter electrode and the reference electrode.Transmission electron microscope was used to characterize the morphology of the prepared samples. Fig. 1 A is the TEM image of m-v oxide. It is found that a certain number of thin nanoplates are overlapped with each other. Fig. 1B is the high resolution TEM image of m-v oxide, it shows a couple of well-resolved and interlaced fringes with interplanar distances of 0.233, 0.276, 0.213, 0.230 nm, which are assigned to the lattice distance for the (111) planes of CuO, the (110) planes of NiO, the (200) planes of Cu2O, the (102) planes of Ni2O3 respectively.After the obtained nanoplates are dropped into the Au colloid and stirred for 2 h, plenty of dark dots are emerged on the surface of the nanoplates, as shown in Fig. 1C. This proves that the AuNPs are successfully decorated onto the oxide. In addition, the HRTEM image of AuNPs/m-v oxide (Fig. 1D shows the atom lattice fringe for the dark dot is 0.235 nm, which corresponds to the Au (111) plane. This can be further inferred that the AuNPs/m-v oxide composite was successfully prepared. Fig. 2 is the XRD patterns of the prepared AuNPs/m-v oxide (curve a) and m-v oxide (curve b), as well as the standard spectrums of Au (curve c), NiO (curve d, PDF#47–1049), CuO (curve e, PDF#48–1548), Ni2O3 (curve f, PDF#14–0481) and Cu2O (curve g, PDF#34–1354). As compared to the standard spectrums, both the characteristic peaks for NiO, CuO, Ni2O3 and Cu2O are all appeared in the XRD spectrums of AuNPs/m-v oxide (curve a) and m-v oxide (curve b). It is worth noting that for the AuNPs/m-v oxide, the diffraction peaks of Au (38°, 44.3°, 64.5°, 77.5°) are partially overlapped with the diffraction peaks of m-v oxide, indicating that the deposited Au nanoparticles are small and highly dispersed on the surface of m-v oxide [31].In order to understand the elemental composition and valence state information of the composite material, the XPS measurement was carried out and the corresponding results are shown in Fig. 3 . Fig. 3A shows the whole spectrum of AuNPs/m-v oxide (curve a) and m-v oxide (curve b). Fig. 3B-3E is the amplified binding energy spectrums for C 1 s, Au 4f, Ni 2p and Cu 2p. Binding energy in all spectra is calibrated based on the carbon standard binding energy of 284.6 eV. In Fig. 3B for C 1 s, beside the standard C 1 s binding energy peak at 284.6 eV from adventitious reference carbon, the binding energy of 287.6 eV, whichis observed both in AuNPs/m-v oxide (curve a) and m-v oxide (curve b), corresponds to the signals of C-O and OH-C = O bonding[13,32]. The peaks of Au 4f XPS spectrum (curve a in Fig. 3C at 84.0 eV and 87.7 eV are allocated to Au 4f 5/2 and 7/2, which is the typical Au0 valence state [33]. However, in the XPS spectrum of m-v oxide (curve b in Fig. 3C, these is just a smooth baseline in these energy range. This difference confirms the formation of AuNPs for the AuNPs/m-v oxide. In Fig. 3D for Ni 2p, there are two accompanying satellite peaks at 861.2 eV and 879.5 eV, and the other four peaks. These peaks are related to Ni 2p 3/2 and 1/2 [34], and confirms the existence of Ni2+ and Ni3+ [35–37]. In Fig. 3E for Cu 2p, the binding energy at 932.5 eV and 952.6 eV are the peak of the Cu+ 2p 3/2 and 1/2 orbital, and the peaks at 934.4 eV and 954.6 eV are related to the 2p 3/2 and 1/2 orbital peaks of Cu2+. These peaks prove the co-existence of Cu+ and Cu2+ in the oxide [38]. The XPS results are consistent with the XRD results and further prove that there is multiple valence states for Cu and Ni in the prepared nanomaterials, which is contribute to the oxidation process to obtain better catalytic performance.The highly dispersed structure of nanoparticles in the catalyst helps to improve the catalytic performance for methanol oxidation. Fig. 4 shows SEM elemental mapping images for C (A), O (B), Ni (C), Au (D) and Cu (E) in AuNPs/m-v oxide.It can be observed that C, O, Ni, Au and Cu signals are all detected in the scanning area, and the dispersion in the sample are very uniform. Combining the above XRD and TEM results, it can be fully proved that the AuNPs/m-v oxide is successfully synthesized with good crystallinity.The anodic oxidation peak current density is usually used to evaluate the catalytic ability of catalyst. Herein, the electrocatalytic performance of two nanomaterials on MOR is evaluated by comparing their electrochemical performance. Fig. 5 is the electrochemical response of the AuNPs/m-v oxide and m-v oxide modified electrodes in 1 M potassium hydroxide electrolyte containing 0.5 M methanol. In the blank 1 M potassium hydroxide electrolyte, the AuNPs/m-v oxide (curve c) and m-v oxide (curve d) modified electrode show no obvious redox peak. After the addition of 0.5 M methanol, both the AuNPs/m-v oxide (curve a) and m-v oxide (curve b) modified electrode exhibit obvious oxidation current at potential of 1.20 V, and the oxidation peak current density (21.1 mA/cm2) for AuNPs/m-v oxide/GCE is 3.4 times bigger than m-v oxide/GCE (6.2 mA/cm2). This confirms that the m-v oxide contributes to the electrochemical catalysis of MOR, and the existence of AuNPs further promotes this process. In addition to the catalytic current density, the onset potential is another important parameter for the electrochemical catalyst. The onset potential for MOR on the AuNPs/m-v oxide/GCE is about 0.5 V, which is much lower than the m-v oxide/GCE (0.7 V). The enhanced electrocatalytic performance of AuNPs/m-v oxide may be attributed to the factor that the well-distributed AuNPs can act as electron absorbers to promote the oxidation of Cu and Ni cations [25,39–40], making metal ions reach higher oxidation state. The higher oxidation state stimulates the rapid charge transfer on the electrode/electrolyte interface and improves the catalytic activity in MOR.To study the kinetics property, the relationship between the peak current density and scan rate of AuNPs/m-v oxide/GCE and m-v oxide/GCE was investigated, which is shown in Fig. 6 . The CV response of different modified electrodes in 0.1 M potassium chloride electrolyte with 2.0 mM Ferri/Ferro-Cyanide at different scan rates (from 30 to 100 mV/s) was recorded. It can be seen from the results that both for the AuNPs/m-v oxide/GCE (Fig. 6A and m-v oxide/GCE (Fig. 6B, the peak current is continuously enlarged as the scanning rate increases. For AuNPs/m-v oxide/GCE (Fig. 6C, the linear equation between the anode peak current density value and the square root of scan rate is j (mA/cm2) = 0.1526 v 1/2 –0.66, and the correlation coefficient R2 = 0.9923. The linear equation between the cathode peak current density and the square root of scan rate is j (mA/cm2) = − 0.1305 v 1/2 + 0.4397 with the correlation coefficient R2 = 0.9952. For m-v oxide/GCE, the linear equation between the anode peak current density and the square root of scan rate is j (mA/cm2) = 0.0558 v 1/2 –0.0938, the correlation coefficient R2 = 0.9981. The linear equation for the cathode peak current density is j (mA/cm2) = − 0.0366 v 1/2 –0.0855, the correlation coefficient R2 = 0.9833. The relationship between the peak current density and the scan rate indicates that both for the AuNPs/m-v oxide/GCE and m-v oxide/GCE, the kinetics process is diffusion control process in the reaction [41].The CV response of AuNPs/m-v oxide/GCE and m-v oxide/GCE in 1 M potassium hydroxide electrolyte with different concentrations of methanol (0.3 ∼ 1.0 M) were tested, as shown in Fig. 7 A and 7B. It can be observed that for the two modified electrodes, the peak current density increases as the methanol concentration increases. As the methanol concentration increases from 0.3 to 1.0 M, the anodization peak current density for m-v oxide/GCE is increased from 3.7 to 7.4 mA/cm2, while this value is increased from 17.0 to 30.5 mA/cm2 for AuNPs/m-v oxide/GCE. Fig. 7C shows the linear relationship between the oxidation peak current density of the two catalysts and the methanol concentration, the slope values of AuNPs/m-v oxide and m-v oxide are 19.5 and 5.1 respectively. A larger slope value indicates a higher electron transfer efficiency of the catalyst [41]. The bigger peak current density and larger slope value proves the electrochemical catalytic effect of AuNPs/m-v oxide is better than that of m-v oxide, which is coincident with the above results. This phenomenon can be attributed to the following reasons: First, via the bifunctional mechanism, the mixing multivalent metal oxides can effectively adsorb substances such as hydroxyl groups, and can convert CO-like intermediates into CO2, thereby improving electrocatalytic performance [42]. In addition, with the excellent electronic conductivity, the existence the AuNPs can accelerate the electron transfer rate between metal oxides and electrolyte. Third, the AuNPs absorbed on the surface of the metal oxide can collect negative electrons to a large extent, which promotes the oxidation of Cu and Ni cations to higher chemical valence state, so as to improve catalytic efficiency in MOR [25].In order to evaluate the catalytic performance of nanocomposite, the comparison between this work and other published catalysts is shown in Table 1 . Comparing the experimental results for AuNPs/mv oxide, it shows higher current density than other catalyst reported in Table 1 in alkaline conditions, confirming the higher electrochemical catalytic performance for methanol oxidation reaction in alkaline conditions.The reproducibility of the AuNPs/m-v oxide catalyst was studied by investigating the oxidation peak current of five AuNPs/m-v oxide/GCE in 1 M potassium hydroxide containing 0.5 M methanol. These electrodes were modified under the same condition for the CV test. The relative standard deviation (RSD) of peak current density for five electrodes was calculated to be 2.75%, which means a remarkable reproducibility.In practical applications, the stability of the catalyst is also an important parameter for its commercial application, therefore, the long-term stability of the catalyst is tested by monitoring the i-t curve of AuNPs/m-v oxide/GCE and m-v oxide/GCE in potassium hydroxide (1 M) electrolyte containing 0.5 M methanol at 1.20 V for 7200 s. Comparing the current curves obtained from two modified electrodes in Fig. 8A, a sharp drop in current density is clearly observed in the first period of time, it may be due to the formation of double-layer capacitors, which is a kinetic process. As the intermediates continue to accumulate and occupy the active sites of the catalyst during the methanol oxidation process, the current slowly decays, and finally reaches a steady state. The current density of AuNPs/m-v oxide stabilizes at 1.016 mA/cm2 after 6000 s, which is 1.24 times higher than the steady current density of m-v oxide (0.823 mA/cm2). The higher steady current density of AuNPs/m-v oxide indicates the stronger anti-poisoning ability in the electro-oxidation process. Fig. 8 B shows the slopes of the Tafel plots of AuNPs/m-v oxide and m-v oxide, the slopes of the two catalysts are 41.87 and 81.81 mV/dec respectively. All these results show that AuNPs/m-v oxide nanocomposite has excellent catalytic activity and anti-poisoning performance against various toxic intermediates [20,52].Cu-Ni oxide with the decoration of AuNPs was effectively synthesized through traditional solvothermal method and surface absorption. The morphology, structure, chemical valence and element composition of the nanomaterial were characterized by a series of techniques like TEM, XRD, XPS and EDS. XRD and XPS characterizations prove that the Cu and Ni are existent with multi-valent state in the oxide. Different nanomaterial modified electrodes of Cu-Ni oxide/GCE and AuNPs/Cu-Ni oxide/GCE were fabricated in the procedure and employed to investigate the electrochemical catalytic effect in MOR. Electrochemical experiments confirmed that AuNPs/m-v oxide has better electrochemical performance than m-v oxide. Via the bifunctional mechanism, together with the excellent electronic conductivity of AuNPs, the AuNPs/m-v oxide exhibits outstanding electrochemical catalysis effect in MOR process. It shows a high oxidation peak current density of 21.1 mA/cm2, good tolerance and stability to toxic intermediates. Further works are still carrying on improving the performance by optimizing the component dose and surface modification of the nanocomposites. What’s more, considering the easy fabrication and high catalytic efficiency, it provides a promising alternation of MOR electrochemical catalyst for further catalytic application.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.Thanks for financial support from the National Natural Science Foundation of China (Grant Nos. 21861018), Natural Science Foundation of Jiangxi Province (20202ACBL214012, 20182BCB22010), Jiangxi Provincial Key Laboratory of Functional Molecular Materials Chemistry (20212BCD42018), the Youth Jinggang Scholars Program in Jiangxi Province, Key Laboratory of Testing and Tracing of Rare Earth Products for State Market Regulation and Qingjiang Excellent Young Talents Program of Jiangxi University of Science and Technology.

Cu-Ni oxide nanocomposite was synthesized via solvothermal method with nickel (II) nitrate and copper nitrate as Ni and Cu resource respectively. AuNPs/oxide nanocomposite was obtained by dispersing the oxide into the pre-synthesized Au colloid. Transmission electron microscope (TEM) shows that the synthesized oxides are nanoplate morphology. Interplanar distance from the high resolution TEM (HRTEM) image shows that the Ni and Cu in the oxide are multivalent state. X-ray diffraction (XRD) and X-ray photoelectron spectroscopy (XPS) characterization further proves that the Ni and Cu are multivalent state, and the Au is Au0 form in the nanocomposte. EDS-mapping images confirm large numbers of AuNPs are uniformly absorbed onto the ultra-thin multivalent Cu-Ni oxide nanoplate. The electrochemical properties of multivalent oxide (m-v oxide) with or without AuNPs decoration are investigated in 0.1 M potassium chloride electrolyte with 2.0 mM Ferri/Ferro-Cyanide. When employed for the electrochemical catalysis in methanol oxidation reaction (MOR), the AuNPs/m-v oxide exhibits better electrochemistry performance than m-v oxide. The current density of AuNPs/m-v oxide (21.1 mA/cm2) is 3.4 times bigger than that of m-v oxide/GCE (6.2 mA/cm2). What’s more, it shows good electrochemical stability and lower slope of the Tafel plot (41.87 mV/dec).

There is a broad variety of possible applications for dopant transition metal oxides with different morphological structures. The possibility of tailoring their electronic properties by doping and quantum containment of carriers is one main feature that has made these materials special. The semiconductor materials have drawn significant focus due to their possible applications for instance catalysts, and sensors for preventing environmental problems [1–5].Among many semiconductor materials, Mn3O4 (hausmannite) has gained research focus due to its unique structural properties together with intriguing physicochemical properties that are of high significance for battery applications [6]. It is usually prepared by hydrothermal pathway, carburization technique, chemical decomposition, ball milling, sun-freezing, self-combustion, co-precipitation, chemical reduction [7–13]. Furthermore, Mn3O4 nanostructures are not only utilised for potential application, but also play a significant part in electromagnetic applications.Microwave-assisted combustion techniques have increased control over the size and shape of the synthesized nanomaterials and have created similar and mono-dispersed nanomaterials. Increased control over the scale and shape of the end Ni-doped Mn3O4 is the advantage of this approach over other approaches. The combustion method is very facile and only takes a few minutes, and it has been widely applied to the preparation of various nano-scale oxide materials. This synthetic technique makes use of the heat energy liberated by the redox exothermic reaction at a relative low igniting temperature between the metal nitrates and urea or other fuels. Furthermore, the combustion method is also safe, instantaneous and energy saving. The advantage of microwave combustion method is that it leads to a highly exothermic reaction, which in turns led to a direct formation of spherical particles. The microwave heating causes the uniform distribution of temperature between the surface and the bulk material, and there by leading to the fast formation of nanoparticles [14–17].The aim of the present research is to focus on the preparation of Ni-doped Mn3O4 by a user-friendly microwave combustion method and in addition the examination of magnetic, optical, morphological, and structural properties.In a standard synthesis process, 6.0 ml of Mn(NO3)2 solution (0.4 M) is added to Ni(NO3)2 solution. In addition to that, 8.0 ml of urea (0.4 M) was introduced to Mn(NO3)2, and then the contents continuously stirred. Manganese precursor alongside nickel nitrate and urea (fuel) was put within a domestic-type convection microwave oven (Make - IFB; Design − 20SC2) and subsequently exposed to microwave energy for ten minutes with a power of 1200 Watt together with 2450 MHz microwave frequency. The uniform resulting substance initially began to boil followed by vaporization occurred as well as the releasing of the gases whilst the microwave combustion. If the substance of chemical sources is prolonged to the automated stage of combustion, the substance will evaporate and right away produce a solid. The solid material obtained was properly washed with ethanol, after that dried at 80 °C for 2 h, and lastly, the obtained materials labeled as Mn3O4 (a) Ni-doped at 0.01(b), 0.02(c) and 0.03(d).The nickel doped and pristine Mn3O4 samples X-ray diffraction patterns were scanned with the Rigaku X-ray diffraction meter utilizing CuK radiation. High-resolution scanning electron microscopy (HR-SEM) pictures were obtained by the Philips XL30 ESEM, that is installed together with energy-dispersive X-ray spectroscopy. High-resolution transmission electron microscopy (HR-TEM) pictures had been captured through the Philips EM 208 transmission electron microscope by applying an accelerated voltage of 200 kV. The UV–visible (Cary100) and also the Cary Eclipse Fluorescence Spectrophotometers had been utilized for recording diffuse reflectance spectra as well as optical properties for the Mn3O4 nanostructures prepared by present approach.XRD was tested for finding the crystalline structure of Mn3O4 and Ni-doped Mn3O4 and in addition, the doping behavior of Ni 2+ in the different samples as shown in Fig. 1 (a). The sharp peaks having 2-theta values 32.32, 36.09, 38.03, 44.39, 50.71, 58.45, 63.84, and 67.65 can be assigned to the (101), (112), (103), (211), (004), (220), (105), (321), (224) and (400) planes of Ni-doped Mn3O4. Diffraction peaks corresponding to the structure of the Mn3O4 hausmannite could also be distinguished (JCPDS No. 24–0734). However, the characteristic peaks attributed to Mn3O4 are mainly sharp and broad which may be due to the crystalline structure of synthesized Mn3O4. It is also noted that the most intense peak is gradually shifted to a higher angle and is widened by increasing the Ni content as clearly shown in Fig. 1(b) of this. This widening and shifting of the diffraction peak with Ni doping strongly recommend that Ni ions have successfully replaced the Ni ion with the Mn3O4 lattice.There are a number of reasons for this feature, such as inversion of the cation distribution (Mn3+ into tetrahedral interstices and Ni2+ into octahedral interstices). When the doped amount of Ni was increased to oxide peak at about intensity peaks 48° (marked by asterisk) was detected. The crystal structure and their physical properties depend on the Mn site symmetry. Mn3+ ion has an electronic configuration of t2g 3 eg0. It usually occupies the intensity peaks octahedral sites in spinel structure which theoretically has no orbital angular momentum. Similarly, Mn ion has a zero orbital momentum at the B site with electronic configuration of t2g3 eg2 in the spinel structure. One would expect that of intensity peaks Ni substitution and these opposite trivalent spins are found to destroy the structure. However, Ni substitutions can induce intensity peaks variations of the exchange coupling into this mixed spinel system. In other words, one understands that Mn valence is reduced in the Td (i.e. from trivalent to divalent Mn) site with a complete reaction at dopant becomes a normal spinel-like (intermediate spinel) structure [18,19].In addition, the observed line widening of the diffraction peaks indicates that the synthesized materials are within the nanometer range. The full width at the half-maximum (FWHM) of the major peaks increases with an increase in doping concentrations, which can be attributed to a decrease in crystalline size. The peak position and the FWHM are obtained by fitting the measured peaks with two Gaussian curves in order to find the true peak position and width corresponding to monochromatic Cu Ka radiation. Since the systematic error decreases as the Bragg angle increases, the values of the average grain size and the lattice parameter of the samples were calculated for the reflection peak which possesses both higher angle and reasonable intensity. As the particle size decreases, the crystal lattice becomes less aligned leads to broadening of the XRD pattern. Hence, there is an inverse relation between NPs size and the sharpness of the XRD peaks [20] . The mean crystallite for all samples was measured by employing the Scherrer equation [21]. L = 0.89 λ β cos θ Wherein, the mean crystal size (Å) and X-ray source wavelength (1.5404 Å) are denoted as L and λ correspondingly. The full width at half maximum (FWHM) and diffraction angle of the respective peaks (radians) are labelled as β and θ respectively. A possible factor for the trend of decreasing crystallite size is that the Ni concentration increasing of the continuation for the ordering process among crystallite size is currently observed in Fig. 1c . The crystal size of all the samples is in the order of a nanometer. This confirms that the Ni-doped Mn3O4samples are nanomaterials and that the size of the crystals varies depending on the doping.The explanation for the smaller sizes of the Ni-doped samples is assumed to be the fact that Ni, which is added within the device, is settled within the lattice of Mn3O4 and thus forms bonds with the unstable oxygen atoms of Mn3O4 or due to oxygen desorption. This means that Ni doping can lower the nucleation rate of Mn3O4 and in turn, doping atoms can influence the size of the particles. Thus with Ni doping, Ni-doped Mn3O4 samples are smaller in size than the undoped Mn3O4 [22]. This difference due to the strain, crystallite size, and lattice parameters that can induce a greater broadening in the diffraction peak (microstructural and microstrain) for the broadening peaks Fig. 1d. Diffraction studies can be helpful when it is important to understand the state of the chemical mixture of the elements involved or the steps in which they exist. The advantage of the diffraction method over traditional chemical analysis is that it is simpler, requires only a relatively small amount of sample, and is therefore non-destructive.UV–visible absorption technique is an effective non-destructive instrument for the analysis of optical properties. Absorption spectra of undoped and Ni-doped Mn3O4 were recorded in the 200–800 nm wavelength range as shown in Fig. 2 . Absorption spectra depend on many factors, such as oxygen vacancy, morphology, bandgap, and impurity. The cut-off wavelength region is 400–600 nm, indicating the photo-excitation of electrons from the valence band to the conduction band. It is also observed that the absorption edge has continuously changed to a higher wavelength (yellow-green shift) as the Ni content rises to the host Mn3O4 and as a result, the size is reduced [23,24].In general, the PL properties of Ni-doped Mn3O4 are highly determined by their defect states (i.e. defect speeds, defect load states, defect concentrations, etc.). The PL emission spectra and the findings of the samples are shown in Fig. 3a . As seen in the diagram. 3b , the emission peak of the Ni-doped Mn3O4 is shifted to a higher wavelength and its emission peak varies from 520 to 550 nm. However, the intensity of this function is very poor and size-independent [25,26]. On the other hand, the nano-sized Ni-doped Mn3O4 allows for a further degree of independence and size reliance on extreme PL emissions. High band-edge luminescence is consistent with a low dislocation density as well as a low surface defect density, as these defects appear to quench the band-edge radiative recombination ( Fig. 3c ). Since the size-related PL spectra intensity of nano semiconductors can be quantified, it is possible to calculate an optical particle size using the intensity shift measured from the emission spectrum. The crystallite size from PL spectra was calculated from line broadening of the with two Gaussian curves in order to find the true peak position line by the using the Scherer equation ( Fig. 3d ). The shift in the PL peaks could be assigned because of the restriction of size and intensity by the Ni. PL peaks to many interdependent factors, for instance, lattice dislocation, electron–phonon coupling, localization of charge carriers because of the point defects, and interface effects [27–30]. The weakness of the deep-level emissions also indicates that these nanoparticles have a stoichiometric structure, likely having a low density of point defects. The findings from PL studies are compatible with the SEM/TEM and XRD.On the other hand, once location oxygen vacancies had been presented upon the decreased metal oxides, various superficial oxygen-vacancy levels surface preceding and partially overlapping by way of the VB of dopant metal oxides (mainly O2 p and M3 d , with a few places. Additionally, the increase of the VB to CB can easily likewise direct result in the widening concerning and the size of the VB. This kind of theoretical function suggested that oxygen vacancies may perform the function of changing the valence band of metal oxides and dopant oxides. It is also noticed that both emission peaks are gradually moved to longer wavelengthswith Ni content growth. This redshiftwas well correlated with the bandgap narrowing, as demonstrated by the UV absorption spectra. The green emission mechanism can be understood as follows. From the above case, PL spectra have a thin, wider emission band and are mainly used for the application of luminescence as they cover a wide range of the spectral area [31].The morphology and structure of the Ni-doped Mn3O4 NPs are described by a high-resolution electron microscopy (HR-TEM) study. A distinctive HR-TEM image is shown in Fig. 4 . HR-TEM images show nanoparticles with a length of 100–110 nm and a width of 3–5 nm. The nanoparticles observed are formed by aggregation of NPs and the samples are also agglomerated due to the magnetic nature of the materials. For the nanopowder doped material produced according to this method, this tendency toward different aggregation/agglomeration levels and the creation of small agglomerates from a few particles is also noted. The mean particle size value (TEM size, 39–23 nm) is nearer to the average crystallite size (XRD size, 31–27 nm), which is calculated from the XRD data in accordance with the Debye-Scherrer formula ( Fig. 5 ). Minor differences are caused by the measurement or calculation errors. This difference in grain size is explained by the fact that in TEM, only a few grains are examined and measurements are conducted manually, but in XRD, grains are regarded as a whole as well as calculations are performed by applying specific formulas [32]. The overall examination of TEM micrographs shows that certain latex compounds create a covering around the particles, preventing them from agglomerating and therefore contributing to the stability of the nanoparticles. The observed slight difference in particle size value as estimated from the two different techniques (XRD and HR-SEM) may be due to some structural disorder and strain in the lattice resulted from different ionic radii and/or clustering of the nanomaterials [33] . An in-depth high-resolution electron microscopy analysis was performed to examine the nanoscale fine structure of the Ni-doped Mn3O4 device. The high-resolution TEM (HR TEM) image of a single nanoparticle has been taken and is shown in Fig. 6 . It is apparent from the figure that the nanoparticle is single crystalline in nature, with transparent lattice fringes that can be seen on the whole particle. The bulk of the particles are observed to be mainly faceted to crystalline.M−H hysteresis cycles have been reported using a vibrating sample magnetic scale operated at a applied magnetic field of 10 k Oe at room temperature to estimate the impact of Ni conversion on residual magnetization, compulsion and saturation magnetization. Fig. 7 Magnetic parameter values such as saturation magnetization (MS), coercivity (HC), and remanence (Mr) are seen in Table 1 . Factors such as cation replacement, grain size, and A-B exchange interactions have been shown to have a significant effect on the magnetic properties of oxide materials. Increased grain size and decreased A-B superexchange interaction cause canting spins on the surface of nanoparticles that minimize the magnetic characteristics of the samples in question. The magnetic parameters indicated suggest the lack of the soft magnetic nature of Mn3O4 by the replacement of nickel [34].The electrical conductivity (σ) was determined through the electrode area (A) and also from the sample thickness (t) of the sample by employing the following equation. σ = t AR In which R is the calculated resistance. The electrical conductivity was determined to be in the order of 10−2 S/ cm, which revealed a progressive tendency with the dopant concentration until 0.03%. The incorporation of Ni on pristine Mn3O4 samples lattice was observed to be beneficial in order to increasing the electrical conductivity on the whole temperature range of measurement.When 0.03% Ni is doped, a decrease in conductivity is observed. The maximum conductivity value of 2.13 × 10–2 S/cm observed for Mn3O4 undoped and Ni doped Mn3O4 samples at 500 °C was found to be around two times higher than that of pure Mn3O4 sample (0.98 × 10−2 S/cm).The activation energy was calculated using the Arrhenius' equation given below: σ = σ o exp - E a kT → Wherein Ea is the activation energy of electrical conduction, k can be called Boltzmann constant, T is generally denoted for temperature and σo is the pre-exponential factor. The activation energy needed for the movement of the charge carriers was determined through the Arrhenius plot which is displayed in Table 2 . The activation energy reduces by increasing in temperature because of the thermally initiated mobility of charge carriers from one atomic site to another according to hopping conduction mechanism [35]. It could be noticed that the activation energy of pristine Mn3O4 and Ni doped Mn3O4 structure reduced until 0.03% which showed the minimum activation energy of 0.69 eV when compared to pristine Mn3O4 (0.87 eV). Because of the spinel structure's porosity, Ni ions may be dispersed over the nanoparticle’s surface (as demonstrated by the TEM and XRD studies in the present work). Even if an attempt is made to include a very high dopant concentration in such nanoparticles, the local energy of the dopant is greater than the energy in the larger part of the semiconductor. Thermodynamically, it may be disadvantageous for the dopant to stay in the nanoparticle as a dopant and be excluded from the “quality” of the semiconductor on the surface of the particle [36,37].The catalytic test of pure Mn3O4 synthesized by and Ni-doped Mn3O4 is proved as follows. In 100 ml beakers (five sets), 0.1 mmol−1 concentration of 4-nitrophenol aqueous solution (50 ml) was injected concurrently with 0.529 mol/dm3 concentration of H2O2 (made at the time for 50 ml). To the above mixture, synthesized different Mn3O4 catalysts (0.001 mol) were added into the different beaker. For comparison, one beaker does not contain any catalyst. The above mixture were continueously stirred. For each of the sets, the periods necessary for full decolorization from yellow colour were recorded and in Table 3 the values were summarized. It has been found that 4-nitrophenol color changes earlier in the presence of a catalyst than in the absence of a catalyst. Among them, Ni-doped Mn3O4 (0.03%) showed the greatest catalytic activity (decolorization takes place in 15 min) in comparison to the pure and other Ni-doped Mn3O4 samples. It is significant to point out that the concentration of Ni has direct impact on the size, DRS, VSM and PL emission spectra. It has been pointed out that among all the factors, the surface chemical state is the most significant factor since it is related to the electron transfer and separation efficiency at the interface. The presence of an appropriate amount of Ni ions on the catalyst surface plays an important role on improving catalytic performance [38–40]. Though, it is significant to assume that despite the doping concentrations used in the present work, there are some interactions between them. Regardless of the dopant concentration applied in this research, it is crucial to expect that they interact.Ultra-fine Mn3O4 Ni-doped nanoparticles were effectively synthesized through microwave techniques. Ni substitution has resulted in a remarkable increase in the structural, optical, and magnetic properties of Ni-doped Mn3O4. Structural, phase, purity, particle size, was studied by powder XRD, TEM, HR-TEM, and VSM spectral analysis, respectively. The Crystalline size of the synthesized nanoparticles was measured using the Debye-Scherer equation, which decreased from 31 nm to 27 nm. PL spectra intensity structural defects in the lattice, which including oxygen vacancy, replacement, and interstitial atoms, which cause localized chargelosses. The effect of doping leads to a decrease in crystallite size. The size of the crystal grain depending on certain doping-induced nucleation centers and the growth stress disruption as a result ofthe difference in the atomic radius among Mn and Ni dopants.The author P. Tamizhdurai have affiliation with university madras direct or indirect financial interest. The authors extend their appreciation to the Deanship of Scientific Research, This research was funded by Princess Nourah bint Abdulrahman University Researchers Supporting Project Number (PNURSP2022R19), Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia. Also the author Dr Abualnaja appreciated Taif University Researchers Supporting Project number (TURSP-2020/267), Taif University, Taif, Saudi Arabia. We further confirm that the order of authors listed in the manuscript has been approved by all of us.We wish to confirm that there are no known conflicts of interest associated with this publication.This statement is signed by all the authors to indicate agreement that the above information is true and correct.This research was funded by Princess Nourah bint Abdulrahman University Researchers Supporting Project Number (PNURSP2022R19), Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia. Also the author Dr Abualnaja appreciated Taif University Researchers Supporting Project number (TURSP-2020/267), Taif University, Taif, Saudi Arabia.

Ni-doped Mn3O4 nanoparticles (NPs) were synthesized by a simple one-pot microwave combustion procedure utilizing urea as a fuel. X-ray diffraction, transmission electron microscopy (TEM), diffuse reflectance spectroscopy, Photoluminescence spectra, and vibrating sample magnetometer. The particle size and the crystalline size measured from the HR-TEM monographs and XRD study suggest the similarity of the data collected from these two measurements. Photoluminescence (PL) spectra demonstrated increased luminescence amplitude with increased Ni concentration. Thus, the present study determines the time required for 4-nitrophenol yellow to colorless by Ni-doped Mn3O4 and Mn3O4 samples.

The increasing request for energy, chemicals, and materials globally due to megatrends needs an intervention to ensure that these supplies are produced efficiently using carbon-neutral processes and renewable feedstocks. Under this emerging socio-economic picture, biomass exploitation allows for a broad range of biofuels and value-added biochemicals (Attard et al., 2020; Clark, 2019). A desirable and practically unused candidate to obtain these commodities is almond hulls, i.e., an unavoidable food waste produced in large amounts in the almond processing industry. Almond production has augmented noticeably recently, e.g. global production has risen from 0.7 Mt. in 2007 to 1.5 Mt. in 2020 (Driedfruit.net, January 2022; Factfish.com, January 2022). This lignocellulosic material is also known as almond green shell cover and accounts for more than half of the total almond weight (Esfahlan et al., 2010). It comprises carbohydrate matter (cellulose, hemicellulose and pectin), along with lignin, tannin-like poly-phenolic compounds and ash (Esfahlan et al., 2010).One way to improve food security is by decreasing food waste to reduce environmental impacts; developing sustainable processes to efficiently manage this unavoidable food waste by-product is necessary (Dutta et al., 2022; Mak et al., 2020). However, little work has been conducted, and the publications reported to date primarily covered the extraction of some constituents (saccharides, bioproducts and/or phenolic species) rather than addressing strategies to valorise the entire material (Remón et al., 2021b). For example, Pinelo et al. (2004), Sfahlan et al. (2009) and Sfahlan et al. (2009) extracted phenols from almond hulls to produce antioxidants. Concerning the production of carbohydrates, Offeman et al. (2015) conducted hot water extraction experiments to recover the carbohydrate content of hulls obtained from different varieties of almonds. Ferrandez-Villena et al. (2019) used almond hulls to manufacture renewable particleboards, while González et al. (2005) analysed the combustion of different almond hulls. These works provide valuable insights to recover some almond hulls constituents, but they do not provide alternative solutions for integral valorisation within a biorefinery strategy. Therefore, it is still required to develop new routes to achieve complete and more sustainable management of the whole material. Among the different approaches for biomass valorisation, hydrothermal hydrogenation might be an attractive route to transform it into a plethora of valuable liquids for the energy sector and pharmaceutical, cosmetic and textile industries. These include water-soluble oligomers, oligosaccharides, saccharides, sugar alcohols, polyhydric alcohols, carboxylic acids, furans and nitrogen and phenolic compounds (Ribeiro et al., 2021).Hydrothermal hydrogenation uses water at hydrothermal conditions. It is conducted at moderate temperatures (150–300 °C) and relatively low H2 pressures (20–80 bar) in the presence of a catalyst (Ribeiro et al., 2021). Therefore, not only hydrogenation reactions take place. In contrast, a series of cascade, ‘one-pot’ transformations, mainly comprising hydrolysis, hydrogenation, and hydrogenolysis reactions co-occur. This ‘one-pot’ strategy aids in shortening the reaction time, diminishing unwanted by-products formation, and helps reduce the consumption of energy, solvents and reagents. Concerning the catalyst, metal (Fe, Co, Ni, Pd, Pt, Ru, Rh, Ir, Ag and Au) supported catalysts have frequently been used (Ribeiro et al., 2021). Ru is considered one of the most promising active phases when balancing crucial factors such as activity, selectivity, deactivation resistance and price. In particular, Ru-based catalysts are very active in hydrogenation reactions. They are economically more competitive than other metals such as Pt and Au and suffer from lesser deactivation than Ni-based catalysts. Carbonaceous materials have commonly been used as supports due to their outstanding chemical immovability in non-oxidising media and bespoke surfaces combined with stunning textural properties (Cardoso et al., 2021; Cardoso et al., 2018; Ochoa et al., 2018; Remón et al., 2016a; Zhu et al., 2013). Besides, carbon-neutral processes can be used for their synthesis (Pinilla et al., 2017; Remón et al., 2021a), linking very well with the emerging biorefinery and circular economy ideologies (Dutta et al., 2022; Mak et al., 2020). Active metals supported on carbon materials have been used as catalysts for the hydrogenation of biomass-derived compounds.Notwithstanding these excellent features, work addressing the use of catalysts based on Ru supported on carbonaceous materials for the hydrothermal hydrogenation of biomass is scarce. For example, Matsagar et al. (2020) used a commercial Ru (5 wt%)/C catalyst for the hydrogenation of furfural to tetrahydrofurfuryl alcohol at mild reaction conditions. In another work, Dutta et al. (2019) critically reviewed the work conducted on the hydrogenation of levulinic acid to produce γ-valerolactone (GVL) over noble metal catalysts. This review also shows that the use of Ru-based catalysts is scarce. Even though these and other works in the literature have addressed the hydrothermal hydrogenation of biomass structural compounds, it is essential to note that the reactivity of biomass could be different from that of these carbohydrates alone. This accounts for several interactions between species taking place in the upgrading process. For example, Ribeiro et al. (Ribeiro et al., 2016; Ribeiro et al., 2017b) found synergetic interactions between cellulose and hemicellulose during their hydrothermal hydrogenation over a 0.4 wt% Ru/CNT catalyst. Remarkably high sorbitol (74%) and xylitol (76%) yields were achieved during the co-valorisation of both carbohydrates through a two-step procedure, firstly 2 h at 170 °C, followed by 4 h at 205 °C. Besides, it is difficult to predict the behavior of actual biomass based on its structural composition due to different factors affecting its recalcitrance (Pu et al., 2013). This information suggests that it is vital to address the hydrothermal hydrogenation of actual biomass. Still, work conducted on the hydrothermal hydrogenation of actual biomass using carbon-supported metal catalysts is not very well reported. Most publications have focused on producing sugar alcohols and glycols from forestry and agricultural residues at some fixed processing conditions.For sugar alcohols, Palkovits et al. (2010) addressed the hydrothermal hydrogenation of spruce wood chips over a Ru supported on carbon (Ru/C) catalyst at 160 °C and 50 bar H2 pressure, achieving a biomass conversion of 59% with a sorbitol yield of 36%. Guha et al. (2011) used a 2 wt% Ru supported on activated carbon (Ru/AC) catalyst for converting beet fibre. They reported an arabitol yield as high as 83% when the process was conducted at 155 °C for 24 h, using 50 bar of H2. Käldström et al. (2011) hydrothermally hydrogenated bleached birch kraft pulp over a (Ru/C) catalyst at 185 °C and 20 bar of H2 for up to 30 h, leading to the production of xylose (up to 0.12 mol/mol biomass) and glucose (up to 0.03 mol/mol biomass). Zhou et al. (2015) transformed Jerusalem artichoke tube into hexitols (32% mannitol and 61% sorbitol) by hydrothermal hydrogenation over a 3 wt% Ru/AC-SO3H catalyst, conducting the process at 100 °C and 60 bar H2 for 5 h. Yamaguchi et al. (2016) studied the hydrothermal hydrogenation of different feedstocks (Japanese cedar, eucalyptus, bagasse, empty fruit bunch and rice straw) over a 3 wt% Ru - 1 wt% Pt/C catalyst, achieving a total sugar yield as high as 55%. Ribeiro et al. (2017a) used cotton wool, cotton textile, tissue paper and printing paper to produce sugar alcohols at 205 °C and 50 bar of H2 with a 0.4 wt% Ru supported on carbon nanotubes (Ru/CNT) catalyst. Except for printing paper, complete substrate conversion was attained with all materials, with sorbitol yields ranging from 51 to 56% in all the cases. Li et al. (2018) achieved a high polyols production during the hydrothermal hydrogenation of cornstalk (25% sorbitol, 12% xylitol and 5% ethylene glycol) and beechwood (18% sorbitol, 13% xylitol and 4% ethylene glycol) over Ru/C at 200 °C and 30 bar of H2 for 8 h.Regarding biomass conversion into glycols, Ribeiro et al. (2021) converted different waste materials (Eucalyptus wood, corncob, cotton wool and tissue paper) into ethylene glycol, using a 0.4 wt% Ru/CNT catalyst at 205 °C, 50 bar H2 for 5 h. The biomass conversion and ethylene glycol yields were as follows: eucalyptus wood (X = 74%, Y = 25%), corncob (X = 91%, Y = 14%), cotton wool (X = 100%, Y = 42%) and tissue paper (X = 100%, Y = 34%). In another work, Pang et al. (2018) used a 5 wt% Ru/AC catalyst for glycols production from ball-milled Miscanthus at 245 °C and 50 bar H2 for 6 h. Under such conditions, the whole material was converted into a mix of valuable liquid products, including ethylene glycol (34.6%), 1,2-propylene glycol (8.1%), glycerol (8.8%), 1,2-butanediol (9.7%) and sorbitol (4.3%). Li et al. (2018) converted cornstalk into polyalcohols and alkyl cyclohexanes over different Ru/C catalysts. The Ru/C catalyst reduced at 300 °C showed the best performance when the process was conducted at 200 °C and 30 bar H2 for 8 h. The molar yield of alkyl cyclohexanes was 97.2%, with a total polyalcohol yield of 52.7% (24.5% sorbitol, 12.2% xylitol, and 16.0% C2–C4 polyols).These publications afford valuable information on the hydrothermal hydrogenation of biomass using Ru catalysts supported on different carbonaceous materials. Although some work has been conducted on the synthesis of Ru/CNF catalysts, these were not employed to this end. For example, Yang et al. (2016) synthesized several Ru/CNF catalysts with different Ru loadings (0.09–0.64 wt%) by the one-pot conversion of Ru-functionalised metal-organic framework fibres. These were tested in the hydrogenation of lactic acid (Lac-Ac) to γ-valerolactone (GVL), with the best results (96% Lac-Ac conversion with 95% GVL yield) being achieved using the 0.27 wt% Ru/CNF catalyst. Nevertheless, to the best of our knowledge, carbon nanofibres (CNF) have never been employed to synthesise Ru-based catalysts to this end. Also, the effect of the processing conditions is not yet well understood using actual biomass. In previous work in our research group, Frecha et al. (2019) used a 0.4 wt% Ru/CNF catalyst for the hydrothermal hydrogenation of cellobiose (a cellulose model compound). The effect of the processing time (0−3 h) was analysed on the cellobiose conversion and reaction products distribution at 180 °C employing an initial H2 pressure of 4 bar. Complete cellobiose conversion was attained within the first 30 min of reaction, with a sorbitol yield progressively increasing from 5 to 46% over the course of the reaction, showing the promising properties of this catalyst for the hydrothermal hydrogenation of biomass due to its excellent activity in hydrolysis/depolymerisation and hydrogenation reactions.Given this background, this work explores, for the first time, the hydrothermal hydrogenation of almond hulls (a lignocellulosic unavoidable food waste) over a carbon-neutral Ru/CNF catalyst, with a very low (0.4 wt%) Ru content. Initially, the effects of the processing conditions on the distribution of the products (gas, liquid and solid) and the detailed chemical composition of the liquid phase have been thoroughly analysed. Additionally, a possible reaction pathway has been developed to explain the formation of the most representative liquid products. Then, the process has been optimised for the selective production of value-added chemicals, including oligomers, carboxylic acids, sugar alcohols and polyhydric alcohols. Finally, the energetic aspects of this process have been analysed and discussed. Therefore, considering the lack of publications exploring the use of Ru/CNF as a catalyst, along with the negligible literature covering the impact of the processing conditions on the hydrothermal hydrogenation of biomass in general and almond hulls in particular, this work symbolises a step forward in this area, providing novel information on the chemical, catalytical and energetic aspects of this process.A ruthenium supported on carbon nanofibers (Ru/CNF) catalyst, previously used to hydrolytically hydrogenate cellobiose (Frecha et al., 2019), was employed in this work. It consists of Ru nanoparticles of around 1.2 nm supported on CNF, resembling a fishbone structure (observed by Transmission Electron Microscopy), with a Ru content of 0.4 wt% (calculated by Inductive Coupled Plasma). The synthesis comprises two key stages: the synthesis of the fibres and the subsequent deposition of Ru onto the CNF by incipient wetness impregnation, utilising RuCl3 as the Ru precursor. The detailed procedure is provided in the supporting information. The support (CNF) contribution to biomass depolymerisation was addressed in previous work using cellulose as the substrate. Due to their low acidity, the contribution of the CNF did not substantially improve that of water at hydrothermal conditions (Frecha et al., 2021). Besides, the performance of the catalyst was compared to that of the CNF alone in previous publications. The experiments were performed at similar conditions (180 °C, 40 bar H2, 3 h) to those used in this work (Frecha et al., 2019). When CNF were used as a catalyst, cellobiose was depolymerised to glucose and fructose, with levulinic acid, HMF and humins being produced as main by-products. This indicated that hydrogenation reactions did not occur to a substantial step. In contrast, when the experiments were conducted in the presence of Ru/CNF, the liquid phase was made up of sorbitol, xylitol, cellobitol and glucose, thus highlighting the hydrogenation properties of the Ru/CNF catalyst in comparison to the original CNF. Besides, in another work (Remón et al., 2019b), the same CNF were tested for the hydrodeoxygenation of guaiacol (a bio-oil model compound) and their activity was compared to that of a Mo2C/CNF catalyst. Again, the CNF showed little activity for hydrogenation. Additionally, for the hydrothermal hydrogenation of almond hulls, previous work conducted revealed that hydrolysis and depolymerisation reactions are promoted by the acidity provided by water at hydrothermal conditions (Remón et al., 2021b) and the CNF, while Ru species mostly catalyse hydrogenations (Frecha et al., 2019; Remón et al., 2019b).The hydrothermal hydrogenation tests were conducted in a small batch, high-pressure reactor (Berghof Products, BR-40 series, 45 mL). Before the reaction, the almond hulls were mixed with the required amount of catalyst (Ru/CNF) in a planetary miller (PM 100 CM, Retsch, Germany), comprising a zirconia vessel (50 mL) and 10 zirconia balls of 10 mm each. This mix-milling step was conducted at room temperature and a rotation speed of 600 rpm for 30 min to diminish mass transfer limitations between the catalyst and almond hulls. The solid mixture (almond hulls and catalyst) was then loaded into the reactor, along with 20 mL of deionised water. The reactor was closed and filled in with N2 to achieve a pressure higher than that used at the reaction conditions to confirm its airtightness. Subsequently, the reactor was purged with H2 and filled in with the required amount of H2 to achieve the initial H2 pressures used in the experiments, i.e., 20, 35 and 50 bar. For these initial pressures, the final H2 pressure achieved at the reaction conditions were 26, 39 and 50 bar at 150 °C; 33, 50 and 66 bar at 190 °C; and 40, 60 and 80 bar at 230 °C. Due to the excess of H2 used, minimal pressure variations took place during the experiments. A ramp time (from room temperature to the reaction conditions) of around 35–45 min (depending on the reaction conditions) and a rotation speed of 1000 rpm were used for all the experiments. Once the reaction terminated, the reactor was quenched with cold water to achieve initial conditions as speedily as possible. A gas sample was then collected and analysed. The reactor was opened, and its content recovered as a final step. Subsequently, a solid-liquid extraction was accomplished in a funnel. The solid was dried at 105 °C for 24 h and quantified gravimetrically, while the aqueous stream was kept for further analysis.The effects of the temperature (150–230 °C), initial H2 pressure (20–50 bar), reaction time (20–360 min) and catalyst/biomass ratio (0.25–1 g/g) were analysed using as the response variables the distribution of the overall reaction products (gas, liquid and solid yields) and the chemical composition of the liquid phase. These intervals are commonly used for the hydrothermal hydrogenation of biomass or related structural model compounds. Besides, it must be borne in mind that the Ru loading in the catalyst is as low as 0.4 wt% and that the support consists of renewable-based carbon nanofibers produced from biomass. These values account for a Ru/biomass loading as low as 0.1 to 0.4 g Ru/g biomass. In previous work, we addressed the hydrothermal treatment of almond hulls for biofuels production in the absence of hydrogen and catalyst. We reported that low biomass/water ratios favour hydrolysis and depolymerisation reactions. On the contrary, high biomass/water ratios promote pyrolysis and thermal decomposition reactions, favouring gas production. This latter accounts for the lower amount of water present in the reaction medium. The present work is directed towards producing value-added liquid products; therefore, a low biomass/water ratio (5 wt%) was used.The calculations and analytical methodologies employed for their determination are listed in Table 1 . The composition of the gas phase was determined using a micro gas chromatograph. The chemical composition of the liquid phase was determined by Gas and High-Performance Liquid Chromatography (GC and HPLC). Elemental analyses were conducted using a Carlo Erba EA1108 Elemental Analyser using the Channiwala and Parikh (2002) empirical formulae to determine the HHV of the spent solids. Thermogravimetric analyses were conducted on a NETZSCH TG 209F1 Libra TGA209F1D-0277-L apparatus, using air as the carrier gas (50 mL STP min−1), increasing the temperature from 25 to 1000 °C (10 °C min−1). Detailed information regarding these calculations is included in the supporting information. Besides, details covering the experimental is provided as supplementary material and reported elsewhere (Frecha et al., 2019; Frecha et al., 2021; Remón et al., 2021b; Remón et al., 2019a; Remón et al., 2018c).The experiments to address the influence of the processing parameters were planned following a 24 Box-Wilson Central Composite Face Centred (CCF, α: ±1) design. The data were then analysed through a 95% confidence (p-value = 0.05) ANOVA (to determine significance) coupled with a cause-effect Pareto test (to calculate relative importance). For both tests, codec variables (between −1 and +1) were utilised, thus making the factors directly comparable. To analyse the results, it is essential to note that the codec formulae obtained from the ANOVA of the 28 runs were used to develop interaction plots showing the main effects and interactions detected. These figures were used to address the impact of the processing conditions and interactions on the process. In addition, in these figures, when possible, some experimental points were added to graphically show that the lack of fit is not significant. This detailed methodology warrants a clear and accurate analysis of the experimental data (Remón et al., 2018a). More information on this methodology is given as supporting information.The almond hulls used in this work were from Marcona almonds harvested in Spain. The complete preparation procedure and characterisation results of the material are fully reported in our previous publication (Remón et al., 2021b). Very briefly, almond hulls were dried at 60 °C overnight to prevent moulds formation during storage. Then, they were knife milled and sieved to a particle size of ca. 100–200 μm. The dried hulls were characterised by proximate and ultimate, elemental and fibre analyses. The proximate analysis showed that almond hulls were made up of 6.72 ± 2.87 wt% moisture, 62.72 ± 1.93 wt% volatiles, 18.77 ± 0.57 wt% fixed carbon and 11.80 ± 0.37 wt% ash (primarily K, Mg and Ca). Structurally, they comprised 12.60 ± 0.77 wt% cellulose, 19.40 ± 1.17 wt% hemicellulose, 25.10 ± 2.47 wt% lignin, 7.81 ± 0.23 wt% proteins, 11.80 ± 0.87 wt% ash and 16.57 ± 1.27 wt% others (mainly waxes and lipids). Overall, the material consists of 44.23 ± 1.38 wt% C, 4.65 ± 0.21 wt% H, 49.88 ± 1.52 wt% O and 1.25 ± 0.07 wt% N and has a HHV of 15.74 ± 0.50 MJ/kg. These values are in line with those reported in the literature (Aktas et al., 2015; González et al., 2005). Table 2 outlines the experimental hydrothermal hydrogenation conditions used and the experimental results attained. These include the overall distribution of reaction products (yields to gas, aqueous and solid), along with the detailed chemical composition of the aqueous stream. Table S1 lists the detailed chemical composition of the liquid phase. The influences of the reaction parameters on these results according to the ANOVA and cause-effect Pareto principle analyses (contemplating all runs conducted) are included in Table S2.The hydrothermal hydrogenation of almond hulls over our Ru/CNF catalyst leads to the formation of three main reaction products: a gas stream, an aqueous fraction and a spent solid product. The yields to these fractions depend on the reaction conditions and vary by 0–5%,49–82% and 13–51%, respectively. The gas stream primarily consists of CO2, along with unreacted H2. Depending on the processing conditions, the spent solid product essentially comprises unreacted biomass, together with reacted biomass from depolymerisation and repolymerisation reactions. The aqueous phase includes value-added liquid chemicals of different nature. CO2 as the only product in the gas suggests that its formation primarily occurs by decarboxylation, decarbonylation, reforming and thermal cracking reactions (Cheng et al., 2017; Wang et al., 2012; Xu et al., 2013). In the scope of this publication, the influence of the processing conditions on the properties of these two fractions (gas and solid) has not been discussed in full. The cause-effect Pareto test (Table S2) reveals the gas yield is affected mainly by the temperature and pressure, both individually and combined, while the liquid yield depends on the temperature and two binary interactions: temperature-catalyst and pressure-time loading. At the same time, this latter interaction and the temperature are the variables exerting the most significant influence on the solid yield. The detailed impact of the effects of the processing conditions and most meaningful interactions on these yields are summarised in Fig. 1 .The impact of the temperature on the overall distribution of the reaction products is directed by the catalyst loading. When a low catalyst/biomass ratio (0.25 g/g) is used, its influence relies on the initial H2 pressure (Fig. 1 a/b, e/f and i/j). For a low initial H2 pressure (20 bar), without regard to the reaction time, an increase in the temperature leads to increases in the gas (especially between 160 and 200 °C) and liquid yields at the expense of the solid yield; with these changes being more marked for a longer than a short reaction time. These developments suggest the beneficial impact of the temperature on the process, which allows almond hulls conversion into gas and liquid species via hydrolysis, depolymerisation, deamination, reforming and thermal cracking reactions (Dimitriadis and Bezergianni, 2017; Gollakota et al., 2018; Kumar et al., 2018; Thiruvenkadam et al., 2015).Spreading the H2 pressure from 20 to 50 bar not only significantly modifies the overall distribution of these products but also the influences of the temperature and reaction time, in some cases. In particular, such an increase in the H2 pressure drops the gas yield. This pressure upturn facilitates the diffusion of water into almond hulls, leading to a more efficient solid-water interaction (Remón et al., 2021b; Remón et al., 2021c). The amount of protons in the reaction medium also increases, owing to a more significant water dissociation at high pressure, promoting acid-catalyzed reactions (Schienbein and Marx, 2020). At the same time, such a pressure spread also augments the H2 availability in the reaction media, which promotes hydrogenation and hydrogenolysis reactions (Cheng et al., 2017), favouring liquid and solid formation (Kumar et al., 2018; Thiruvenkadam et al., 2015).Furthermore, the impacts of this pressure spread on the liquid and solid yields rely on the reaction time. On the one side, for a speedy process (20 min), the liquid yield increases and the solid yield drops. This suggests that hydrolysis, depolymerisation, hydrogenation and hydrogenolysis reactions, boosting water-soluble liquid species, take place to a more substantial extent. On the other, for a lengthy treatment (360 min), the effect of the pressure depends on the temperature. While at low temperature (160–190 °C), increasing the pressure does not alter the liquid and solid yields, both rise when higher temperatures are used (between 190 and 230 °C). The competition between different reactions might account for these differences. On the one side, an increase in the reaction temperature leads to a lesser spread of hydrogenation and hydrolysis reactions. This latter accounts for both the exothermic character of these transformations (Xu et al., 2013) and the lesser amount of H2 available in the liquid phase owing to the lower H2 solubility in water at a high than low temperature (Wang et al., 2012). On the other, increasing the reaction time promotes secondary decomposition reactions, thus favouring the transformation of some liquid products, such as furan compounds, into solids (Xu et al., 2013).Therefore, these results indicate that there is compensation between the positive effect of augmenting the initial H2 pressure and the negative impact of using long reaction times at a low temperature. Conversely, such compensation does not take place at high temperatures for long reaction times. As a result of these variances, different outcomes are perceived for the liquid and solid yields with the temperature when a high initial H2 pressure (50 bar) is used. Regardless of the reaction time, increasing the temperature increases the liquid yield and drops the solid yield between 180 and 195 °C by reason of the positive kinetic influence of the temperature on the process (Remón et al., 2021b; Remón et al., 2021c). With further increased up to 230 °C, the liquid yield decreases and the solid yield increases on account of the transformation of liquids into solid species.Increasing the catalyst loading also modifies the reaction products distribution. These changes depend on the initial H2 pressure and are more critical for a longer than a short process duration. On the one hand, for a speedy process (20 min), at 20 bar (Fig. 1 a/e/i vs c/g/k), increasing the catalyst/biomass ratio from 0.25 to 1 g/g significantly decreases the gas yield and increases the solid yield without substantially modifying the liquid yield. An increase in the catalyst loading favours biomass depolymerisation (via hydrolysis, hydrogenation and hydrogenolysis reactions, yielding water-soluble liquids) over the direct thermal decomposition of the solid material to gases (Lam and Luong, 2014). At the same time, an upsurge in the catalyst amount also endorses the subsequent decomposition of these liquids formed. These can evolve towards solid species, such as humins, via dehydration and repolymerisation. Conversely, at 50 bar, the liquid yield increases at the expense of the solid yield, especially at temperatures higher than 200 °C, due to the pressure promoting liquid production (Remón et al., 2021b; Remón et al., 2021c). On the other hand, when a long reaction time is used (360 min), increasing the catalyst amount rises the gas and liquid yields at a high temperature at the expense of solid formation, without regard to the initial H2 pressure. These transformations take place to a more significant extent, given the more prolonged exposure of the material to hydrothermal conditions (Remón et al., 2021b; Remón et al., 2021c). As a result, for a high catalyst/biomass ratio (1 g/g), the temperature depicts two distinctive influences depending on the initial H2 pressure (Fig. 1 c/d, g/h and k/l). At 20 initial bar H2, an increase from 160 to 230 °C progressively upturns the gas and liquid yields and diminishes the solid yield regardless of the reaction time. These variations are particularly more marked for a longer than a short reaction time as described earlier (Prado et al., 2016).An increment in the initial H2 pressure leads to a decrease in the gas yield, as commented for a low catalyst/biomass ratio. For a short reaction time (20 min), such an increase in pressure significantly upsurges the liquid yield and drops the solid yield, regardless of the temperature. These variations result from the beneficial kinetic influence of the pressure when a short reaction time is applied, thus promoting biomass decomposition into liquid species but preventing secondary reactions from occurring substantially. Contrariwise, when a long reaction time is used (360 min), the effect of the pressure is only significant at a high temperature (190–230 °C). An increase in pressure leads to a slight decrease and increase in the liquid and solid yields, respectively, due to the transformation of liquid species into solid products by dehydration, condensation and repolymerisation (Xu et al., 2013). This pressure influence also modifies the effect of the temperature. Thus, at 50 bar, between 160 and 190 °C, the gas yield is meagre, while the liquid yield increases at the expense of the solid yield. These variations account for depolymerisation, hydrogenation and hydrogenolysis reactions favoured at high pressure and low temperature in the presence of a catalyst (Wang et al., 2012; Xu et al., 2013). Conversely, a further increase up to 230 °C increases the gas yield very sharply at the expense of the liquid yield, with the solid yield being practically unaffected due to the transformation of some liquid species into gases using high temperatures for long processing times (Dimitriadis and Bezergianni, 2017; Gollakota et al., 2018).Owing to these phenomena, the impact of the reaction time is directed by the catalyst/biomass ratio and initial H2 pressure. In particular, the reaction time significantly influences the liquid and solid yields when a low catalyst loading is used (0.25 g/g), while its impact on the gas yield is not significant from a practical point of view (Fig. 1 a/e/i vs b/f/j). At a low initial H2 pressure (20 bar), lengthening the process leads to an increase in the liquid yield at the expense of the solid yield. Conversely, at 50 bar, such an increase in the process duration diminishes the liquid yield and increases the solid yield. Rising the pressure promotes a greater spread of the reactions occurring in the liquid phase, favouring liquids production and their subsequent transformation into less polymerised products. These can then evolve towards solid species by dehydration, condensation, and repolymerisation (Xu et al., 2013), as described above. For a high catalyst loading (1 g/g), the influence of the reaction time is significant at low pressure (20 bar). Under such conditions, lengthening the process substantially increases the gas and liquid yields and diminishes the solid yield. Besides, the higher the pressure, the lesser is the impact of the reaction time, as the positive kinetic influence of the former can mask the effect of the latter. This development accounts for the beneficial impact of the catalyst on the liquid species transformation into gaseous products via cracking, deoxygenation, deamination (Dimitriadis and Bezergianni, 2017; Gollakota et al., 2018), and solid species through dehydration, condensation, and repolymerisation (Xu et al., 2013).The liquid phase comprises a complex pool of different species whose composition depends on the reaction parameters. It includes oligomers (46–81 wt%), saccharides (2–7 wt%), sugar alcohols (2–15 wt%), polyhydric alcohols (1–8 wt%), carboxylic acids (7–31 wt%), furans (0–3 wt%), nitrogen-containing species (0–1 wt%) and phenolic compounds (0–2 wt%). Oligomers are produced mostly from the depolymerisation of cellulose and hemicellulose, leading to the presence of cello-oligosaccharides and xylo-oligosaccharides in the liquid product. Ligno-oligomers resulting from the depolymerisation of lignin are also present in this fraction. Saccharides result from the hydrolysis and depolymerisation of cellulose and hemicellulose. The hydrogenation of these latter fractions leads to the formation of sugar alcohols and polyhydric alcohols, while carboxylic acids and furans are mostly produced via dehydrogenation reactions. Phenolic compounds are produced from the depolymerisation of ligno-oligomers, while nitrogen-containing species result from the degradation of the proteins present in the original biomass. Table S1 lists the detailed chemical composition of the liquid phase, while Fig. 2 shows a reaction pathway covering the detailed formation of liquid products from the structural components (cellulose, hemicellulose, lignin and proteins) in almond hulls.Cellulose decomposition occurs via a first hydrolysis step, yielding cello-oligosaccharides (Jiang et al., 2020; Jiang et al., 2019; Remón et al., 2020; Remón et al., 2018b; Remón et al., 2018c). These can progressively depolymerise to glucose (Davila et al., 2019; Jiang et al., 2020; Jiang et al., 2019), which can be subsequently transformed into fructose via isomerisation (Li et al., 2018). When a hydrogen-rich atmosphere is achieved at hydrothermal conditions, these species can undergo different transformations (Li et al., 2018). These include hydrogenations yielding sugar alcohols, such as sorbitol and mannitol (Manaenkov et al., 2019; Sun and Liu, 2011), dehydrogenations to produce gluconic acid, and/or dehydration, generating 5-hydroxymethylfurfural (5-HMF) (Remón et al., 2018c). This latter can be subsequently decomposed into levulinic and formic acids by hydrolysis. Simultaneously, glucose and fructose can be further decomposed into small oxygenates via the retro-aldol reaction. On the one hand, the former can evolve towards erythrose and 2-hydroxy acetaldehyde (Manaenkov et al., 2019), while the latter can decompose into 2,3-dihydroxypropanal and 1,3-dihydroxypropan-2-one. Erythrose may lead to the formation of erythritol and/or 1,2-butanol via hydrolytic dehydration (Li et al., 2018). Additionally, 2-butanol can be subsequently produced from both species and evolve to butane-2-one. In addition, ethane-1,2-diol can be formed from 2-hydroxy-acetaldehyde (Manaenkov et al., 2019), whose subsequent decomposition might lead to the formation of ethanol and/or acetaldehyde, both of which can subsequently evolve towards acetic acid (Remón et al., 2018d).On the other hand, 2,3-dihydroxypropanal and 1,3-dihydroxypropan-2-one can be produced from fructose via the retro aldol reaction. Both species can be decarboxylated, yielding ethane-1,2-diol, which can evolve towards acetaldehyde via dehydration, and/or 2-hydroxy-acetaldehyde by dehydrogenation. These chemicals can be decomposed into ethane-1,2-diol through decarbonylation (King et al., 2010; Lin, 2013). Acetaldehyde and 2-hydroxyacetaldehyde can be produced by dehydration and dehydrogenation, respectively (King et al., 2010; Lin, 2013; Wawrzetz et al., 2010). The former can lead to ethanol and acetic acid, while the latter can be decomposed into methanol (King et al., 2010; Lin, 2013; Wawrzetz et al., 2010). In addition, glycerol can also be produced by the hydrogenation of 2,3-dihydroxypropanal and 1,3-dihydroxypropan-2-one, both of which can be dehydrogenated yielding 1-hydroxypropan-2-one (Remón et al., 2016c; Remón et al., 2016d; Xu et al., 2022). This can undergo further hydrogenation towards propane-1,2-diol (Gandarias et al., 2010; King et al., 2010; Lin, 2013; Wawrzetz et al., 2010; Zhang et al., 2012). Additionally, lactic acid can be produced from 1,3-dihydroxypropan-2-one via rearrangement (dehydration/hydration) (Onda et al., 2008; Xu et al., 2021). At the same time, propane-1,2-diol can subsequently be dehydrated to form propane-2-one and/or propionaldehyde, which can be hydrogenated to propane-2-ol and propane-1-ol, respectively (Gandarias et al., 2010). Afterwards, ethanol might be produced from propane-2-ol by cracking and hydrogenation (Lin, 2013).Hemicellulose depolymerises yielding hemi-(xylo)-oligosaccharides, which progressively evolve towards xylose formation (Jiang et al., 2018; Remón et al., 2019a; Remón et al., 2018c). This saccharide can be subsequently hydrogenated to produce arabitol, xylitol and/or threitol, and/or dehydrated yielding furfural, which can be decomposed into formic acid (Putro et al., 2016). At the same time, xylose can lead to 2-hydroxyacetaldehyde and 2,3-dihydroxypropanal/1,3-dihydroxypropan-2-one via the retro aldol reaction (Putro et al., 2016; Sun and Liu, 2011). Ethane-1,2-diol can be produced from both species via hydrogenation and decarboxylation, respectively, while glycerol can also be attained from the hydrogenation of 2,3-dihydroxypropanal/1,3-dihydroxypropan-2-one (Putro et al., 2016; Sun and Liu, 2011). Then, both alcohols can evolve towards forming the same species as described above for cellulose.The reactivity of lignin and proteins at the processing conditions of this work is lower than that of the carbohydrate fraction, attending to the number and amounts of products formed. At the conditions tested, lignin depolymerisation leads to the formation of ligno-oligomers that progressively depolymerise and decompose towards different phenolic compounds: phenol-2-methoxy-4-propyl, phenol,2-6-dimethoxy, phenol,2-methoxy and phenol (Li et al., 2012; Li et al., 2018; Madsen et al., 2017; Madsen and Glasius, 2019; Yang et al., 2018). Proteins can be hydrolysed, yielding amino acids, which can be further decarboxylated, decarbonylated and deaminated (Kumar et al., 2018; Yang et al., 2015), leading to the formation of amides (propenamide), amines (2-pentanamine) and ammonia, respectively. Additionally, pyridines, such as 3-pyridinol, can be produced from the reaction between ammonia and the furans produced from the carbohydrate content (Madsen et al., 2017).The influence of the processing conditions according to the Pareto test is listed in Table S2. For the most abundant species, this analysis reveals that the proportion of oligomers in the liquid is mainly affected by the temperature and reaction time, along with the initial H2 pressure and catalyst loading. The proportion of saccharides relies on the temperature and catalyst loading, while the relative amount of sugar alcohols mainly depends on the reaction temperature and its interaction with the reaction time. The reaction time and its interaction with the pressure and catalyst loading are responsible for the changes observed in the proportions of polyhydric alcohols. Carboxylic acids are primarily affected by the reaction temperature and some interactions with the time and pressure, while the proportion of ketones is substantially influenced by the catalyst loading and reaction time. The detailed influence (from the ANOVA of all runs) of these effects and interactions on the proportions of the most abundant liquid species are plotted in Fig. 3 .Different outcomes are observed when a low catalyst/biomass ratio (0.25 g/g) is used (Fig. 3 a/e/i/m/q/u and b/f/j/n/r/v for 20 and 360 min, respectively). The reaction temperature does not significantly alter the chemical composition for a quick treatment (20 min) and low initial H2 pressure (20 bar). An increase from 180 to 230 °C leads to moderate decreases in the relative amounts of oligomers and saccharides. It increases the concentration of sugar alcohols, with the proportions of polyhydric alcohols, carboxylic acids and ketones being unaffected practically. These transformations suggest an initial conversion of oligomers (mostly cello-oligosaccharides and hemi/xylo-oligosaccharides) into saccharides via hydrolysis (Jiang et al., 2020; Jiang et al., 2019; Remón et al., 2020; Remón et al., 2018b; Remón et al., 2018c), and the subsequent transformation of these species into sugar alcohols by hydrogenation (Manaenkov et al., 2019; Putro et al., 2016; Sun and Liu, 2011).An increase in the initial H2 pressure from 20 to 50 bar modifies the composition of the liquid effluent. At a low temperature (160–200 °C), such an increase augments the proportions of oligomers at the expenses of the concentrations of saccharides, sugar alcohols and carboxylic acids. This accounts for the rise in the liquid yield occurring when the pressure of the system increases, which favours water penetration and promotes the decomposition of almond hulls into water-soluble oligomers (Remón et al., 2021b; Remón et al., 2021c; Schienbein and Marx, 2020). However, under such conditions, these species do not evolve towards less-depolymerised species, and consequently, their relative amounts in the liquid phase increase. This seems to indicate that the subsequent conversion of these water-soluble macromolecules is the rate-determining step. On the contrary, at a high temperature (210–230 °C), the concentration of carboxylic acids increases and the proportions of oligomers and saccharides diminish when the H2 pressure increases. This pressure spreads not only promotes the formation of oligomers but also their subsequent transformation into other products (Remón et al., 2021b).As a result of these different outcomes, when an elevated initial H2 pressure (50 bar) is applied, the proportion of oligomers decreases between 160 and 230 °C, while the relative amounts of saccharides, sugar alcohols and carboxylic acids increase. The progressive transformation of saccharides accounts for this, as their conversion into small oxygenated species, such as carboxylic acids, is promoted by the positive kinetic influence of the pressure and temperature.(Jiang et al., 2020; Jiang et al., 2019; Remón et al., 2020). However, these alterations are particularly important between 160 and 190 °C for sugar alcohols and from 190 to 230 °C in the case of carboxylic acids. The solubility of H2 in water diminishes with augmenting the reaction temperature (Wawrzetz et al., 2010) and hydrogenation reactions are also less favoured at a high temperature due to their exothermic nature (Cheng et al., 2017; Wang et al., 2012; Xu et al., 2013). Thus, the beneficial kinetic impact of the pressure boots the transformation of oligomers differently depending on the temperature, with hydrogenations being favoured at a low temperature and dehydration reactions occurring to a more substantial extent at a high temperature.Additionally, the effect of the reaction time is ruled by the initial H2 pressure and reaction temperature. At 20 bar, lengthening the duration from 20 to 360 min rises the proportions of sugar alcohols and polyhydric alcohols at a low temperature (160–190 °C). These variations are accompanied by diminishments in the relative amounts of oligomers, carboxylic acids and ketones. Increasing the reaction time promotes the gradual conversion of oligomers into less polymerised species, mostly saccharides, along with their hydrogenation to sugar alcohols and polyhydric alcohols. These transformations are favoured at a low temperature due to the endothermicity of hydrogenation reactions (Cheng et al., 2017; Wang et al., 2012; Xu et al., 2013) and the greater amount of H2 dissolved in the liquid phase (Wawrzetz et al., 2010). Conversely, at a high temperature (190–230 °C), such an increase in the reaction time increases the proportions of oligomers at the expenses of the relative amounts of sugar alcohols and carboxylic acids. Prolonging the reaction time at a high temperature promotes the initial biomass conversion into water-soluble oligomers and the transformation of some low molecular mass oxygenates, such as carboxylic acids, into gases (Lorente et al., 2019; Remón et al., 2021b; Remón et al., 2019c; Remón et al., 2021c). These developments account for the increments observed in the proportions of oligomers and the diminishments occurring for sugar alcohols and carboxylic acids. At 50 bar, the effect of the reaction time is more marked.Regardless of the temperature, an increase in the reaction time leads to a sharp decrease in the proportion of oligomers, accompanied by increases in the concentration of saccharides, sugar alcohols and polyhydric alcohols. The concentration of carboxylic acids increases at a low temperature and decreases at a high temperature, while the relative amount of ketones is unaffected. These outcomes are thought to result from the positive influence of the reaction time on the process, which favours the transformation of oligomers into saccharides via hydrolysis (Lorente et al., 2019; Remón et al., 2021b; Remón et al., 2019c; Remón et al., 2021c) and their subsequent transformation into sugar alcohols and polyhydric alcohols (Zhou et al., 2012). In this case, the thermodynamic H2 limitation at a high temperature is less detrimental, as it can be compensated by a more prolonged exposure at hydrothermal conditions along with the greater H2 pressure applied.Owing to these differences, diverse outcomes are observed with varying the temperature and initial H2 pressure for a reaction time of 360 min. At 20 bar, increasing the temperature increases the proportion of oligomers and decreases the relative amount of saccharides. The former is associated with the increase observed in the liquid yield with increasing the temperature, while the latter suggests converting saccharides into other species; with these transformations being boosted when a high temperature is applied for a long reaction time. For example, the proportion of polyhydric alcohols initially decreases between 150 and 190 °C and then stabilises with a further increase in the temperature. Contrarily, a trend-off is observed for the relative amount of sugar alcohols between 160 and 190 °C, followed by a sharp decrease with a further temperature increment to 230 °C. These transformations suggest that hydrogenations resulting in sugar alcohols occur more significantly than the transformation of saccharides into sugars alcohols via a first retro-aldol reaction and subsequent hydrogenation. Conversely, at a high temperature, hydrogenations are less predominant due to the lower H2 solubility in water (Wawrzetz et al., 2010), which favours the conversion of saccharides, first by the retro-aldol reaction, and then by a series of dehydration, hydration and decarboxylation reactions, increasing the concentrations of carboxylic acids and ketones.For a long reaction time (360 min), the effect of the initial H2 pressure is different. Notably, a pressure spread from 20 to 50 bar H2 increases the proportion of saccharides and carboxylic acids at the expense of the relative amount of oligomers, regardless of the temperature, due to the positive influence of the pressure. Besides, the concentrations of sugar alcohols and polyhydric alcohols diminish at a low temperature and increase at a high temperature, while the proportion of ketones is unaffected. At a low temperature, these diminishments are related to the substantial increase occurring in the proportion of carboxylic acids, primarily acetic acid, which denotes that the beneficial pressure kinetic impact helps shift the conversion of saccharides into carboxylic acids via a tandem of consecutive transformations, including the retro-aldol reaction coupled with hydrogenations and dehydrations as described in the reaction pathway (Jiang et al., 2020; Jiang et al., 2019; Remón et al., 2020). As a result, the effect of the temperature is less critical, probably due to the higher pressure and longer reaction time used. Notably, the proportions of oligomers, saccharides and ketones are barely affected by the temperature. On the contrary, the proportion of polyhydric alcohols increases linearly. Simultaneously, a maximum and a minimum occur for the relative amounts of sugar alcohols and carboxylic acids, respectively. An initial augment between 150 and 190 °C increases the proportion of the former at the expense of the relative amount of the latter. In contrast, a subsequent increase leads to an opposite outcome. These trends suggest that the hydrogenation of saccharides occurs to a substantial extent at a low temperature, accounted for by the positive effect of the pressure and reaction time, combined with an appropriate H2 solubility in water (Wawrzetz et al., 2010) and an appropriate reaction temperature promoting hydrogenation reactions (Cheng et al., 2017; Wang et al., 2012; Xu et al., 2013). Conversely, an increase in the temperature exerts a positive kinetic influence on the retro-aldol conversion of saccharides. Thus, the subsequent evolution of these species towards carboxylic acids is preferential over the direct hydrogenation.Increasing the catalyst loading modifies not only the liquid phase chemical composition but also the effects that the reaction temperature and initial H2 pressure have on the process (Fig. 3 a/e/i/m/q/u vs c/g/k/o/s/w and b/f/j/n/r/v vs d/h/l/p/t/x for 20 and 360 min, respectively). For a catalyst/biomass ratio of 1 g/g, the influence of the reaction temperature does not depend on the initial H2 pressure or reaction time for the most abundant species in the liquid effluent. An increment in the temperature between 150 and 200 °C leads to a decrease in the proportion of oligomers at the expense of the relative amounts of sugar alcohols and polyhydric alcohols. This results from the beneficial impact of the temperature on hydrolysis and depolymerisation combined with a temperature range wherein hydrogenations are thermodynamically promoted. On the contrary, a subsequent increment up to 230 °C increases the concentration of oligomers and drops the proportions of sugar alcohols and polyhydric alcohols. These differences might be a consequence of the kinetic and thermodynamic influence of the reaction temperature. An initial increase in the temperature favours the conversion of biomass into water-soluble oligomers, promoting their subsequent transformation into saccharides, sugar alcohols and polyhydric alcohols by hydrogenation in the presence of a high amount of catalyst. These latter transformations are favoured at a low temperature due to the more significant H2 dissolved in the aqueous medium and the endothermic character of hydrogenation reactions (Cheng et al., 2017; Wang et al., 2012; Xu et al., 2013). However, a prolonged increment in the temperature hinders these transformations. Consequently, the relative amount of oligomers increases due to the positive influence of the temperature on biomass hydrolysis and the limited extension of hydrogenation reaction at a high temperature. These phenomena are more marked when a high catalyst loading is used, as hydrogenations occur to a greater extent.On the contrary, the proportions of saccharides, carboxylic acids and ketones are influenced by the initial H2 pressure and reaction time. For a 20 min reaction time, the concentration of saccharides is not directed by the reaction temperature; a trend-off is observed without regard to the initial H2 pressure. Conversely, the relative amount of carboxylic acids increases with rising the temperature from 150 to 230 °C, irrespective of the initial H2 pressure; yet, a sharper increase is observed for a low than high H2 pressure. The beneficial impact of the temperature on saccharides decomposition via the retro-aldol reaction and the subsequent promoting effect of the temperature on dehydration, decarboxylation and hydration reactions, leading to the formation of carboxylic acids, might be responsible for such developments (Remón et al., 2016b). Concurrently, the relative amount of ketones diminishes between 150 and 230 °C when the process is conducted at 20 bar due to the sharp upturn occurring for carboxylic acids, with the effect of the temperature being less and less important as the initial H2 pressure increases. In this case, greater H2 pressure can compensate for the lower solubility of H2 at a high temperature.In addition, the initial H2 pressure also directs the chemical composition. For a speedy treatment (20 min), different outcomes occur. On the one side, an increase from 20 to 50 bar H2 increases the relative amount of polyhydric alcohols at the expenses of the proportions of saccharides and sugar alcohols. These developments suggest that the retro-aldol reaction might be quicker than the direct hydrogenation of saccharides to sugar alcohols, favouring the formation of polyhydric alcohols over sugar alcohols under such conditions. In this case, spreading the H2 pressure kinetically boots these transformations, owing to the rise in the total pressure of the system. On the other, the pressure for the proportions of oligomers, carboxylic acids and ketones relies on the temperature. While the concentration of carboxylic acids increases at a low temperature (150–190 °C) due to decreases occurring in the relative amounts of oligomers and ketones, the opposite trend is observed at a high temperature (190–230 °C); i.e., the relative amounts of oligomers and ketones increase at the expense of the carboxylic acids concentration, when the initial H2 pressure rises. An increase in the pressure positively influences hydrolysis and hydrogenation reactions at a low temperature, thus facilitating the first conversion of biomass into water-soluble oligomers and their transformation into less depolymerised species via hydrolysis and hydrogenation. Conversely, hydrolysis and depolymerisation occur significantly at a high temperature, with the subsequent hydrogenation of sugars being the limiting step. This can hinder the conversion of saccharides and increases the proportion of oligomers in the liquid.Furthermore, an increase in the reaction time modifies the effect of the pressure. Augmenting the H2 pressure from 20 to 50 bar, using a reaction time of 360 min, leads to a substantial spread in the relative amount of saccharides, especially between 150 and 190 °C, temperatures at which the relative amount of polyhydric alcohols also increases. Such increases occur along with a depletion in the concentration of oligomers. Under such conditions, the positive influence of the processing time and the H2 pressure promote hydrolysis and hydrogenation reactions combined with the higher amount of H2 dissolved at a low temperature (Wawrzetz et al., 2010) and the exothermicity character of hydrogenations (Cheng et al., 2017; Wang et al., 2012; Xu et al., 2013). Besides, at higher temperatures (190–230 °C), the same spread in the H2 pressure upsurges the relative amount of oligomers. It diminishes the proportion of polyhydric alcohols due to the positive and negative impacts of the initial H2 pressure at a high temperature on hydrolysis and hydrogenation reactions.Owing to these distinctive effects exerted by the temperature and initial H2 pressure, the impact of the reaction time relies on these two variables. At 20 bar, it is particularly important at a high temperature (190–230 °C). Within this range, lengthening the process (from 20 to 360 min) increases the relative amounts of oligomers and polyhydric alcohols, leading to decreased concentrations of saccharides, sugar alcohols and ketones. This might be accounted for by hydrolysis and depolymerisation reactions, yielding saccharides, being faster than the following transformation of these species into other chemicals. At 50 bar H2, when a temperature between 150 and 190 °C is used, lengthening the process increases the proportions of saccharides, polyhydric alcohols and carboxylic acids and diminishes the relative amounts of oligomers, sugar alcohols and ketones. In this case, an increase in the reaction time promotes oligomers depolymerisation, and also it shifts the decomposition of saccharides via the retro-aldol reaction. Conversely, when the process is conducted at a high temperature (190–230 °C), such an increase in the reaction time increases the relative amounts of oligomers and sugar alcohols at the expenses of the concentrations of carboxylic acids and ketones. At a high temperature, these latter species can be easily converted into gaseous products (Lorente et al., 2019; Remón et al., 2020; Remón et al., 2021b; Remón et al., 2019a), which connects very well with the proliferation seen in the gas production at a high temperature and using a long reaction time.Five likely optimisation scenarios were sought to transform almond hulls into valuable liquids, using the formulae obtained from the ANOVA of the experimental results (Table 3 ). The lack of fit of the models developed in this work is not significant with 99% confidence (p-value > 0.01). Besides, the predicted R2 of all these models were higher than 0.95, allowing their use for prediction purposes. This proves validation for the optima obtained in this work. In this regard, it must be borne in mind that these optima serve as the starting point for future process commercialisation and scale-up of the process as they may depend on the biomass and type of reactor. The gas and solid yields were minimised in these optima, while the liquid yield was maximised to ensure the selective conversion of the material into liquid species. The first optimisation considers the transformation of almond hulls into water-soluble oligomers, while the second minimises their production to obtain a liquid product containing less polymerised species (oxygenates), i.e., primarily saccharides, sugar alcohols, polyhydric alcohols and carboxylic acids. The third and the fourth aim at transforming almond hulls into a liquid product containing high amounts of sugar alcohols and carboxylic acids, respectively. Additionally, the fifth includes the concurrent production of alcohols (sugar alcohols and polyhydric alcohols) and carboxylic acids. All restrictions have been assigned with different importance (from least important, 1, to most important, 5) for these case scenarios so that operating conditions satisfying all the criteria could be sought. The overall yield has been provided with a relative importance of 3, while 5 has been given to the chemical composition to ensure that quality (purity) prevails over quantity.Opt. 1 shows that 91% of almond hulls can be converted into an oligomer-rich (74 wt%) liquid mixture conducting the process at 230 °C and 35 bar H2 for 360 min using 1 g cat/g biomass. On the contrary, Opt. 2 reveals that the production of small liquid oxygenated compounds (58 wt%) prevails over the formation of oligomers (42 wt%), conducting the process at 206 °C and 39 bar H2 for 20 min using 1 g cat/g biomass. Nano- and ultra-filtration can resolve the separation of oligomers from saccharides and other liquid oxygenates (Feng et al., 2009; Vegas et al., 2006). This would allow the selective transformation of almond hulls into oligomers (primarily cello- and xylo-oligomers) in high yield and purity. Opt. 1 would convert up to 67% of the original material into oligomers assuming an ideal separation. Cello- and xylo-oligomers are widely used as natural, renewable-based prebiotic materials owing to their excellent physicochemical and physiological properties (Bian et al., 2014; Carvalho et al., 2013; Lin et al., 2017; Miguez et al., 2018).Concerning the production of valuable liquid oxygenates, sugar alcohols and carboxylic acids are the most abundant species in the liquid. The former (Opt. 3) production can be maximised (61% liquid yield, containing 19 wt% sugar alcohols), conducting the process at 150 °C and 32 bar H2 for 20 min using 0.25 g cat/g biomass. Likewise, oligomers can be easily removed from the aqueous effluent, which would result in the transformation of up to 30% of the original material into oligosaccharides and the other 30% into an aqueous product containing up to 38 wt% of sugar alcohols. Considering the total amount of carbohydrates (cellulose and hemicellulose) in almond hulls, this equals a sugar alcohols yield of 36% with respect to the entire carbohydrate content. This represents an improvement in yield and time reduction compared to the data reported by Palkovits et al. (2010) and Li et al. (2018) for other lignocellulosic (wood chips, cornstalk and beechwood) materials. On the contrary, Opt. 4 demonstrates that the latter are maximised (72% liquid yield, containing 26 wt% carboxylic acids) when the hydrothermal reaction is conducted at 198 °C and 42 bar H2 for 20 min using 1 g cat/g biomass. After oligomers separation, this represents the conversion of up to 37% of the almond hulls into an aqueous mixture comprising up to 50 wt% of carboxylic acids. This equals 58% of the carbohydrate content of the biomass converted into carboxylic acids, and it is one of the best results reported in the literature so far.Additionally, Opt. 5 suggests that alcohols and carboxylic acids can be maximised concurrently (68% liquid yield, containing 15 wt% sugar alcohols, 8 wt% polyhydric alcohols and 25 wt% carboxylic acids) at 187 °C and 35 bar H2 for 360 min using 1 g cat/g biomass. The difference between Opt. 1 and Opt. 5 agrees with the more significant H2 dissolved in the aqueous medium and the endothermic character of hydrogenation reactions, as discussed previously. As such, 37% of the raw biomass could be converted into an upgraded aqueous fraction containing 28 wt% of sugar alcohols and 46 wt% of carboxylic acids. These fractions can be easily fractionated by distillation considering the standard boiling point of the sugar alcohols (216–230 °C), polyhydric alcohols (188–290 °C) and carboxylic acids (101–122 °C) produced. Fig. 4 shows a schematic flow diagram to selectively transform almond hulls into oligomers, alcohols and carboxylic acids. These species are renewable-based substitutes for petroleum-based chemicals of paramount interest for the energy sector and pharmaceutical, cosmetic and textile industries. As such, these promising results, combined with the environmental friendliness and holistic features of this hydrothermal process, exemplify a landmark step change to managing and valorising unavoidable food waste.Nonetheless, the bottleneck of this process might be the separation and reusability of the catalyst. This accounts for a ball-milling step before the reaction to increase the biomass catalyst effective contact. Such a pretreatment substantially increases the reactivity of the biomass, diminishing solid-solid mass transfer limitations, but hampers the subsequent separation of the spent biomass from the spent catalyst. Given this, a possible solution to improve the sustainable and economic aspects of this process when it comes to addressing future scale-up and possible commercialisation might be the separation and recovery of Ru (Swain et al., 2013) from the spent solid (spent almond hulls and carbon nanofibers used as a support). Then, the recovered Ru could be used for preparing a new fresh catalyst, while the solid carbonaceous material could be subjected to a hydrothermal treatment to produce liquid and solid biofuels concurrently (Remón et al., 2021b), and/or it might be pyrolysed, gasified or combusted to produce energy. This would close the loop and allow the total usage of the original almond hulls and catalyst within a biorefinery unit.A theoretical energy assessment has been conducted to address the viability of the process and provide added value to the spent solid produced. Table 4 shows the elemental analyses and HHVs of the spent solids produced at optimum conditions. This characterisation reveals that the remaining solid is an energy-dense material (26–33 MJ/kg), which combustion could provide the required energy for the process. As the CNF and almond hulls are renewable, carbon-neutral materials, CO2 emissions will be environmentally neutral. According to the thermogravimetric analysis conducted in an oxygen atmosphere, such decomposition occurs at two specific intervals: i.) 200–380 °C and ii.) 400–620 °C (Fig. S.2). The decomposition of the fresh catalyst reveals that this takes place between 420 and 620 °C, which indicates that the former interval accounts for the combustion of the spent biomass (almond hulls), while the latter corresponds to the decomposition of the carbon nanofibres. Table 4 shows the amount of material decomposed at a temperature lower than 380 °C, which correlates well with the catalyst/biomass ratio. Given these data, combusting the spent solid using a temperature lower than 400 °C would be a plausible strategy to recover the spent catalyst from the solid mixture. The theoretical energy for the process could be approximated as that required to heat the reaction mixture (biomass, catalyst, water and H2) from room to the reaction temperature, assuming an ideal adiabatic reactor without heat losses. Thus, considering that most of this energy corresponds to the energy required for heating 20 mL of water (due to the significant excess of water compared to the solid material and H2), Table 4 shows that the theoretical energy for the process shifts between 11 and 17 KJ. Besides, the combustion of the spent solid material provides between 18 and 40 KJ. This corresponds to around 40–60% more energy than theoretically required. Consequently, the controlled combustion of the spent solid material could also serve as a suitable strategy to recover the spent catalyst and provide the energy required for the process.Since the combustion of the spent solid takes place within two intervals, different options arise for process intensification. One might be recovering the Ru from the spent solid and proceeding with the combustion of the whole material. Another could comprise the combustion of the whole material and recover the Ru within the ash content. Then, the Ru recovered could be impregnated onto fresh CNF to prepare more catalyst. Besides, temperature-controlled combustion might also be plausible. This would be intended to selectively combust the biomass content at a temperature lower than 400 °C to separate the biomass from the spent catalyst and provide energy for the process. This would leave a solid material comprising the Ru/CNF catalyst and biomass ashes, which could be recovered and re-used. However, the catalytic properties of this catalyst might have been altered. Therefore, all these options must be carefully compared and studied experimentally in future work.This work has explored the hydrothermal hydrogenation of almond hulls over a carbon neutral Ru/CNF catalyst for the first time. The influence of the operating conditions has been thoroughly addressed, and the process has been optimised for the selective production of value-added liquid species. The processing conditions exerted a significant influence on the hydrothermal hydrogenation of almond hulls, controlling the yields to gas (0–5%), liquid (49–82%) and solid (13–51%) and the chemical composition of the aqueous product. This stream primarily comprised oligomers (46–81 wt%), saccharides (2–7 wt%), sugar alcohols (2–15 wt%), polyhydric alcohols (1–8 wt%) and carboxylic acids (7–31 wt%). The temperature and reaction time influenced the extension of hydrolysis, depolymerisation, deamination, hydrolysis, hydrogenation and dehydration reactions. Additionally, the initial H2 pressure and catalyst loading kinetically promoted these transformations, facilitating the production of different liquid species depending on the other processing conditions. The extensions of these reactions are ruled by the amount of H2 effectively dissolved in the reaction medium and the prevalence of hydrogenations over dehydration/decarboxylation reactions or vice versa. Process optimisation revealed that up to 65–67% of the original biomass could be converted into i.) high-purity (mostly, cello- and xylo-) oligomers alone, and/or ii.) oligomers (31 wt%) and small oxygenates (17 wt% carboxylic acids, 11 wt% sugar alcohols and 6 wt% polyhydric alcohols) concurrently. Combined with the environmentally friendly and holistic features of this hydrothermal process, these promising results exemplify a landmark step change to managing and valorising unavoidable food waste. The energy assessment revealed that the combustion of the spent solid takes place at different temperatures and could provide the energy required for the process. As a result, Ru could be recovered from the spent solid mixture prior to/after its combustion, with the energy produced exceeding the theoretical energy for the process. Alternatively, the temperature-controlled combustion at temperatures lower than 400 °C would also provide energy for the process, leaving a resultant solid material mainly comprising the spent Ru/CNF catalyst. Future experimental work should be directed towards experimentally addressing and exploring all these options, which would close the loop and allow the total usage of the original almond hulls and catalyst within a biorefinery unit.Javier Remón: Conceptualisation, Methodology, Validation, Formal analysis, Investigation, Data curation, Writing - Original Draft, Writing - Review & Editing, Supervision, Project administration, Funding acquisition.Raquel Sevillla-Gasca: Validation, Formal analysis, Investigation, Data curation.Esther Frecha: Methodology, Validation, Formal analysis, Investigation, Data curation.José Luis Pinilla: Conceptualisation, Writing - Review & Editing, Resources, Supervision, Project administration, Funding acquisition.Isabel Suelves: Conceptualisation, Writing - Review & Editing, Resources, Supervision, Project administration, Funding acquisition.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.The authors wish to express their gratefulness to FEDER and the Spanish Ministry of Science, Innovation and Universities (Grant Number ENE2017-83854-R) for providing financial support. Besides, Javier Remón is grateful to the Spanish Ministry of Science, Innovation and Universities for the Juan de la Cierva (JdC) fellowships (Grant Numbers FJCI-2016-30847 and IJC2018-037110-I) awarded. Supplementary material Image 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.scitotenv.2022.154044.

The almond industry leaves behind substantial amounts of by-products, with almond hulls being the primary residue generated. Given that one way to improve food security is by decreasing waste to reduce environmental impacts, developing sustainable processes to manage this by-product is necessary. Herein, we report on the hydrothermal hydrogenation of almond hulls over a carbon-neutral Ru supported on carbon nanofibres (Ru/CNF) catalyst, addressing the temperature, H2 pressure, time and catalyst loading. These variables controlled the distribution of the reaction products: gas (0–5%), liquid (49–82%) and solid (13–51%), and ruled the composition of the liquid effluent. This aqueous fraction comprised oligomers (46–81 wt%), saccharides (2–7 wt%), sugar alcohols (2–15 wt%), polyhydric alcohols (1–8 wt%) and carboxylic acids (7–31 wt%). The temperature and reaction time influenced the extension of hydrolysis, depolymerisation, deamination, hydrolysis, hydrogenation and dehydration reactions. Additionally, the initial H2 pressure and catalyst loading kinetically promoted these transformations, whose extensions were ruled by the amount of H2 effectively dissolved in the reaction medium and the prevalence of hydrogenations over dehydration/decarboxylation reactions or vice versa depending on the catalyst loading. Process optimisation revealed that it is feasible to convert up to 67% of almond hulls into merchantable oligomers at 230 °C, 35 bar initial H2, using 1 g catalyst/g biomass (0.4 g Ru/g biomass) for 360 min. Additionally, decreasing the temperature to 187 °C without modifying the other parameters could convert this material into oligomers (31 wt%) and small oxygenates (17 wt% carboxylic acids, 11 wt% sugar alcohols and 6 wt% polyhydric alcohols) concurrently. The theoretical energy assessment revealed that the total and partial combustion of the spent solid material could provide the required energy for the process and allow catalyst recovery and reutilisation. This environmental friendliness and holistic features exemplify a landmark step-change to valorising unavoidable food waste.

Electrochemically produced hydrogen is considered an essential energy storage technology in the process of decarbonization, connecting intermittent renewable energy sources with our everyday life in a CO2-neutral energy economy. In the temperature range below 100 °C, alkaline water electrolyzers (AWE) which employ liquid alkaline solutions as the electrolyte and proton exchange membrane water electrolyzers (PEMWE) based on proton conductive solid electrolytes mainly govern the market [1,2].Anion exchange membrane water electrolysis (AEMWE) is considered one viable solution to combine the high current densities and (asymmetric) pressure operation of membrane-based electrolyzers with the benefits of the alkaline environment. These advantages feature stable low-cost platinum group metal (PGM)-free catalysts, transport media and bipolar plates made of nickel or stainless steel. In the past years, anion exchange ionomers have reached higher technological maturity concerning hydroxide conductivity [3,4] and chemical stability [5,6] in an alkaline environment. Nevertheless, the durability of state-of-the-art membrane materials is expected to be higher in water than in alkaline solution. Common degradation pathways are, e.g. via nucleophilic attacks of hydroxide on the cationic functionalities in the ionomer [7].Hydroxide conductivity is commonly equated with a high pH environment established by the ionomer and in particular the associated hydroxide ions balancing its fixed positive charges. With such materials present in membrane and catalyst layer, it is expected that the operation of AEMWE systems in pure water is sufficient to thermodynamically stabilize PGM-free catalysts [8]. In the field of fuel cells, this could be shown in multiple studies [9,10]. For water electrolysis, on the other hand, long-term stable operation was not possible so far when using PGM-free catalysts in pure water. Instead, Ir and Pt are often employed for oxygen evolution- (OER) and hydrogen evolution (HER) reactions to achieve stable long-term operation of water-fed AEMWE [11,12], which misrepresents the original intention to replace costly materials. For operation with PGM-free catalysts, additional electrolyte (e.g., KOH or K2CO3) is commonly used to externally establish a high pH environment around the catalyst. Furthermore, these supporting electrolytes increase the system's conductivity allowing to achieve higher current densities [13].In the past, notable AEMWE cell performances were reported for CuCoOx-based catalyst, formerly commercially available as Acta 3030 from Acta S.p.a., curr. known as Enapter. Fig. 1 illustrates the best performing cells from a literature review on previously realized AEMWE performances employing CuCoOx as the OER catalyst in different electrolytes. For further information on the individual works and additional data, see SI. It has to be noted that the compared performance data was reported for different AEM materials, manufacturing strategies, and operating temperatures. Also, the preconditioning for the MEAs varied between the different reports. Nevertheless, this survey clearly shows that operation in KOH solutions resulted in the highest performances reported so far. However, durability has only been investigated by Faraj et al. for 500 h in 1 wt% K2CO3 solution with a low-density polyethylene-based AEM [14]. To the best of our knowledge, no similar reports on long-term stability in water can be found in the literature.Cobalt oxides are widely discussed in literature as potential OER catalysts [15] – even in acidic environments [16]. On the other hand, rather noble Cu has been shown to electrochemically dissolve in various pH and potential ranges [17,18], which renders thermodynamic stability of this catalyst an open question.In the present study, we investigate catalyst- and catalyst layer stability for an AEMWE system with CuCoOx-based anode in different electrolytes aiming to establish a generic understanding of durability determining factors for AEMWE. For this purpose, we focus on the anode only, as this electrode features the most significant potential for cost reduction. We specifically investigate the impact of the liquid feed (pure water and 0.1 M KOH) on the electrolyzer performance and relate it to electrochemical and mechanical stability descriptors representative of the catalyst and membrane employed.CuCoOx as the oxygen evolution catalyst was prepared according to a literature-reported procedure by dissolving 4.5 g CuSO4(H2O)5 and 10 g CoCl2(H2O)6 in 200 mL deionized water and adding 150 mL 0.75 M Na2CO3 solution while stirring. The purple precipitate was filtered off, washed thoroughly with deionized water, and dried under vacuum for 12 h and ground before calcination at 400 °C for 5 h [30]. Product control of the black powder was performed via X-ray diffractometry on a Bruker D8 Advance instrument using a Cu X-ray source. An associated diffractogram is depicted in Figure S1. From energy-dispersive X-ray spectroscopy (EDX), a stoichiometric composition of Cu0.5Co2.5O4 was determined, which is in line with other literature reports [29].Cathode electrodes (cPTEs) were fabricated via a spray coating procedure employing an ExactaCoat device (SonoTek), as reported elsewhere [31]. H24C5 gas diffusion layers with a microporous layer (Freudenberg) were coated with an ink comprised of 1 wt% solids consisting of 10 wt% AP1-HNN8-00 Aemion binder (Ionomr Innovations) and 90 wt% HiSPEC 9100 (Johnson Matthey) in a 1:4 solvent mixture of 1-propanol in deionized water. To obtain the final ink, the catalyst was weighed into a glass bottle and the full amount of water was added. Subsequently, the Aemion binder dissolved in the 1-propanol proportion of the ink solvent was added to the bottle. Homogenization was conducted by immersing an ultrasonic horn (UP200St, Hielscher) for 30 min into the ink while stirring and cooling the bottle with ice. This procedure was constant throughout all electrode batches to ensure a comparable effect of the homogenization procedure on the ionomer [32] as well as potential contaminations [33] from the ultrasonic horn. For spray coating, the gas diffusion layers were fixed on the heating plate associated to the spray coater at 80 °C. The spray path was meander shaped with a pitch of 1.5 mm and an offset of 0.5 mm to ensure a high level of homogeneity of the catalyst layers. The loading was monitored via the weight increase of a reference sample with fixed area, which was also placed on the spray coater. The final platinum loading was 0.5 mg cm−2 in all prepared samples.Anode electrodes (aPTEs) were fabricated by spray coating CuCoOx-inks with a solid content of 1 wt% and different binder:catalyst ratios onto 220 μm thick nickel felts (fiber diameter: 14 μm, 85% porosity, Bekaert). In contrast to the Pt/C based cathode inks but in accordance to previous reports on IrO2-based aPTE systems [31,34], slow particle precipitation of the oxide in the spray coater's syringe during the spray coating process resulted in a higher binder content in the final catalyst layer than in the original ink, even for optimized compositions. Therefore, the final catalyst layer composition for the anodes was determined via thermogravimetric analysis (TGA) of the catalyst layers scratched off the spray coating masks. For this purpose, samples were heated in air from 30 to 1000 °C with a heating rate of 10 K min−1 in a Setsys CS Evo TG-DTA device (SETARAM). A binder content of 2 and 8 wt% in the ink's solid content resulted in approximately 10 and 30 wt% polymer in the final catalyst layer, which was found to be reproducible for the material combination employed in this study. Ink preparation was performed in a similar fashion as for the cathodes employing the same solvent ratio. For the anodes, the ink was prepared the day before electrode fabrication, ultrasonicated for 30 min, stirred overnight and ultrasonicated again right before the spray coating procedure following a similar spray pattern as described above. The nickel felt material was cut into squares of 5 cm2 via laser cutting, deburred and cleaned with 2-propanol and water before fixing them into PTFE frames on the spray coater at 120 °C.For electrode heat treatment, samples were heated to 220 °C in an N2-purged furnace (OTF-1200X-S-II, MTI) within 30 min and kept at this temperature for another 30 min before leaving it to cool down to ambient temperature.Membranes (AP1-HNN8-50 Aemion, Ionomr Innovations) and electrodes were soaked in 1 M KOH solution (prepared from potassium hydroxide Emsure, Merck) for 48 h and 12 h, respectively, before MEA assembly to activate the materials. For electrolysis testing in water, the materials were immersed in pure water before cell testing to remove residual KOH.The cell fixture for electrolysis tests was an in-house setup, comprised of aluminum endplates on either side with inlets for heating elements (HPL, Türk + Hillinger GmbH) and corresponding in- and outlets for the electrolyte. To prevent any contact of the electrolyte and the housing, PTFE tubes were led through the endplates to the flow fields and fixed using a thermoelectric flanging tool (Bola). The MEA was clamped between two identical flow fields made of titanium with a 5 μm gold layer coating. A straight flow design similar to the one reported by Bühler et al. [34] was used in this study. MEA compression was adjusted using reinforced PTFE foils (HighTechFlon) with a thickness of 0.18 mm at the cathode and two individual gaskets at the anode with thicknesses of 0.075 and 0.150 mm to prevent electrolyte leakage. The cell temperature was monitored using thermocouples (Type T, Omega) inserted into the flow fields.In this study, all electrolysis measurements were performed on a self-constructed test bench (Figure S2) employing a peristaltic pump (Masterflex L/S, Cole Parmer) to circulate the electrolyte in PTFE tubing through a heating bath (Aqualine AL25, Lauda) before entering the electrolysis cell with a flow rate of 40 mL min−1. Two 2 L PTFE bottles (VWR) purged with a 100 mL min−1 nitrogen stream were employed as electrolyte reservoirs and gas-liquid separators. An Octostat (Ivium) was used to control electrochemical experiments.After stabilizing the cell temperature, a standardized test protocol was employed. First, a short break-in procedure was performed with subsequent constant voltage holds going in three steps from 1.8 V to 2.0 V and 2.1 V with a holding time of 200 s each. Following this, two polarization curves were recorded with a hold time of 30 s and varying current step sizes as a trade-off between high resolution in the activation region and reasonable measurement time. In the low-current density region from 1 to 10 mA cm−2 the step size was 1 mA cm−2, which was increased to 10 mA cm−2 in the range of 10–100 mA cm−2. Above current densities of 100 mA cm−2, the step width of 100 mA cm−2 was kept constant. Every single current step was followed by a galvanostatic impedance scan between 100 kHz and 100 Hz to determine the high-frequency resistance (HFR) from the Nyquist plot's x-intersection [35]. After a constant voltage hold at 1.9 V, further polarization curves were recorded. A VMP-300 potentiostat (BioLogic) with three 10 A booster boards was employed for long-term measurements after the previous experiments.All electron microscopy experiments were carried out on a Zeiss Crossbeam 540 focussed ion beam scanning electron microscope (FIB SEM). For this purpose, the electrodes were attached to aluminum sample stubs using carbon tabs. The images were recorded with a beam current of 2 nA at an acceleration voltage of 5 kV. A four-quadrant backscatter detector was used for image acquisition.Catalyst spots on glassy carbon (Sigradur, HTW GmbH) were prepared via drop-casting 10 μL of individual inks of CuCoOx and 10, and 30 w% AP1-HNN8-00 binder with a solid content of 5 wt% in 1-propanol. A custom-made three-electrode electrochemical scanning flow cell (SFC) depicted in Figure S3 was used for activity measurements of the catalyst spots. The Ar-purged electrolyte feed (0.05 M KOH or 0.05 M NaH2PO4 buffer) passes a C counter electrode and into the SFC. The catalyst spots on the working electrode can be automatically contacted with the electrolyte of the SFC. A capillary channel inside the SFC connects the reference electrode (Ag/AgCl, Metrohm). The electrolyte outlet is connected to the sample introduction system of an ICP-MS (Nexion 350x, PerkinElmer), which allows online quantification of dissolved species from the catalyst. The ICP-MS sensitivity was calibrated daily using Ge (40 μg L−1) as an internal standard, with a 4-point calibration line mixed from standard solutions (Cu, Co 1000 mg L−1 Merck Centripur). The electrolyte flow rate was measured to be 3.6 μL s−1. For detailed information about the experimental setup, measurement, and calibration procedures, please refer to our previous works [18,36,37].Annealing of SFC samples was performed by heating the glassy carbon plate in an N2 atmosphere in a furnace (GHA 300, Carbolite Gero) to 220 °C in 30 min, keeping this temperature constant for another 30 min and leaving it to cool down to room temperature in an inert atmosphere.Trace elemental analysis of the liquid electrolyte after electrochemical MEA characterization was performed by acidifying 2.5 mL of the anode compartment (total amount: 1 L) with 5 mL 0.1 M HNO3 and injecting it directly into the ICP-MS employing the same calibration line.For swelling experiments, samples of 25 μm thick AP1-HNN8-50 Aemion membranes (Ionomr Innovations) were cut into rectangular pieces of 10 × 50 mm and dried overnight at 50 °C in vacuum. The initial sample size was determined before immersing the membrane pieces into the respective solutions of 0.05 M KOH, 0.1 M KOH, or 1 M KOH for 24 h at either room temperature or 70 °C. The samples for swelling in deionized water were subjected to ion exchange in 1 M KOH solution for 48 h, rinsed, and soaked thoroughly with deionized water to prevent any additional KOH contamination in the system. Dimensional swelling was determined by measuring the resulting sample size according to eq. (1): (1) d i m . s w e l l i n g = ( V f i n a l − V i n i t i a l ) V i n i t i a l ⋅ 100 % Mechanical properties were investigated using the samples from the dimensional swelling experiment after equilibration at room temperature. A Shimadzu EZ-SX universal testing machine with a 100 N load cell was used to perform uniaxial tensile tests. Elastic moduli were determined from stress-strain curves measured at ambient conditions with a constant crosshead speed of 5 mm min−1. At least three samples of each type were measured and the previously measured dimensions of the swollen samples were employed for data fitting.Replacement of costly Ti-based current collectors and precious metal catalysts such as IrO2 represents the most considerable potential for cost-reduction in AEMWE compared to PEMWE. For the anodes in our study, CuCoOx catalyst layers with 10 wt% or 30 wt% Aemion binder were applied to porous nickel felts via spray coating to prepare porous transport electrodes (PTE). Cathodes were based on the same stable Pt/C gas diffusion electrodes employing 10 wt% Aemion [31] in the catalyst layer to exclude chemical cathode dissolution effects. The general MEA design is depicted in Fig. 2 a, whereas Fig. 2 b and c feature scanning electron microscope (SEM) images of the resulting anodic catalyst layer structures with a CuCoOx loading of 2 mg cm−2.For PEMWE MEAs, an optimum (Nafion) binder content of roughly 10 wt% in the final catalyst layer was reported previously by Bühler et al. for PTE-based MEA designs [34] based on Ti porous transport layers with a similar structure. This value was used as a starting point for our investigations anticipating similar behavior of AEMWE and PEMWE operated in pure water. As shown in Fig. 2 d, the cell in pure water feed with 10 wt% binder showed a poor performance reaching only 20 mA cm−2 and decayed to a negligible minimum in less than 1 h of constant voltage operation (data not shown). In an attempt to increase the amount of “alkaline” functionalities around the catalyst, the binder content in the anodic catalyst layers was accordingly increased to 30 wt%. Under the same operating conditions, the onset potential was in a similar range of 1.6 V and the current density could be doubled to 40 mA cm−2 at 1.8 V, still far too low to compete with any existing technology. Thus, similar electrodes were tested with a 0.1 M KOH electrolyte feed. It was found that even by the addition of these low amounts of KOH the performance could be increased by a factor of 20–45 at 1.8 V and lower the onset potential to 1.5 V (Fig. 2 d), which is in good agreement with recent literature [11].An interesting outcome of the above experiment was the different performance trends concerning binder content in the anode. While the cell performance in pure water still improved with increasing the polymer content from 10 to 30 wt%, the opposite was found for alkaline solution. The transition from operation in pure water to dilute KOH entails significant changes in the membrane- and contact resistances due to different swelling behavior of the materials in the respective electrolyte. As the nickel-based porous transport layer is expected to contribute to the catalytic activity of the catalyst (see Figure S4), a simple model system was required.Therefore, a scanning flow cell (SFC) setup (Figure S3) was employed to investigate the effect of different binder contents in a CuCoOx catalyst layer on activity in alkaline- (0.05 M KOH solution) and neutral pH (0.05 M phosphate buffer solution). Surprisingly, the catalyst's activity (Fig. 2 e) did not follow the same trend observed in the full cell setup concerning binder content. Overpotentials determined for a current densitiy of 0.1 mA cm−2 were 1.55 V at pH 12.7 and 1.67 V at pH 7 for the samples with 10 wt% ionomer, whereas for the 30 wt% ionomer samples 1.62 V (pH 12.7) and 1.77 V (pH 7) were observed. In both investigated pH environments, activity is higher for the sample with the lower binder content. In terms of electrolyte pH, even mild KOH concentrations are sufficient to increase the catalyst's OER activity drastically compared to pH 7.The combination of full cell and three electrode measurements suggests that Aemion binder alone is not enough to supply sufficient OH− species to PGM-free catalysts. Only in presence of an alkaline electrolyte (feed), OER performance of CuCoOx was sufficient. This result follows recent doubts in the literature that alkaline pH environments are merely established by the presence of anion exchange ionomers in catalyst layers while feeding pure water [38,39]. Cao et al. even observed that the local pH in an Aemion membrane does not rise above 9.3, where OER occurs, and hydroxide is consequently consumed [40]. However, as discussed below, the binder's properties change with temperature and electrolyte type, which needs to be considered in data interpretation.In three-electrode SFC experiments, it becomes apparent that the binder content lowers the activity of CuCoOx, potentially due to reduced accessibility of the catalytic sites. Similar observations can be found in literature for LSV analysis also for other anion exchange ionomers [41,42]. In our experiments, this binder influence is more pronounced in alkaline solution than at pH 7. Similarly, the higher binder content in the KOH-operated full cell exhibits a slightly worse performance. In this system, the overall resistance is lower due to the electrolyte's high conductivity; the catalyst and membrane connectivity through the binder is no longer vital (primary OH− ion source). Thus, the role of electrode binders in AEMWE differs significantly depending on the feed electrolyte used.After evaluating the effect of pH and binder content on the catalyst's OER activity, the critical question of stability needs to be addressed. The presented findings suggest that the AEM cannot supply enough “alkaline” functionalities in a neutral electrolyte feed, which draws attention to the PGM-free catalyst's thermodynamic limitations. While Co is expected to form stable oxides in alkaline environments and at high potentials, neutral conditions favor the dissolved Cu2+ and Co2+ species [8,15,18]. To better understand the interplay of electrolyte pH, binder content, and catalyst stability, we performed online inductively coupled plasma mass spectrometry (ICP-MS) measurements that track Co, Cu, and I dissolution from the catalyst layer at open circuit potential (OCP).Firstly, it should be mentioned that the iodine leaching rate was used to monitor the ion-exchange process of the ionomer from its iodide form during contact with the liquid electrolyte of the SFC (Fig. 3 ). It dissolves immediately upon contact (marked by the vertical line and an asterisk, ∗) and declines over the first 100 s for both electrolytes, indicating a completed ion-exchange process. The iodine leaching rate further scales with the amount of ionomer in the catalyst layer.It is noteworthy that even at OCP, the catalyst's dissolution differs drastically between the different pH regimes. A first intense dissolution peak is immediately observed for both Cu and Co in a neutral environment upon contacting the catalyst at OCP (Fig. 3 a), indicating the thermodynamic instability at this pH. Afterward, the initially high Cu contact dissolution peaks reside back to the baseline. Such a decrease of Cu leaching over time indicates the dealloying of the catalyst's surface, leaving behind a mostly stable CoOx surface layer. Similar dealloying behavior of thermodynamically unstable metals is often reported in Pt-based PEM fuel cell binary catalyst systems [43–46]. Interestingly, the metals' dissolution behavior in alkaline media is more complex and severely impacted by the polymer activation process. Directly at contact (marked with an asterisk, ∗ Fig. 3 b), the catalyst is significantly more stable in an alkaline environment than in neutral conditions. Co exhibits a negligible dissolution rate within the first 50 s, while Cu shows slight dissolution. However, once most iodide ions are exchanged from the ionomer, the catalyst film suffers from severe particle detachment indicated by the highly noisy dissolution signal (indicated with a pound, #, Fig. 3 b). On the other hand, the severe particle detachment in an alkaline environment points toward a loss of the ionomer's mechanical strength and its ability to guarantee adhesion of catalyst particles in its hydroxide form in an alkaline environment.From swelling experiments with 25 μm membranes made of high IEC Aemion (Fig. 4 a), we observed a significant electrolyte uptake already at ambient temperature, which was very pronounced for the SFC relevant KOH concentrations and pure water. Fortin et al. observed previously that the relationship between swelling and film thickness is not linear for Aemion materials [47]. Therefore, thin films as featured in a catalyst layer (Fig. 2 b and c) are thus expected to expand even more than the membranes employed in our swelling experiments, which could explain the particle detachment in SFC measurements.For ionomers, which have been investigated more thoroughly already, e.g., Nafion, there are quite a few reports in the literature on the effect of heat treatment on mechanical thin film properties [48]. Reduced binder swelling is expected to be favorable for durable catalyst layers [49]. Fig. 4 a suggests that after heat treating Aemion for 30 min at 220 °C, swelling in the different electrolyte concentrations becomes negligible and does not change significantly even at elevated temperatures. Simultaneously, the thermal treatment results in a more than 1000-fold increase in the elastic modulus of Aemion (Fig. 4 b).Considering long-term AEMWE operation, catalyst stability during OER is an essential requirement. Fig. 5 a and b show the dissolution of Co and Cu during a standard galvanostatic hold at 5 mA cm−2 for 3 min in neutral and alkaline electrolyte. Here, the fundamental importance of pH for non-PGM materials becomes apparent. At pH 7 (Fig. 5 a), both Co and Cu dissolve during the OER, which is expected according to thermodynamics. Initially, Cu exhibits an up to ten times higher dissolution rate than Co but declines due to Cu depletion at the catalyst surface. The relatively lower dissolution rate with higher amounts of ionomer is suspected to originate from mass transport of the dissolved ions in the ionomer itself. Similar observations were made for Nafion ionomers in fuel cell environments previously [50]. With the dissolution data at hand, operation in deionized water was no longer considered in this study due to a lack of electrochemical stability of CuCoOx under these conditions. In the alkaline region (Fig. 5 b), on the other hand, both metal concentrations remain below the detection limit of the ICP-MS, while only particle detachment represented by the spikes in the dissolution signal is still observed due to the above-described mechanical phenomenon.Last, we could confirm experimentally that heat treatment of the catalyst layers at 220 °C before the measurements is an efficient way to improve their mechanical stability in SFC measurement. This facile pretreatment step resulted in minimized particle detachment during OER in an alkaline environment, as indicated by the absence of dissolution signal spikes in Fig. 5 c.Some previous works in the literature (overview see supplementary information) report catalyst loadings in the range of up to 30 mg cm−2 for CuCoOx-based anodes [3], while more recent works even implement CuCoOx-based thin films as catalyst layers without any addition of ionomer binder [25,26]. Thus, we aimed to elucidate the influence of the absolute catalyst loading on our AEMWE system's performance operated in a 0.1 M KOH solution. Samples in the range of 1–4 mg cm−2 CuCoOx loading were fabricated. While the overall cell performance improved slightly with higher loadings, the electrolyte reservoir was significantly colored after the measurements with non-heat treated electrodes, which is indicative for catalyst particle detachment. The electrode's heat treatment allowed to significantly improve the cell's performance with higher loadings. Fig. 6 a suggests that the onset potential could be significantly decreased from 1.51 V for the 1 mg cm−2 sample to 1.49 V with 4 mg cm−2 CuCoOx in the anode. As featured in Figure S5, the catalyst layer structure did not exhibit visible changes after the heat treatment, and catalyst layer activity is not expected to be adversely affected (Fig. 4 c).It has to be noted that the electrode heat treatment resulted in a substantial increase in the cell's HFR (in the range of 50 mΩ cm2, see Figure S6) compared to the untreated electrode. However, particle detachment from the catalyst layer during operation could be mitigated by this approach resulting in a clear electrolyte solution even after a 70 h durability experiment for the 4 mg cm−2 samples. After operating the MEA with 4 mg cm−2 CuCoOx loading without additional thermal treatment in 0.1 M KOH at 70 °C for roughly 10 h, the amount of Co and Cu dissolved was found to be 56 μg L−1 and 263 μg L−1, respectively. For a similar sample with additional heat treatment, the amount of Co dissolved in the electrolyte feed was only 3 μg L−1 after the long-term measurement depicted in Fig. 6 b. The amount of dissolved Cu for this sample was 263 μg L−1, which again supports the surface leaching effect also observed for the SFC measurements in an alkaline pH environment (Fig. 5 b and c). Our study, therefore, highlights the positive effect on performance and mechanical properties through heat treatment of Aemion binders in catalyst layers. A better understanding of the chemical structure changes through possible cross-linking as well as optimization of the process remain a challenge for future studies.Fortin et al. reported a similar effect of catalyst layer degradation in MEAs employing Ir black as OER catalyst and the same Aemion binder (7 wt% in the catalyst layer) resulting in degradation rates in the range of ∼3 mV h−1 in 0.1 M KOH at 50 °C and a constant current of 500 mA cm−2 [47]. With our PGM-free anode featuring 10 wt% Aemion in the catalyst layer, it was possible to achieve lower degradation rates of 0.87 mV h−1, even at a higher operating temperature. Moreover, the absolute cell performance is among the best reported for a CuCoOx-based anode (see Figure S7).These results are promising; however, future studies would require longer runtimes to fully evaluate the stability of employed binders for comparison to greater lifetimes as recently shown by Motealleh et al. for AEMWE systems with PGM-free electrodes employing perfluorosulfonic acid (PFSA) binders [51]. Particularly in an alkaline electrolyte feed, the high chemical stability of PFSA binders was found to be favorable for the performance of AEMWE cells [52]. Here, the polymer itself acts as a purely mechanical binder (such as PTFE in the study by Pavel et al. [21]), whereas the hydroxide supply is established by the electrolyte base [53].This study on CuCoOx and Aemion based anodes in AEMWE revealed that the electrode binder's role and behavior differs with changing electrolyte feed. As stability of CuCoOx in a feed of neutral electrolyte could not be enhanced by increased AEI binder content in the catalyst layer, cell operation in pure water was found unfavorable.MEA operation in dilute KOH solution overall resulted in drastically improved cell performances. From the studied cases it could be elucidated that the system does not solely rely on the binder's hydroxide conductivity under these conditions. Further, the high pH environment, essential for thermodynamically stable PGM-free catalysts, is sufficiently established by the external KOH feed. However, for stable operation, a thorough understanding of the ionomer's mechanical properties – particularly in its active form – is vital to guarantee the catalyst's electrochemical stability and a durable adhesion between PTLs, catalyst layers, and membrane. In the case of Aemion, it was found that heat treatment at 220 °C drastically increased the electrode lifetime and should be considered for stable AEMWE operation with PGM-free electrodes. A thorough investigation of the changes within the ionomer structure caused by the heat treatment will be helpful to tailor this process for different kinds of PGM-free OER catalysts.Supporting Information – PDF-file containing additional data and a literature survey on CuCoOx-based AEMWE devices. B. M.: Conceptualization, Formal Analysis, Methodology, Investigation, Visualization, Writing – Original Draft (lead) F. S.: Conceptualization, Formal Analysis, Investigation, Methodology, Visualization, Writing – Original Draft (supporting), M. H.: Investigation, Writing – Reviewing and Editing (supporting), M. B.: Resources, Writing – Reviewing and Editing (supporting), D. A.: Investigation, Writing – Reviewing and Editing (supporting), D. M.: Investigation, Writing – Reviewing and Editing (supporting), S. C.: Supervision, Writing – Reviewing and Editing (supporting), S. T.: Funding Acquisition (lead), Supervision, Writing – Reviewing and Editing (supporting), R. P.: Supervision, Writing – Reviewing and Editing (lead).The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.The authors acknowledge financial support from the Bavarian Ministry of Economic Affairs, Regional Development, and Energy. Furthermore, the fruitful discussions on test bench design with Marco Bonnano, and the help of Dominik Kraus with TGA measurements and Dirk Döhler with XRD.The following is/are the supplementary data to this article: Multimedia component 1 Multimedia component 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.ijhydene.2021.11.083.

Anion exchange membrane (AEM) water electrolysis is considered a promising solution to future cost reduction of electrochemically produced hydrogen. We present an AEM water electrolyzer with CuCoOx as the anode catalyst and Aemion as membrane and electrode binder. Full cell experiments in pure water and 0.1 M KOH revealed that the optimum binder content depended on the type of electrolyte employed. Online dissolution measurements suggested that Aemion alone was not sufficient to establish an alkaline environment for thermodynamically stabilizing the synthesized CuCoOx in a neutral electrolyte feed. A feed of base is thus indispensable to ensure the thermodynamic stability of such non-noble catalyst materials. Particle loss and delamination of the catalyst layer during MEA operation could be reduced by employing a heat treatment step after electrode fabrication. This work summarizes possible degradation pathways for low-cost anodes in AEMWE, and mitigation strategies for enhanced system durability and performance.

No data was used for the research described in the article.While the cement sector is critical for sustainable growth in many countries, it generates large volumes of toxic waste on a daily basis, including CO2 emissions and cement kiln dust (CKD). Toxic waste, such as CKD, may be very harmful to human life, animals, and plants, regardless of whether it reaches the ground, streams, or even air. The dangers associated with CKD arise from its high alkalinity and high concentration of heavy metals [1]. Daily, massive amounts of CKD are generated at two cement plants located in the vicinity of Qena city (Upper Egypt). Additionally, for every ton of cement clinker produced [2], around 54–200 kg of CKD is generated, and 600–700 kg of CO2 gas is released into the environment [3]. Herein, we attempt to suggest successful solutions and safe ways for eliminating this type of industrial waste (CKD) that contributes to environmental pollution while maintaining a balance between the environment and natural resources through waste recycling and mitigating the cement industry's negative environmental impact. Additionally, examining the economic benefits of converting CKD, a potentially harmful byproduct, into a useful material such as impurity-free hydroxyapatite (HAPT). During our literature survey, only one published article was found explaining the formation of HAPT from CKD [4], but their product was found to be mixed with several elements and SiO2.Numerous recently published articles detailed the methods for preparing HAPT from a variety of natural sources, including eggshell [5,6], catfish bones and animal bones [7], cuttlebone and bovine bone waste [8,9], pigeon bone waste [10], porcine byproducts [11], cockle shells or snail shell waste [12,13], turtle shell [14] and natural phosphate rocks [15]. On the other hand, several articles are concerned with the preparation methods of hydroxyapatite, including solid-state and wet chemical precipitation methods [5], hydrothermal porous hydroxyapatite preparation [16], the microemulsion method [17] and replica method [18]. Finally, HAPT with angiogenic-osteogenic properties was prepared for biomedical applications in bone repair via direct refluxing of the cuttlebone with (NH4)2HPO4 [9] or via HAPT modification with gum tragacanth [19].Apart from its biomedical applications, HAPT has a variety of industrial applications as an active compound or catalyst, including catalytic oxidation of volatile organic compounds [20], solventless self-aldol condensation of butyraldehyde to 2-ethylhexenal [21], CO-oxidation reaction [22] and ethanol conversion [23]. Furthermore, HAPT is widely used as catalyst support in many reactions, such as: dry reforming of methane using HAPT-supported nickel catalysts [24], N-oxidation of tertiary amines with H2O2 by W/HAPT [25], and steam reforming of ethanol over cobalt-supported HAPT [26]. Based on our tracing about the different uses of HAPT, as a heterogeneous catalyst, during the conversion of sec-butanol to trans-2-butene or any other products. We were unable to locate any published article that examined this reaction in conjunction with HAPT under any conditions. As a result, our research herein is regarded as the first successful scientific attempt to produce trans-2-butene directly from sec-butanol at relatively low temperatures in a single step. We converted several industrial wastes into pure catalysts early on, such as aluminum dross tailings (ADT) to high surface area γ-Al2O3 [27–29], via γ-AlOOH boehmite, using a variety of precipitation methods and techniques. Additionally, we prepared a series of 1–10 wt% FeOx/Al2O3 catalysts, based on γ-AlOOH recovered from (ADT) and another waste (SPW) steel-pickling chemical waste [30]. These catalysts (1–10 wt% FeOx/Al2O3) calcined at 600 °C demonstrated high activity in ethanol dehydration to diethyl ether and ethene production [30].This ground-breaking work exemplifies a contemporary and relevant set of points, which include the following:(1) For the first time, hydroxyapatite (HAPT) was prepared easily and directly from cement kiln dust (CKD) as industrial waste at 500 °C. As a result, our work herein aims to identify the economic benefits associated with CKD as a harmful byproduct through its conversion into a value-added material.(2) Our simple method results in a HAPT with a hierarchical mesopore structure composed of thin sheets and flakes. Additionally, it is characterized by a dense population of surface acidic sites of varying strength that is uniformly generated and distributed across its surface.(3) This wide range of mesopores and the strong surface acidity of HAPT enabled it to be used as a potent catalyst for the first time during the conversion of sec-butanol (SB). The catalyst showed superior activity in converting SB to t-2-butene with a selectivity of 99% at a low-temperature range of 200–300 °C.A CKD sample was obtained from Misr Qena Cement Plant- Qeft at Qena governorate - Egypt, with a greyish-green colour, where its chemical composition was analyzed by the producer and presented in Table 1. The total wt% of analyzed items in (Table 1) was equal to 92.26%. TG analysis of an original sample of CKD from RT up to 500 °C showed three successive steps with about 7.4% mass loss, Fig. S1 (supplementary information). This could be related to the removal of moisture as well as the desorption of CO2 that is linked to the sample's surface during the storage of CKD in the air for a long time. TG-curve and FT-IR spectrum of original CKD, each one interprets the other and both match well together; see Fig. S1. Furthermore, the clinker in the cement industry contains SO3 and Cl in the form of K2SO4, while chloride form stable compounds with alkalies such as K2O but more volatile than sulphate.Most of the chemicals herein were purchased as analytical grade pure chemicals from El Nasr Pharmaceutical Chemicals Co. Egypt-ADWIC, including nitric acid (HNO3, 50–55%), oxalic acid [H2C2O4.2 H2O], ammonia solution (25%) and diammonium hydrogen phosphate [(NH4)2HPO4].Before HAPT preparation, an estimation of the extent of Ca-ions that would be recovered using the locally delivered CKD was performed. 5 g of CKD sample was added to 50 mL deionized water with continuous stirring using a magnetic stirrer (1500 RPM at room temperature). Then, 12.5 mL of (50–55%) HNO3 was added dropwise until effervescences ceased. The resulting mixture was filtered off through crucible Gooch (Sintered Glass-G3), and the solid residue was washed with deionized water several times. The dissolved portion of CKD in the filtrate was calculated as 4.57 g concerning the weighed undissolved part after the complete dryness of the crucible. The filtrate was kept for use in the next step. Following that, 100 mL of 4% oxalic acid was added to the filtrate, followed by diluted aqueous ammonia solution until the pH reached 8.5–9.0 to precipitate calcium as CaC2O4·H2O. The precipitate boiled, settled, and filtered immediately through ashless filter paper. Finally, after complete dryness of the ashless filter paper, the precipitate was calcined at 500 °C for 2 h in a muffle furnace, then cooled and weighed as = 4.463 g of CaCO3. According to the obtained gravimetric analysis results, we recovered approximately 54.7% of CaO from the dissolved portion, while CaO was calculated to be 50% of the original CKD sample. According to these findings, Ca-hydroxyapatite (HAPT) sample was prepared in a new experiment from cement kiln dust as follows: 10 g of CKD was dissolved in the required volume of nitric acid (H2O: HNO3; 4:1) with continuous stirring and gentle heating of the mixture at 60 °C. To eliminate the undissolved part (UDP) of CKD from the clear solution of Ca-ions, filtration was applied through crucible Gooch Sintered Glass-G3. The calculated amount of (NH4)2HPO4 was dissolved in deionized water and added to Ca-solution in a 250 mL beaker. The resulting mixture was heated at 60 °C over a hot-plate, and then aqueous ammonia solution was added from a burette with continuous stirring until the complete precipitation of Ca-HAPT in a strong basic medium [31] as shown in (Eq.1). (1) 10Ca(NO3)2·4 H2O + 6(NH4)2HPO4 + 8NH4OH →Ca10(PO4)6(OH)2 + 20NH4NO3+ 46 H2O The white precipitate of HAPT was washed several times with deionized water, dried in an oven at 120 °C for 12 h, then ground in a mortar and stored. The obtained HAPT was calcined at 500 °C for 3 h in a 100 mL.min−1 oxygen flow.Thermal analyses of HAPT as prepared were carried out using thermogravimetric (TG) and differential scanning calorimetric (DSC) techniques. These experiments were performed at a 10 °C.min−1 heating rate in 40 mL.min−1 N2-gas flow rate, using a 50 H Shimadzu thermal analyzer-Japan. The instrument is equipped with data acquisition and handling system (TA-50WSI), and highly sintered α–Al2O3 was applied as reference material in DSC experiments.X-ray Diffraction analysis (XRD) of the calcined samples at 500 °C, as well as the original CKD sample, were analyzed by X-ray powder diffraction (XRD) using a Brucker AXS-D8 Advance diffractometer (Germany), equipped with a copper anode generating Ni-filtered CuKa radiation (k = 1.5406 Å) from a generator operating at 40 kV and 40 mA, in the 2θ range between 20°− 80°. The instrument is supported by interfaces of DIFFRAC plus SEARCH and DIFFRAC plus EVA to facilitate an automatic search and match of the crystalline phases for identification purposes with the COD crystallographic database.The FTIR spectra were recorded using a Magna-FTIR 500 (USA), between 4000 and 250 cm−1, operating Nicolet Omnic software and adopting the KBr disk technique.Surface textural properties of HAPT samples that were calcined at 500 °C (viz. specific surface area, pore volume, and mean pore radius) were calculated from nitrogen adsorption-desorption isotherms recorded at liquid nitrogen temperature (i.e.–196 °C) using automatic Micromeritics ASAP2010 (U.S.) equipped with online data acquisition and handling system operating BET and BJH analytical software. All samples were degassed at 200 °C and 10−5 Torr for 2 h before measurements (1 Torr = 133.3 Pa).The morphological characterization of the prepared samples was performed by Scanning Electron Microscopy (SEM) using a FEI Quanta 250 FEG MKII with a high-resolution environmental microscope (ESEM) using XT microscope Control software. EDX dot mapping analysis combined with FESEM was used to determine the homogeneous dispersion of sample constituents. Scanning Electron Microscopy (SEM) images of samples were carried out on a FEI Quanta 250 FEG MKII with a high-resolution environmental microscope (ESEM) using XT Microscope Control software linked to an Electron Dispersive X-ray (EDX) detector. The EDX used was a 10 mm2 SDD Detector-x-act from Oxford Instruments, which utilizes Aztec® EDX analysis software.The total number of acidic sites (sites.g−1) over HAPT at 500 °C was measured using the temperature-programmed desorption (TPD) of tetrahydrofuran (THF, 99 +%, stabilized with 0.025% butylated hydroxytoluene-Sigma) condensed phase, as a probe molecule. The experimental details can be explained as follows [32,33]. 50 mg of sample preheated at 350 °C for 1 h in the air before the probe molecule is exposed. 20 ± 2 mg covered sample with THF were subjected to thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC) analyses at a heating rate of (10 °C.min−1) in dry N2 flowing (40 mL.min−1), using a 50 H Shimadzu thermal analyzer (Japan). The thermal analyzer is equipped with a data acquisition and handling system (TA-50WSI). α–Al2O3 was used as the reference material in DSC measurements. The mass loss due to desorption of THF during TG experiments from the acidic sites was determined to measure the total surface acidity as sites.g−1. The following (Eq.2) is used to estimate the total number of surface acidic sites [34]. (2) Total No . of surface acidic sites ( sites / g ) = moles of desorbed THF x Avogadro ′ s number ( sites / mol ) weight of TG sample ( g ) Catalyst activity experiments were performed at atmospheric pressure in a conventional fixed bed U-shaped quartz reactor. The catalytic activity and selectivity of catalyst samples for decomposing sec-butanol (SB) to products, mostly in the temperature range of 200–300 °C, were investigated. In each experiment, 0.2 g of catalyst sample was preheated at 400 °C inside a fixed-bed continuous flow reactor for 1 h in an airflow (100 mL.min−1) before measurements; then, the temperature was gradually decreased to 200 °C. Sec-butanol (SB) (Fluka, ≥99%) vapors were generated by passing a stream of air (100 mL.min−1) through the liquid SB in a glass saturator thermostatically stabilized at 5, 10 and 15 °C. The gas hourly space velocity of 30 L.gcat −1.h−1 was used in all the experiments. The expected reaction products as cis-2-butene (c-2-butene), trans-2-butene (t-2-butene), and methyl ethyl ketone (MEK) were analyzed and detected using an online gas chromatograph (Shimadzu GC-14), equipped with a data processor model Shimadzu Chromatopac C-R4AD. An automatic sampling was continuously performed using a heated gas sampling cock, type HGS-2, at 140 °C, using a hydrogen flame ionization detector (FID) and a stainless-steel column (PEG20 M 20% on Chrmosorb W, 60/80 mesh, 3 m × 3 mm) at 75 °C. The % conversion of SB and % selectivity of products were calculated [35] using the following (Eqs.3 and 4):- (3) % SB conversion = [ No . moles SB ] i n − [ No . moles SB ] o u t [ No . moles SB ] i n × 100 % (4) Product selectivity ( % ) = [ No . moles of product ] [ Total no . moles of products ] × 100 % [No. moles SB]in and [No. moles SB]out = the number of SB moles in the feed and outlet streams, respectively.The thermal decomposition behavior of precipitated HAPT was investigated using TG and DSC techniques, as presented in ( Fig. 1(a)). Starting from RT and increasing to 500 °C, the TG profile exhibited consecutive steps with % mass loss equal to 24.89%. Additionally, a very small weight loss step between 500 and 600 °C was monitored, associated with a 1.75% weight loss. The corresponding DSC profile recorded for HAPT revealed a series of endothermic peaks in the temperature range of RT-179 °C, due to the slow dehydration step of the precipitated HAPT, followed by a strong endothermic peak associated with the main thermal decomposition process of the prepared HAPT. This enormous step may be ascribed to the decomposition of NH4NO3 [36] that formed between HAPT particles, as presented in (Eq.1), during the precipitation process. Besides, there are two small broad endothermic peaks at 329 and 474 °C. These peaks may be due to the decomposition of the residual traces of NH4NO3 in depth inside the bulk of HAPT particles. Thermal analyses results of precipitated HAPT indicate that calcination at 500 °C for 3 h in oxygen is suitable for preparing well-defined HAPT, as will be discussed later by XRD and FT-IR analyses.To compare the original CKD and the purified obtained HAPT sample, XRD analysis of both samples was employed, as shown in Fig. 1(b). The primary constituent of CKD was tricalcium silicate (C3S) [Ca3(SiO4)O], as confirmed by the most intense Miller index (312) at 2θ = 29.2°, besides other six planes at (013), (203), (101), (204), (313) and (431) which correspond to 2θ = 22.8, 26.3, 31.2, 32.2, 33.8 and 39.1°, respectively (COD 9014362). The second compound in CKD was identified as dicalcium silicate (C2S) [Ca2SiO4] based on a group of diffraction peaks at 2θ = 32.8° (111), 35.7° (122), 41.0° (221), 43.0° (230), 48.3° (240), 56.2° (062), 57.0° (223), 60.6° (170), 64.5° (243) and 72.8° (115) (COD 1546025). The last five diffraction peaks are extremely faint, see Fig. 1(b)-CKD. These results are consistent with data published recently [37,38]. Finally, two diffraction peaks associated with Ca(OH)2 were observed in the XRD pattern of CKD at 2θ = 46.9° (102) and at 50.7° (110) [39]. The X-ray diffraction pattern of HAPT at 500 °C clearly revealed 26 diffraction peaks (Fig. 1(b)), six of which are the most intense peaks for a highly crystalline mixture of both Ca10(PO4)6(OH)2 (COD 1100066) and Ca5(PO4)3(OH) (COD 9002213) at 2θ = 25.2, 26.6, 28.1, 28.9, 42.4 and 45.3° which correspond to (201), (002), (012), (210), (302) and (023), respectively. Additionally, two additional Ca-compounds were identified in the XRD pattern of HAPT prepared at 500 °C, i.e. Ca3(PO4)2 with two diffraction peaks at 30.1° (012) and 33.4° (110) (COD 1521426) as well as a single diffraction peak related to the presence of Ca(OH)2 at 36.6° (002) (COD 9009098). The XRD analysis (Fig. 1(b)) demonstrated that the prepared sample of HAPT was free from any compounds of other elements detected in the XRF of the original CKD, as shown in Table 1. The most intense diffraction line (002) at 26.6°, in the case of HAPT, was used to calculate the crystallite size to be 23.6 nm using the Scherrer equation [40]. Our XRD analysis of HAPT prepared at 500 °C is consistent with results from recent publications [41–43].According to many authors [44–46], the XRD patterns of the prepared HAPT mostly exhibited three main planes (002), (211), and (300) as the main growth planes of HAPT crystals. The plane (211) is greatly sensitive and influenced by calcination temperature as well as the Ca-source used to prepare HAPT [46]. An interesting paper should be considered carefully [46] that recently explained the stepwise growth of the different planes of HAPT, especially plane (211), during heating HAPT with different heating rates, i.e. 3, 6, and 9 °C.min−1. and in-situ recording X-ray diffractograms in the temperature range of 400–900 °C. Therefore, with enlargement of the area in the range of 2θ = 24–35°, in a new window ( Fig. 2), the XRD pattern of HAPT at 600 °C prepared from CKD (sample not included) showed a strong (211) plane, in comparison with the faint very weak plane in case of HAPT at 500 °C (the subject of this paper). On the other side, both (002) and (300) planes sharply appeared in diffractogram of HAPT at 500 °C.Furthermore, to prove that Ca-source plays an important role in the type and structure of the resulting HAPT, we used the same applied precipitating agent, i.e. (NH4)2HPO4 diammonium hydrogen phosphate, and different sources. The other sample of HAPT-NO3 at 500 °C we used as a comparative catalyst from a pure Ca-source (Ca(NO3)2) (in the section of decomposition of sec-butanol) exhibited a completely different XRD pattern from that of HAPT-CKD at 500 °C, see Fig. 2. An attention should be made to the following observations: i) HAPT-NO3 at 500 °C was characterized essentially by the polymorph hydroxyapatite structure [47] with formula Ca5(PO4)3(OH) according to (COD- 9002213). ii) The three main planes (002), (211) and (300) sharply and easily distinguished in XRD pattern of HAPT-NO3. iii) at 500 °C, at the same calcination temperature, the XRD pattern of the corresponding HAPT-CKD gave a very faint plane at (211), as explained earlier, with two sharp planes at (002) and (300). Our prepared HAPT-CKD at 500 °C was identified by matching the recorded diffractogram with standard card (COD-1100066) and was mainly composed of Ca10(PO4)6(OH)2 with little traces of polymorph Ca5(PO4)3(OH). The positions of the recorded planes in the two diffractograms are quite different due to matching each one with a different standard card, i.e. (COD- 9002213) and (COD-1100066). Fig. 3(a) exhibits the FT-IR analysis of the original CKD, as delivered, the precipitated HAPT and HAPT calcined at 500 °C. The spectrum of CKD is completely different from the spectra of HAPT samples. Furthermore, the three spectra have a complex group of bands, especially in the range of 500–1700 cm−1, as shown in Fig. 3(a). In the case of CKD, its spectrum contains a group of bands in the range of 800–1200 cm−1, related to the asymmetric and symmetric stretching vibrations of Si-O bonds [48], due to the presence of both C2S and C3S, as confirmed by XRD analysis (Fig. 1(b)). The bands at 3645, 3432 and 1644 cm−1 are associated with the stretching and bending vibrations of O-H due to adsorbed water and the presence of Ca(OH)2 in CKD [48,49]. A strong absorption band at 1419 cm−1 corresponds to the asymmetric stretching vibration of V 3 C-O due to the formation of CaCO3 in CKD, which is present in atmospheric CO2 [48,50,51]. Moreover, a sharp and small absorption band at 1385 cm−1 could be assigned to the vibration of nitrate [49] that is still adsorbed between CKD particles. Finally, an absorption band appeared at 517 cm−1 that was assigned to Ca-O symmetric stretching vibrations [52] in the spectrum of CKD. The spectra of HAPT as prepared and after calcination at 500 °C for 3 h in oxygen are quite similar and have the same features. Typical absorption bands of phosphate groups were observed in spectra of HAPT in the range of 1112–985 cm−1 that were assigned to ν(PO4 3-) [53], while bands at 662 and 579 cm−1 related to the ν4 bending vibrations of P-O mode in the crystalline HAPT network and the band at 916 cm−1 resulted from the ν1 symmetric P-O stretching vibration [54]. These bands are assigned to the stretching vibrations, symmetric bending and asymmetric bending of vibrations of PO4 3 - [55]. In addition, the broad bands centered at 3445 and 1632 cm−1 indicate the presence of adsorbed water, especially in the fresh precipitated HAPT [56]. Moreover, a broad band at about 1300 cm−1 due to δ(NH4 +) appeared in spectra of HAPT at 500 °C [53], besides a sharp band at 1386 cm−1 that ascribed to the vibration of nitrate [49] as traces that formed during the preparation of HAPT using HNO3. Finally, an absorption band located at 754 cm−1 could be assigned to (P2O7 2-) [57] , and another band related to the symmetric stretching vibrations of Ca-O [58] is shifted to 530 cm−1. Based on the XRD and FT-IR analysis results discussed above, one can conclude that the prepared HAPT at 500 °C is quite an impurity-free sample, as no silicate or other elemental compounds were detected.To get a deep look at the absorption bands, those reflect the structure of the resulting HAPT from CKD, which is not clearly observed in the spectra presented in Fig. 3(a). IR-spectrum enlargement of HAPT-CKD at 500 °C, in the range of 1250–500 cm−1, presented in Fig. 3(b). The expanded spectrum showed a group of stretching vibrations of the phosphate group in HAPT-CKD at 1112 and 1059 cm−1 [53], at 1030 and 983 cm−1 for ν3 vibrational modes of (PO4 3-) [59,60], as well as ν4-mode at 535 cm−1 [59]. Furthermore, the bending vibration modes (ν4) of PO4 3- functional groups were recorded at 662, 603 and 579 cm−1 [31,54]. The stretching vibration of P=O and the bending vibration of P-O are located at 1044, 916 and 559 cm−1, respectively [54,61].The surface characteristics of the HAPT at 500 °C, such as surface area and porosity, were studied using nitrogen adsorption/desorption isotherm, as shown in Fig. 3(c). The figure shows a type-IV isotherm with an H3-hysteresis loop (p/p° >0.9) [62,63]. This clearly indicates the formation of secondary slit-shaped pores of increased size in nanoparticle aggregates [62]. The corresponding pore size distribution calculated using the BJH method from the desorption branch of the isotherm clearly demonstrated the presence of major pores in the range of 2.4 – 6.8 nm with a few pores slightly larger than 11 nm, indicating the evident hierarchical mesoporous characteristics of HAPT, see Fig. 3(c). However, the calculated surface area of HAPT prepared at 500 °C from CKD was 3 m2.g−1. As fundamentals, the abundant mesopores and the hierarchical structure of the HAPT sample would facilitate effective contact of active sites across its surface as a catalyst, as well as mass transport during the catalytic reaction that will be discussed later. The calculated surface area (SBET) of our sample HAPT at 500 °C agrees with recently published values for various HAPT samples prepared from eggshells [64]. Fig. 3(e) exhibits the SEM images of HAPT at 500 °C with different magnifications. At low magnification, the SEM image reveals that the HAPT surface is composed of brittle particles; however, as the magnification increases, more details become visible. Furthermore, the microstructure of HAPT contains a diverse array of mesopores with different pore diameters (as indicated by the yellow arrows in Fig. 3(e). In addition, HAPT particles have a microstructure of transparent thin sheets and flakes (as indicated by red arrows). Each sheet or flake is self-assembled from a large number of loosely packed nanoparticles that connect to form a layered structure, see Fig. 3(e). This distinctive morphology of HAPT particles, as determined from SEM images in Fig. 3(e), offers additional support to the reaction's expected high catalytic properties. The corresponding EDX spectrum obtained from the surface of HAPT confirms the presence of only the Ca, P and O elements, as illustrated in Fig. 3(d). In addition, these results indicated the formation of Ca-deficient hydroxyapatite. Moreover, the calculated weight ratio and a molar ratio of Ca: P elements were 0.78 and 0.61, respectively, compared to the stoichiometric HAPT’s (Ca/P) molar ratio [65] of 1.67. It is worth noting that EDX is a surface technique that reveals only a 10 nm depth below the catalyst's surface, resulting in the discrepancy between the measure and stoichiometric calculation.In order to confirm further the purity of the prepared HAPT from CKD, by a simple method, depending on the dissolution of Ca2+ in a diluted HNO3 followed by separation of this clear solution of Ca(NO3)2 and precipitating HAPT using (NH4)2HPO4 with adjusting the pH of the medium. In addition to the EDX-spectrum (see Fig. 3(d)) that revealed the purity of HAPT-CKD. We performed another accurate technique for the HAPT-CKD, i.e., elemental mapping, as presented in Fig. 4. As an accurate and modern analysis technique, the obtained results of the elemental mapping strongly supported the formation of a pure sample of hydroxyapatite (HAPT-CKD at 500 °C). The whole spectrum revealed that only P, Ca, and O were recorded as only constituents in HAPT at 500 °C (see Fig. 4).Over the last two decades, considerable research has been conducted on the use of HAPT as an active catalyst, either pure or as a catalyst support, in a variety of chemical reactions [20–26]. While none was conducted on the decomposition reactions of sec-butanol (SB), as a result, the work herein represents the first successful attempt to use HAPT as an acidic catalyst to convert SB to trans-2-butene (t-2-butene) via single-step alcohol dehydration. Whether considering the catalyst preparation or the catalytic activity findings, a very high degree of reproducibility has been observed, and the carbon mass balance was almost 100% in all catalytic tests, with no carbon coke formation observed in the spent catalysts. To economically produce t-2-butene from SB in a single step, the following experiments were conducted: (1) The effect of air and N2-gas as carriers The effect of air and N2-gas as carriersTo compare the effect of a costless carrier, such as air and a known-cost carrier, such as N2-gas, on the dehydration reaction. The conversion of SB was studied using 0.2 g of HAPT and Vp of SB of 0.93 kPa in the temperature range of 200–300 °C, using 100 mL.min−1 of air or N2-gas as carriers in two separate experiments with gas hourly space velocity (GHSV) = 30 L.gcat −1.h−1, as shown in Fig. 5(a). At all reaction temperatures, using air as a carrier had a beneficial effect on SB conversion. It increases by approximately 5.2–8.7% with the increasing reaction temperature from 200 to 300 °C, see Fig. 5(a).Furthermore, at a reaction temperature of 200 °C, the % conversion was 40%, then gradually and steadily increased to 91.4 ± 1.8% at 300 °C using an air carrier. As a result, air was used as a costless and effective carrier in all experiments herein. Fig. 5(b) presents the distribution of all products produced during the conversion of SB at the corresponding reaction temperature. The obtained results indicated that the major product was t-2-butene, which had a higher and constant % selectivity of 99%, whereas the other products, c-2-butene and MEK, had a selectivity of less than 1%. Finally, N2-gas, as an inert carrier in catalysis, sometimes do not activate the surface sites over the catalyst responsible for producing the desired product. Therefore, using air as a carrier reduces the reaction costs and initiates the active sites to perform high alcohol conversion at a low temperature, as we studied in Fig. 5(a). Air is superior to nitrogen as a carrier gas in this reaction due to the dissimilar properties of the two gases. Unlike air, which is a mixture of inert and active gases, N2 is an inert gas, and its inertness is due to the covalent bonds formed by the three lone pairs of electrons in its molecule [66]. Thus, the potential of air to initiate the active sites of the catalyst and regenerate the used active sites responsible for producing the desired product is greater than that of nitrogen. Furthermore, GC analysis identified only the product trans-2-butene and unreacted SB in each injection during the reaction in the temperature range of 200–300 °C (see Figs. 5,6,9). (2) Effect of the reactant space velocities on the catalyst reactivity Effect of the reactant space velocities on the catalyst reactivityThree different reactant space velocities were used during the dehydration of SB over HAPT, namely 20, 30 and 60 L.gcat −1.h−1, in the temperature range of 200–300 °C. As shown in Fig. 5(c), increasing the space velocity resulted in a gradual and significant decrease in % of the conversion of SB at all reaction temperatures. This decrease in the % conversion is readily apparent when WHSV is increased from 30 to 60 L.gcat −1.h−1. This phenomenon obviously indicates that at a higher GHSV value (60 L.gcat −1.h−1), insufficient time was allowed for the reacting molecules of SB to convert to products [67]. Furthermore, the WHSV has a significant effect on reaction kinetics, and the high value of WHSV results in a rate-limited conversion [68]. Therefore, we utilized the intermediate value (30 L.gcat −1.h−1) in all experiment, which produced favorable results. (3) Comparison of HAPT-CKD derived from CKD and HAPT-NO3 derived from Ca(NO3)2 Comparison of HAPT-CKD derived from CKD and HAPT-NO3 derived from Ca(NO3)2 To demonstrate the validity of CKD as a Ca-rich source for preparing HAPT, as a benefit of that huge industrial waste-CKD, we prepared another sample of HAPT-NO3 under the same conditions using pure Ca(NO3)2 (BDH-analytical grade). A comparison of the catalytic activities of both samples was performed during the conversion of SB, in the temperature range of 200–300 °C, using WHSV = 30 L.gcat −1.h−1, as shown in Fig. 6. At all reaction temperatures, HAPT-CKD exhibited incomparable activity towards the conversion of SB to t-2-butene. The % conversion of SB using HAPT-CKD, in the temperature range of 200–250 °C, was 9.2–9.8 times that of HAPT-NO3, see Fig. 6. As the reaction temperature increased gradually to 275 and 300 °C, a significant activation occurred for HAPT-NO3. This could be attributed to the distribution of different acidic sites over both samples, as discussed in the following paragraph, as well as the crystallite size of both samples that were determined by XRD analysis (HAPT-NO3 XRD is not shown). The crystallite size of HAPT-CKD and HAPT-NO3 samples were 23.64 and 32.09 nm, respectively. As a result, the crystallite size of HAPT-CKD is smaller than that of HAPT-NO3.To obtain information about the acidic site's capacity and the distribution of these sites over each sample, temperature-programmed desorption (TPD) of tetrahydrofuran (THF) was performed using TG and DSC techniques [34], as shown in Fig. 7. The use of tetrahydrofuran as acidity-probe molecule instead of the known acidity probe molecules such as pyridine and ammonia is due the low surface area of the prepared HAPT catalysts. According to our recent publication [34], the accurate measurement of surface acidity of the catalysts with limited values of surface area requires a relatively moderate basic molecule as THF with pKb = 16.08, rather than strong basic molecules such as NH3 (pKb = 4.75) or pyridine (pKb = 8.77). In the case of HAPT-CKD, the % mass loss due to desorption of THF from the total acidic sites, in the temperature range of 125–500 °C, was 39.43% see Fig. 7(a). This is approximately 1.77 times the calculated value for HAPT-NO3. This clearly indicates that the acidic sites density is higher in the case of HAPT-CKD, with a total no. of acidic sites of 3.29 × 1021 sites.g−1, while it was 1.84 × 1021 sites.g−1 for HAPT-NO3. The DSC-TPD analysis of THF from both samples revealed important information about the distribution of these acidic sites over their surfaces. Fig. 7(c), the DSC-TPD profile of HAPT-CKD showed the presence of weak acidic sites maximized at 148 °C, a big peak of the moderate acidic sites from 163 °C up to 350 °C, and a small peak due to strong sites maximized at 395 °C. On the other hand, DSC-TPD profile of HAPT-NO3 exhibited a very faint peak maximized at 154 °C related to weak acidic sites, a strong peak maximized at 263 °C associated with the presence of moderate acidic sites, followed by a weak peak at 368 °C ascribed for strong sites, as shown in Fig. 7(c). This broad range of acidic sites over the surface of HAPT, including weak, moderate and strong sites, has been previously observed [22] using the NH3-TPD technique. Consideration should be given to the big peak of the moderate acidic sites in both DSC-TPD profiles of the two samples, as shown in Fig. 7(c). In the case of HAPT-CKD, this massive peak can be divided into two sub-acidic sites using Gaussian line profiles, see Fig. 7(b). Furthermore, the calculated no. of moderately acidic sites under this peak was 2.77 × 1021 sites.g−1 in the case of HAPT-CKD, while it is estimated to be 1.53 × 1021 sites.g−1 in the case of HAPT-NO3. As previously explained and illustrated in Fig. 6, this obvious difference in the number of moderately acidic sites present in each sample will significantly affect the catalytic activity of each sample during the conversion of SB to t-2-butene.It is known that the main adsorption sites of HAPT particles are Ca2+ ions as Lewis acidic sites and O atoms of PO4 3- as Lewis basic sites, and the existence of these sites in different proportions is the dominant factor regulating the catalytic activity and selectivity of HAPT [22–24,69]. Bittencourt et al. [69] studied the adsorption properties of a wide range of probe molecules, including CO, CO2, C2H2, CH4, H2, H2O, NH3, SO2, and BCl3, on the surface of HAPT. They found that all the selected probe molecules are adsorbed preferentially close to the most exposed Ca2+ ions. Thus, we believe that the acidic sites involved in the adsorption of THF molecules and in the dehydration of sec-butanol are mainly the Lewis acidic sites (i.e. Ca2+ ions). As a result of the preceding discussion, THF can accurately and quantitatively count the surface acidic sites [34] over HAPT molecules as we proposed based on the published chemical structure of HAPT [70], see Fig. 8(a). The vapour molecules of THF will directly interact with Lewis acidic Ca2+ cationic centers that are abundantly available over its surface, see Fig. 8(b).Our proposed mechanism is quite similar to one recently published [71] for pyridine adsorption over H2-treated CeO2-MeOx samples. As proposed in Fig. 8(c), these acidic sites over the prepared HAPT-CKD catalyst are responsible for the dehydration of SB to t-2-butene [72], with superior activity and selectivity starting from 200 °C up to 300 °C, see Fig. 5(b). Additionally, the perfect distribution of surface moderate acidic sites on HAPT-CKD makes it easy to snipe and firmly bond water molecules over these sites, see Fig. 8(c). This facilitates the dehydration reaction, and the catalyst became more active as the reaction temperature increased from 200° to 300°C. The produced t-2-butene exhibits superior properties to c-2-butene [73], with known values such as kinetic diameter = 4.31 Å, dipole moment (D) = 0.00 and polarizability equal to 81.8 × 10−25 cm−1. Likewise, it was established in a recent study that c-2-butene is consistently less stable than t-2-butene [74] by 0.9–1.4 kcal.mol−1. Soto et al. [75] and Bedia et al. [76] stated that sec-butanol could attack the active sites over the catalyst surface from different directions, giving multiple adsorption complexes. Therefore, the formation of 1-butene, cis-2-butene and trans-2-butene and their abundances depend on the acid-base properties of each catalyst [75,76]. Furthermore, they calculated the equilibrium (cis/trans) ratio in the homogeneous phase to be approximately 0.6. This value showed that trans-isomer formation is favored because it is thermodynamically more stable [75].Numerous articles examined the kinetics of t-2-butene ignition in a jet-stirred reactor and a combustion bomb [77,78], while others investigated and studied the isomeric flames of butene isomers [79] using in situ molecular-beam mass spectroscopy and gas chromatography techniques. Furthermore, t-2-butene and c-2-butene showed promising thermal performance in the solar organic Rankine cycle (ORC) [80]. A recent article also studied the catalytic isomerization and hydroformylation of butenes [81]. Due to the widespread use of 2-butene isomers as chemical intermediates, the global 2-butene market is expected to expand significantly [82] between 2022 and 2028. As a result, the advancement of numerous industries necessitates additional research into producing more low-cost, high-productivity t-2-butene. This is the major merit of the work herein. (4) Stability and reusability of the catalyst HAPT-CKD Stability and reusability of the catalyst HAPT-CKDTwo distinct experiments were conducted to determine the stability and reusability of the catalyst HAPT-CKD at 500 ℃ during the conversion of SB to t-2-butene:(i) A fresh sample was tested under the same conditions, as previously described using an air-carrier and demonstrated conversion of SB ranging from 40 ± 0.8 to 91.4 ± 1.8% in the temperature range of 200–300 °C, see Fig. 9(a). The % selectivity for t-2-butene was always 99%. Following that, the same sample was regenerated for 1 hr by heating at 350–400 °C in air flow (100 mL.min−1). The temperature is then reduced, and the experiment is repeated using the regenerated sample under the same conditions. The used catalyst after regeneration exhibited increased activity towards the conversion of SB at all reaction temperatures, see Fig. 9(a), while the % selectivity toward t-2-butene remained constant. This phenomenon could be explained by increased activation of acidic sites over the catalyst surface in the range of 200–300 °C. (ii) To confirm the catalyst's robustness and stability, a subsequent four-cycle experiment was performed on the same sample under the same conditions, as shown in Fig. 9(b), in the temperature range of 200–300 °C. In the second cycle, the results showed that the sample attained its maximum catalytic activity at all reaction temperatures. After that, the activity of HAPT gradually decreased in the range of 200–250 °C before slightly increasing again. The trend in catalytic activity observed in Fig. 9(b) over the course of the four cycles is likely the result of two opposing effects on the catalyst surface. The first is the effect of exposing the catalyst to air as an active carrier gas, which activates the acidic sites on the surface of the catalyst and increases its catalytic activity. The second factor is exposure to the reaction feed, which exhausts acidic sites and reduces catalytic activity. In cycle 2, the combination of the two effects appears optimal. The primary objective of these successive cycles is to avoid changes in any of the reaction conditions [83], including a) the catalyst weight, b) the distribution of the sample on the catalyst bed, and c) the exact position at which the catalyst is inside the reactor in the vertical furnace during the reaction. Finally, these experiments demonstrated the durability and reusability of the prepared HAPT-CKD, at 500 °C, as a costless catalyst during the conversion of SB to t-2-butene with a high % selectivity of 99%.Furthermore, the stability of any catalyst can be observed during the applied catalytic reaction. The catalyst loses its reactivity due to coke deposition during the applied reaction or poisoning of the active sites by poisoning molecules. In our case, the stability of our catalyst was tested by a simple experiment. Thermogravimetric analysis of the used catalyst, after four cycles of SB-reaction over it, was done twice compared to a fresh sample (see Fig. S2). Due to the strong dehydration activity of the catalyst (HAPT-CKD at 500 °C), the sample lost about 2.3% of its weight on average, while the fresh sample lost only 0.4% of its weight. This is considered good evidence for the stability and constant reactivity of HAPT-CKD at 500 °C, where no coke is deposited over the catalyst's surface.Hydroxyapatite (HAPT) was successfully prepared as a pure nanocrystalline compound at 500 °C from cement kiln dust (CKD) as industrial waste using a simple and cost-effective method. According to EDX, elemental mapping, and SEM analyses, the resulting HAPT at 500 °C is a Ca-deficient structural sample. Moreover, its surface morphology is characterized by thin sheets and flakes, as well as a significant hierarchical distribution of mesopores. This distinguished structure enhances the catalytic activity of HAPT when used as an active catalyst in the conversion of sec-butanol (SB) to t-2-butene at relatively low temperatures in the temperature range of 200–300 °C. Air was utilized as a carrier rather than N2-gas during the conversion of SB to t-2-butene over the HAPT catalyst. The catalyst demonstrated remarkable activity at relatively low temperatures with an extremely high selectivity of 99% towards t-2-butene. The perfect distribution of acidic sites positively reflected on the catalytic activity of HAPT in comparison with another sample prepared from Ca(NO3)2 under the same conditions. We strongly recommend that HAPT be prepared from CKD as a cost-effective and calcium-rich source. Additionally, as demonstrated by SEM micrographs, due to its distinct surface morphology. We invite our colleagues and researchers to utilize our HAPT-CKD sample in biomedical applications involving bone repair. Mahmoud Nasr: Visualization, Conceptualization, Formal analysis, Investigation, Methodology, Software, Writing – original draft, Writing – review & editing. Samih A. Halawy: Writing – review & editing, Formal analysis, Validation Project administration, Supervision. Safaa El-Nahas: Writing – review & editing, Software, Project administration, Supervision. Adel Abdelkader: Writing – review & editing, Software, Project administration, Supervision. Ahmed I. Osman: Writing – review & editing, Conceptualization, Methodology, Proofreading, Project administration.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.The authors wish to dedicate this work to the spirit of the distinguished Egyptian professor Dr Samih A. Halawy who passed away on the 2nd of September 2022. The authors wish to acknowledge the support of The Bryden Centre project (Project ID VA5048). The Bryden Centre project is supported by the European Union’s INTERREG VA Programme, managed by the Special EU Programmes Body (SEUPB). The authors thank Dr Charlie Farrell for proofreading the revised manuscript.Supplementary data associated with this article can be found in the online version at doi:10.1016/j.apcata.2023.119039. Supplementary material .

This research is regarded as the first successful attempt at directly producing highly pure nanocomposite hydroxyapatite (HAPT) from cement kiln dust (CKD) using a cost-effective preparation method. The crystallite size of HAPT was 23.6 nm and showed a hierarchical mesoporous unit with a Ca-deficient structure according to SEM-EDX analyses. HAPT exhibits a high population of different acidic sites, i.e. weak, moderate and strong acidic sites, as determined by TG and DSC-TPD experiments using tetrahydrofuran as a probe molecule. The wide range of acidic sites over HAPT is clearly and positively enhanced its catalytic activity during the conversion of sec-butanol to trans-2-butene. In addition, our prepared HAPT demonstrated greater catalytic activity when sec-butanol was converted using air as a carrier rather than N2-gas. A comparison between the catalytic activity of HAPT prepared from the waste CKD and pure Ca(NO3)2 was also conducted, showing HAPT derived from waste streams with higher catalytic activity.

Sulfur compound contents before reaction, μg/gSulfur compound contents at reaction time t, μg/gSulfur removal, %The emission of sulfur oxides (SOX) in vehicle exhaust can cause serious environmental problems and threaten human health. Thus, countries around the world have set up strict limits on the sulfide content in fuels, and it is very urgent to develop efficient deep desulfurization technology. In recent years, hydrodesulfurization (HDS) has been widely used to remove organic sulfur compounds, which is conducted at harsh reaction conditions (300–350 °C, 2–10 MPa) with expensive hydrogenation catalysts. HDS can effectively remove aliphatic sulfur compounds (Sampieri et al., 2005). Furthermore, removal of aromatic sulfur compounds is difficult because steric hindrance. Some technologies have been developed to overcoming technical defects of HDS, these desulfurization technologies are adsorption desulfurization (ADS) (Luo et al., 2021), extractive desulfurization (EDS) (Li et al., 2013) and oxidative desulfurization (ODS) (Wang et al., 2018), etc. Among them, the ODS technology has been widely researched due to its advantages such as mild operating conditions, low energy consumption, and high removal efficiency of aromatic sulfides. In the ODS process, organic sulfides can be oxidized into sulfones with strong polarity, and then removed by extraction or adsorption (Zhen et al., 2019). Various oxidants, including H2O2, oxygen and organic peroxides have been applied to the ODS process. H2O2 is widely used due to its lower cost, higher oxidation performance and environmental protection. However, the production and storage of hydrogen peroxide may cause safety and cost issues (Liu et al., 2021). Oxygen is not only environmentally friendly, moreover it is easy to get. Thus, oxygen as the oxidant in ODS, it has attracted the attention of researchers (Wang et al., 2021).In ODS, the reaction between oxygen and sulfide is difficult to be carried out under mild conditions. Therefore, it is very important to develop high activity catalysts. In recent years, some catalysts, such as inert metal-based material (Nakagawa et al., 2019), metal organic frameworks (MOFs) (Tang et al., 2020), polyoxometalate (POMs) (Chi et al., 2019), transition metal oxide (TMO) (Wang et al., 2017) etc., are applied to the field of aerobic oxidative desulfurization (AODS). TMO has been widely concerned in ODS due to their higher activity and lower cost (Wang et al., 2020). For example, Wang et al. (Wang et al., 2020) synthesized V2O5 nanosheets with oxygen vacancies by rapid gas drive stripping method. The sulfur removal can reach 99.7% under the optimum reaction conditions. Shi et al., (2016) proposed a simple sol-gel method to prepare Ce–Mo–O catalysts. The catalyst can completely remove BT and 4, 6-DMDBT, the removal rate of BT is 97%, and the oxidation process does not need to add sacrificial agent. Dong (Dong et al., 2019) et al. has developed ultra-thin a-CO(OH)2 nanosheets with molybdate intercalation, its derived Co–Mo–O mixed metal oxide has shown excellent sulfur removal performance. Wu et al., (2020) found that the interaction of strong metal edges between Pt and h-BN can improve the aerobic oxygen oxidation performance in fuel oil. Liu et al., (2020) established Co–Ni–Mo–O mixed metal oxide nanotubes with a hollow structure preferred to application of AODS. The preparation of the above catalysts often requires harsh preparation conditions such as calcination or the addition of organic reagents to control the structure of the catalyst. These defects are not conducive to the industrialization process of aerobic oxidative desulfurization.Cerium molybdate has been widely applied in the fields of inorganic pigments (Dargahi et al., 2020) and photocatalysts (Xing et al., 2016). At present, there is no report of cerium molybdate used as a catalyst for AODS. In this paper, Ce2(MoO4)3 was synthesized and characterized by FT-IR, XRD, SEM, XPS. Compared with the previous TMO, the synthesis of the Ce2(MoO4)3 can be carried out under mild reaction conditions (lower temperature and shorter reaction time), and the reaction process does not need to add structure promoter and high-temperature calcination. The oxidative desulfurization of model oil was determined by using Ce2(MoO4)3 as the catalyst and oxygen as the oxidant. Furthermore, based on the experimental conclusion, the functions of Ce and MoO4 are illustrated in ODS, and proposed possible reaction mechanism.Ammonium molybdate (98 wt%), Cerium Nitrate Hexahydrate (AR), Decahydronaphthalene (AR), Dibenzothiophene (AR), 4,6-Dimethyldibenzothiophene (AR), Benzothiophene (AR), All above reagents were purchased from Aladdin Reagent Co., Ltd.. Oxygen (99 wt%, Hubei Guangao Biological Technology Co., Ltd.).FT-IR of the synthesized catalysts was characterized by a Nicolet FT-IR spectrophotometer (Nexus 470, Thermo Electron Corporation) with spectral range of 4000-400cm-1 and resolution better than 1.5 cm-1. XRD graphics of the Ce2(MoO4)3 sample were obtained using the Philips diffractometer utilizing high-intensity Cu Kα radiation in X’Pert MPD model (40 kV; 100 mA; 1.5406 Å), and the step scan technique at 2 theta angles range between 10° and 70°. Surface morphology of the Ce2(MoO4)3 was characterized by SEM (ZEISS Gemini SEM 500, Germany). The elemental composition of the Ce2(MoO4)3 was studied by energy-dispersive X-ray spectroscopy (EDS). XPS (PHI5000 Versaprobe II, Japan) was used to survey and evaluate the elemental composition and surface chemical state of Ce2(MoO4)3.In this work, the Ce2(MoO4)3 catalyst was synthesized by the reflux method. 0.5296g ammonium molybdate was dissolved in 40 mL distilled water, marked as solution-A, then, solution-A was added to a triangular flask with a cooling condenser. After that, aqueous solution (40 mL) of 0.8744g cerium nitrate hexahydrate was added dropwise to Solution-A. The triangular flask containing the mixed solution was transferred to an oil bath and stirred at 80 °C for 1 h. The pale yellow precipitation was obtained. The products after centrifugal separation were washed three times with absolute ethanol and distilled water, and dried at 90 °C for 5 h.The model oil with S-content of 250 μg/g was prepared by dissolving 0.718g dibenzothiophene in 500 mL decahydronaphthalene. The AODS reaction was carried out in a three-neck flask. Firstly, 20 mL simulated oil and a certain amount of Ce2(MoO4)3 were added to three-necked flask with reflux device. Then, the three-neck flask was placed in a preheated oil bath. oxygen was injected at a flow rate of 0.2L/min. The AODS reaction begins at a certain temperature and agitation speed. A small amount of upper oil phase is taken as sample every 20 min and the sulfur content of the sample was measured by WK-2D microcoulomb analyzer, then calculation of sulfur removal by formula (1). The reaction device is shown in Fig. 1 . (1) η = ( C 0 − C t C 0 ) × 100 % The FT-IR and XRD characterization results of the as-synthesized Ce2(MoO4)3 are revealed in Fig. 2 . As shown in Fig. 2(a), the infrared absorption peaks at 3382 cm-1 and 1616 cm-1 are attributed to stretching and bending mode of O–H from water adsorbed on the surface of samples (Xing et al., 2016), the smaller peak at 1384 cm-1 corresponds to the bending vibration of the Ce–O–H bond. The narrow peaks between 1070 and 1150 cm-1, the peaks around 870, 712 and 634 cm-1 associate to the stretching vibration peaks of the Mo–O bond (Yousefi et al., 2012). Crystallinity of Ce2(MoO4)3 was determined by XRD analysis. The XRD patterns are shown in Fig. 2(b). From the results, typical characteristic peaks of samples associate to Ce2(MoO4)3 crystals (JCPDS: 00-057-0952) and the structure is amorphous nanocrystal (Kartsonakis and Kordas, 2010). The individual peaks at 2θ angles are obtained to 21.19°, 24.47°, 27.61°, 28.33°, 29.99°, 36.39°and 46.34°, which are assigned to (112), (004), (200), (204), (220), (116), (312) planes of monoclinic Ce2(MoO4)3, respectively. No peak of other crystal phases is observed in the spectrum.The morphology of the as-prepared Ce2(MoO4)3 was investigated by SEM. As shown in Fig. 3 (a), the sample is composed of nanocrystals with rod-like structure and particles. The particle size of nanocrystals is uneven due to lower reaction temperature and shorter reaction time. Moreover, nanocrystals were revealed high dispersibility and low packing density. According to relevant reports (Xing et al., 2016), the morphology of Ce2(MoO4)3 can be changed according to different synthesis conditions. The EDS measurement of the sample is shown in Fig. 3 (b), indicates that the sample is composed of Ce, Mo and O elements, and its molar ratio is approximately n(Ce): n(Mo): n(O) = 2:3:12. It is consistent with the composition of cerium molybdate.For the benefit of obtaining the composition and chemical state of catalyst, XPS analysis of Ce2(MoO4)3 catalysts was performed. As shown in Fig. 4 (a), the survey curves of catalysts further demonstrate the coexistence of Ce, Mo and O elements, each element has a spin-orbit core energy level. Among them, the spin-orbit energy spectrum of Ce elements can be divided into two multiple energy levels (U and V), they correspond to the core-level spin-orbit splitting of Ce 3d3/2 and Ce 3d5/2 (Kanai et al., 2017). The high-resolution XPS spectrum of Ce 3d was shown in Fig. 4(b), the U0, V0, U1, V1 peaks corresponding to Ce3+ state and the remaining 5 peaks corresponding to Ce4+ state (Sakthivel et al., 2015). The presence of Ce3+ and Ce4+ indicates that the Ce2(MoO4)3 catalyst has redox properties.The high-resolution XPS spectrum of the spin-orbit core energy level of O1s was displayed Fig. 4(c). The sharp peak observed at 530.51 eV assigned to lattice oxygen in the Mo–O bond. And the side peak at 532.55 eV can be corresponded the O2 was adsorbed on Ce2(MoO4)3 surface (Sakthivel et al., 2015).The XPS spectrum of the spin-orbit core energy levels of Mo 3d5/2 and Mo 3d3/2, the binding energy at 232.68 eV related to the Mo6+ oxidation state of the Mo 3d5/2 spin-orbit core energy level in Fig. 4 (d). The other peak appears at 235.87 eV is the Mo 3d3/2 spin-orbit core energy level of Mo6+ oxidation state, it can be attributed to the existence of Mo=O or Mo–O bond, but no additional peaks of Mo4+ or Mo5+ are observed. These peaks are consistent with the reported XPS spectra of Ce 3d and Mo 3d (Karthik et al., 2017).In order to study the effect of catalyst structure in ADOS, oxidative desulfurization activity of different catalysts such as Ce2(MoO4)3, CeVO4, Ce2(WO4)3 and Na2MoO4 were investigated under the same experimental conditions. As shown in Fig. 5 , The MoO4 2- has the highest sulfur removal in these catalysts owning the same cation. Efficiencies change follows the sequence MoO4 2- > VO4 3- > WO4 2-. Similarly, based on the fact that the sulfur removal performance of Ce2(MoO4)3 is much higher than that of Na2MoO4. It can conclude that Ce3+ also play a significant role in ODS. The above results indicate that the synergistic effect between Ce3+ and MoO4 2- leads to higher sulfur removal.The reaction temperature is an important factor affecting the desulfurization rate. In AODS, the catalyst can show high sulfur removal at high temperature (>100°C). The oxidative desulfurization activity of Ce2(MoO4)3 at different reaction temperatures was shown the Fig. 6 , sulfur removal increases from 10.8% at 80°C to 99.6% at 100°C in 120 min. This is because increased the temperature can accelerate the collision rate between reactant molecules (Zhao et al., 2007). When the reaction temperature was increased from 100°C to 110°C, the sulfur removal rate increased to 99.6% in 80min. However, high temperature will also cause side reactions of hydrocarbons oxidation and the increase of energy consumption (Eseva et al., 2021). In conclusion, 100°C as the optimum reaction temperature.The catalyst dosage is one of the most important parameters for the industrialization of ODS. The effect of Ce2(MoO4)3 dosage on sulfur removal was investigated. The results are shown in Fig. 7 , catalyst dosage increased from 0.02g to 0.05g results in the sulfur removal of DBT increasing from 58.8% to 99.6%. Nevertheless, the catalyst dosage was increased from 0.05g to 0.06g, the sulfur removal decreased from 99.6% to 90.8%. The results show that increasing the amount of catalyst is beneficial to increasing the number of active sites (Qiu et al., 2016). However, excessive catalyst will cause agglomeration and limit the contact area with DBT, which will affect the diffusion of reactants and products, so reduce the sulfur removal (Eseva et al., 2021). Therefore, the optimal catalyst dosage of 0.05g was selected to oxidative desulfurization process.In this experiment, the Ce2(MoO4)3 has a high oxidation desulfurization activity on DBT in AODS. However, it is very necessary to study sulfur removal performance of catalysts for other sulfides such as BT and 4,6-DMDBT due to the diversity of sulfides in actual fuel. As shown in Fig. 8 , the removal rate of 4,6-DMDBT and BT are 94% and 26%, they are lower than that of DBT. According to the literature (Zhao et al., 2017), the difference of removal efficiency can be attributed to the electron cloud density of sulfur atoms and steric hindrance effect. The methyl groups in 4,6-DMDBT would hinder process of oxidative desulfurization reaction, resulting in a lower removal rate of DBT. The higher electron cloud density results the higher the oxidation desulfurization capacity. The electron densities of the S atom in DBT, 4,6-DMDBT and BT are 5.758, 5.760 and 5.739, respectively (Mao et al., 2017). Therefore, sulfur removal of DBT, 4,6-DMDBT and BT are affected by the steric hindrance and electron density (Otsuki et al., 2000).In ODS process, the presence of olefins and aromatics will affect the removal of organic sulfide. Here, toluene and cyclohexene are selected as the models of aromatics and olefins, effects of their addition on sulfur removal are explored in Fig. 9 . Under optimal reaction conditions, when 5 wt%-toluene and 5 wt%-cyclohexene were added to 20 mL model oil, removal efficiency slightly decreased to 97.6% and 94.1%, respectively. It can be seen that both cyclohexene and toluene have a certain influence on the sulfur removal, which may be caused by the competitive reaction among toluene, cyclohexene and DBT (Liu et al., 2021). The results show that the sulfur removal of DBT in the AODS system is less affected by olefins and aromatics.After the AODS reaction, the catalyst was filtered from the reaction system, and then washed with CCl4 under a magnetic stirrer at 25°C for 30min. Finally, the catalyst was dried at 80 °C for 8 h. Thereafter, the recovered Ce2(MoO4)3 catalyst was used in the next AODS under the optimal conditions. The exhibited results in Fig. 10 illustrates that the sulfur removal of DBT reduced to 94.7% after the five cycles. The recovered Ce2(MoO4)3 catalyst was analyzed by FT-IR analysis measurement. The FT-IR spectrum was shown in Fig. 11 . It can be seen that the fresh catalyst and the recovered catalyst have similar absorption peaks, Therefore, the catalyst has high stability.In order to investigate aerobic oxidation desulfurization mechanism, a free radical capture experiment was designed. P-benzoquinone (·O2 - trapping agents) and isopropanol (·OH trapping agents) were added to AODS system, respectively. The experiment results are shown in Fig. 12 . Sulfur removal after adding isopropanol can reach 97.2%, while the sulfur removal after adding p-benzoquinone is only 12%, this indicates that a large number of superoxide radicals [·O2 -] generated during the reaction are captured by p-benzoquinone, leads to reduce sulfur removal. The results show that [·O2 -] radical is the intermediate activated product of oxidation reaction.After oxidative desulfurization reaction, The Ce2(MoO4)3 was washed by CCl4. The CCl4 solution was evaporated by rotary evaporator to obtain oxidation products of the sulfur compounds, FT-IR characterization results of oxidation products are shown in Fig. 13 (a). Two infrared absorption peaks at 1292 and 1166 cm-1, which correspond to the characteristic absorption peaks of dibenzothiophene sulfone (DBTO2). In addition, the obtained CCl4 solution was analyzed by GC-MS measurement. The results are shown in Fig. 13 (b), the strong peak corresponding to DBTO2 (m/z = 216.0) at 17.604 min was found. The FT-IR and GC-MS analysis proved that DBT was oxidized to DBTO2.In the research of aerobic oxidation of molybdenum-based catalysts (Lü et al., 2013; Ma et al., 2020; Xun et al., 2019), molybdenum sites have mixed valence states, the conversion between different valence states leads to the production of active molybdenum species. This experiment is different from these literatures. Through the XPS characterization, it can be seen that there is only Mo6+ in the Ce2(MoO4)3 catalyst. Mo peroxides may be formed in the presence of oxygen. This phenomenon is very common in the oxidative desulfurization system with molybdenum-based catalyst (Jiang et al., 2019; Zhang et al., 2019). Refer to previous research (Shi et al., 2016, Zhang et al., 2019), the mechanism of oxidative desulfurization was shown in Fig. 14 . First, oxygen molecules are adsorbed on the Ce3+ site of the catalyst to form Ce3+-O2, and some of the Ce3+-O2 further formation Ce4+-[·O2 -] superoxide, then these Ce4+-[·O2 -] interact with Mo sites on the catalyst to produce some active molybdenum species to further oxidize DBT compounds to DBTO2 (Shi et al., 2016; Zhu et al., 2007).The aerobic desulfurization performance of Ce2(MoO4)3 for diesel fuel with S-content of 150 μg/g was also investigated. It can be seen from Fig. 15 , under optimal conditions, the sulfur removal of 21.1% was obtained. Compared with the model oil, the sulfur removal is too low due to the complexity of components in actual diesel. After the AODS, 1 mL of acetonitrile was added to the AODS system to extraction of oxidation products of sulfur, removal of sulfur compounds from diesel was reached 74.8%. When extraction desulfurization in diesel fuel was carried out by acetonitrile as extractant, sulfur removal is only 8.2%. The experimental results show that the AODS-extraction desulfurization system under the action of Ce2(MoO4)3 can still remove most of the sulfide in diesel fuel.Ce2(MoO4)3 was synthesized by a simple reflux method using ammonium molybdate and cerium nitrate hexahydrate as raw materials. Ce2(MoO4)3 was used as the catalyst in AODS systems. Sulfur removal of 99.6% for DBT, 94% for 4,6-DMDBT and 26% for BT were obtained under the optimized conditions of 20 mL model oil, catalyst dosage of 0.05g, oxygen flow rate of 0.2L/min, 100°C, respectively. The introduction of both olefin and aromatic hydrocarbon cannot significantly change the oxidative desulfurization activity of the catalyst. The superoxide radical generated by oxygen under the action of catalyst is the key factor for the oxidation of sulfide. The Ce2(MoO4)3 has a strong regenerative capacity and the sulfur removal reached 94.7% after five cycles.The authors also acknowledge the financial support of the Natural Science Foundation of Liaoning Province (2019-ZD-0064); Doctoral Fund of Liaoning Province (201501105).

Ce2(MoO4)3 was synthesized by a simple reflux method using cerium nitrate hexahydrate and ammonium molybdate as reactants. The as-prepared Ce2(MoO4)3 was characterized by Fourier transform infrared spectroscopy (FT-IR), X-ray diffraction (XRD), Scanning electron microscope (SEM), and X-ray photoelectron spectroscopy (XPS). The removal of dibenzothiophene (DBT) in model oil was studied using Ce2(MoO4)3 as catalyst and oxygen as oxidant. The reaction factors such as reaction temperature, amount of catalyst, and sulfide type on sulfur removal were researched. The results prove that both Ce3+ and MoO4 2- play significant role in the conversion from DBT to DBTO2. The Ce2(MoO4)3 catalyst has an excellent performance for the sulfur removal of DBT. Under the optimum reaction conditions, sulfur removal of 99.6% was obtained. After recycling five times, no significant loss in catalyst activity of Ce2(MoO4)3. Mechanism of aerobic oxidative desulfurization was proposed based on the experiment of free radical capture and infrared characterization.

Data will be made available on request.Power-to-gas processes such as the catalytic hydrogenation of CO2 to CH4 are promising strategies to store renewable energy and accommodate fluctuations in energy consumption and production. Efficient CO2 methanation requires the development of active, selective, and durable catalysts[1,2].Solid base metal oxides such as Al2O3, CeO2, TiO2, MgO, and SiO2 are the most applied supports in the methanation reaction as they help facilitate the adsorption of CO2, but a deep understanding of the role of the support, the active catalyst, and the reciprocity between them is still missing in most of the studies, therefore in this paper, we will specifically focus upon Ru and Ni supported on MgO.Studies have shown that different key factors of design catalysts, such as active metal, metal-support interaction, and promotors influence selectivity to methane [3]. Ni and Ru have been some of the most used metals dispersed on different solid supports with high surface area in CO2 methanation [4–11]. Although researchers have primarily focused on Ni-based catalysts due to their relatively low cost and availability. Noble metal catalysts such as Ru-based are prone to less carbon deposition and show higher catalytic performance at lower temperatures [12]. Notably, lowering the reaction temperature results in a thermodynamic hampering of CO formation and hindering the deactivation of catalysts caused by sintering [13–16]. The reaction is performed on large scales thereby, activity and selectivity improvements and a deeper understanding of the reaction and the associated catalysts can have a significant impact the on viability of the technology.There have been considerable investigation and discussion on the mechanism of CO2 methanation, known as the Sabatier reaction, and there are two possible pathways proposed[17–19]. The two proposed overall mechanisms include 1) direct conversion of CO2 to CH4 where formate species are the primary intermediates; 2) conversion of CO2 to CO via reverse water gas shift reaction followed by hydrogenation of CO to CH4 [20].Multiple studies have been dedicated to investigating the underlying mechanism, which demonstrated that the initial reaction that occurs is the dissociative adsorption of CO2 to form adsorbed CO and O (CO2 → CO* + O*). It is found that the rate-limiting step is the cleavage of the C-O bond of the adsorbed species to adsorbed C and O (CO* → C* + O*). This dissociation can take place by either the H-assisted paths with formate or carbonyl hydride as intermediates or by direct dissociation of C-O to its components, C* and O*[21–27]. To complete the catalytic cycle, it is required that C* is hydrogenated by four dissociated H* and desorbed as CH4. Generally, the activity and selectivity of the catalyst are determined by the active metal bond strength to CO and H, which directly dictate the coverage of the surface. Hence, the support plays an integral part in dictating the bond strength of CO* and H* and can help facilitate the desired selectivity. For example, CeO2−x can utilize the oxygen vacancies to directly associate CO2 to CO* . In contrast, MgO reacts with CO2 to form MgOCO2 (or as a normal carbonate, MgCO3) and is hydrogenated by spillover hydrogen provided from Ru which could function as a bifunctional reaction mechanism[28]. However, the mechanism is still being investigated and discussed.Recently, J. Tan et al.[29] improved the catalytic activity of Ni/ZrO2 catalyst using MgO as a dopant to confine Ni active sites. However, the catalytic improvement was limited, and they demonstrated that MgO had no role in the intrinsic activity. Cimino et al.[30] promoted the Ru/Al2O3 catalyst by alkali metals as the base to enhance CO2 capture from flue gas and subsequent methane formation. The study demonstrated that CO2 capture capacity at room temperature improves in the alkali promoted catalysts which resulted in the most active catalyst for CO2 conversion giving site time yield (STY) of 444 molCH4 molRu −1 h−1 at temperatures of 375 °C. Therefore, we propose a catalytic system consisting of a MgO support with basic properties to enhance CO2 adsorption due to the basic-acid interaction and Ru as active sites. Furthermore, applying MgO as the support is demonstrated to reduce catalyst deactivation caused by sintering and carbon deposition[22–24]. To the best of our knowledge, there is no research that has specifically targeted catalytic activity of MgO-supported Ru catalyst for CO2 hydrogenation.In this paper, we show higher activity, methane selectivity, and stability have been achieved through developments of catalysts, including introducing novel synthesis methods, changing morphologies of the support, optimizing the metal dispersion, and enhancing metal-support interaction[29,31–35]. It is notable from the literature that researchers have been showing increasing interest in solid base metal oxide supports in different industries, such as methane to syngas, biomass to fuels, and CO2 [36–44]. Specifically, we focused on synthesizing a high surface area nano MgO support to disperse active metal of Ru and Ni for catalytic conversion of CO2 to CH4. The synthesized materials were characterized by XRD, nitrogen physisorption, SEM/SEM-EDS, in-situ DRIFT, and TEM. For evaluating the catalytic performance of Ru-based catalysts, different loadings of Ru on high surface area MgO support were tested at different temperatures, and the optimum Ru catalyst was compared to its Ni-containing counterpart. The results show that 5 wt % Ru/MgO catalyst results in the highest yield at 375 °C with some initial activity down to 250 °C. At 375 °C, the catalyst also showed high stability over 50 h on stream of conversion.First, oxalic acid (Sigma Aldrich, ≥99.5 %) was dissolved in distilled water and heated to the boiling point, followed by mixing it with bulk low surface area magnesium oxide (MgO) powder (Sigma Aldrich, 97 %) to precipitate magnesium oxalate (MgC2O4). The solid was separated by filtration, washed with distilled water, dried at 80 °C overnight, and then calcined at 500 °C (5 °C/min ramp) for 4 h yielding a high surface area MgO.An adequate amount of the as-synthesized MgO was taken and impregnated via incipient wetness impregnation method with an aqueous solution of either Ru(NH3)6Cl3 (Sigma Aldrich, 98 %) or Ni(NO3)2·6 H2O (Sigma Aldrich, 98 %). Prior to characterization, the Ru and Ni-containing catalysts were reduced to metallic form under a constant flow of Formier gas (10 % H2 in N2) for 2 h at 450 °C and 500 °C (5 °C/min ramp), respectively. Three different Ru concentrations of 3, 5 and 7 wt % supported on MgO were named as 3Ru/MgO, 5Ru/MgO and 7Ru/MgO, respectively. Similarly, 5 wt % Ni on MgO was named as 5Ni/MgO.N2 physisorption was performed at 77 K on a Micromeritics 3Flex surface area and porosimetry analyzer. Samples were outgassed under vacuum at 400 °C overnight before measurement. The specific surface area (SBET) was calculated from the N2 adsorption data by the BET method in the relative pressure range of 0.05–0.3 (P/P0). Micropore volumes (Vmicro) and total pore volumes (Vtotal) were determined using the t-plot method and from a single-point read at a relative pressure of P/P0 = 0.95, respectively.The particle sizes and morphologies were investigated by scanning electron microscopy (SEM) using a Quanta 200 ESEM FEG operated at 20 kV and by transmission electron microscopy (TEM) on a FEI Tecnai T20 G2 microscope operated at 200 kV. All samples were coated with gold for 1 min under 20 mA current prior to SEM or dispersed directly on a holey carbon grid for the TEM analysis. Energy-dispersive X-ray spectroscopy (EDS) was performed using SEM-EDS elemental mapping by studying the sample with electron scanning microscope (Quanta 200 ESEM FEG) operated at 20 keV and equipped with an Oxford Instruments X-Max 50 mm2 EDS analyzer using Aztec 3.3 Service Pack 1 software for data analysis.All synthesized catalysts were characterized with powder X-ray diffraction at ambient atmosphere and temperature with a HUBER G670 Guinier camera in transmission mode using a CuKα radiation from a focusing quartz monochromator. The data was recorded from 2θ of 5–90° over 1 h.Hydrogen temperature programmed reduction (H2-TPR) was performed upon the incipient wetness impregnated sample and carbon dioxide temperature programmed desorption (CO2-TPD) were performed upon the reduced catalysts and was carried out on a Micromeritics AutoChem II 2920 chemisorption analyzer to study the reducibility of MgO-supported metal catalysts and basicity of MgO, respectively. For H2-TPR analysis, the samples were heated under 5 % H2 in He to 600 °C with heating ramp of 5 °C/min while recording the TCD signal. CO2-TPD was carried out by first heating the sample to 500 °C for 60 min under He atmosphere, followed by treating the sample with pure CO2 flow at 40 °C for 30 min. The last step was heating the sample under He to 500 °C (5 °C/min ramp) to desorb CO2 while recording the TCD signal. In-situ diffuse reflectance infrared Fourier transform spectroscopy (in-situ DRIFT) was carried out using Thermo Nicolet 6700 FTIR equipped with a reactor and Praying Mantis diffuse reflectance accessory from Harrick Scientific Products. The experiments were carried out by reducing the sample in a constant flow of Formier gas followed by cooling it to room temperature in Formier gas (10 % H2 in N2). Hereafter, the mercury cadmium telluride (MCT) detector was cooled to 77 K, and the atmosphere was changed to N2 and heated to 400 °C for at least 30 min, and a background spectrum was recorded. Hereafter the atmosphere changed to 25 ml/min CO2, and spectra were recorded every third minute for an hour. After an hour, the atmosphere was changed to 25 ml/min of Formier gas and spectra were recorded every third minute for an hour. Lastly, the atmosphere was changed to reaction mixture consisting of a flow of 80 ml/min Formier gas (10 % H2 in N2) and 2 ml/min CO2.CO2 methanation reaction was carried out in a stainless steel fixed-bed reactor with a diameter of 5.1 mm (PID Eng&Tech, Microactivity Effi reactor). The set-up was equipped with a thermocouple in contact with the catalyst bed to control the reaction temperature, an automatic liquid-gas separator and mass flow controllers for N2, CO2 and H2. The reactor was loaded with 100 mg of catalyst powder (fraction size 180–355 µm) diluted with 600 mg quartz (fraction size 180–355 µm) and fixed with quartz wool. The catalysts were then reduced at 450 °C or 500 °C (in case of Ni supported catalyst) for 2 h with a heating ramp of 5 °C/min under mix flow of 5 ml/min H2 and 45 ml/min N2. The catalytic tests were performed at atmospheric pressure and temperature range of 200–500 °C using gas composition of 80 ml/min H2, 20 ml/min CO2, and 20 ml/min N2 corresponding to gas hourly space velocity (GHSV) of 60000 ml/gcatalyst/h. The system was maintained for 60 min at each temperature setpoint to reach a steady state. The reaction products were analyzed with an online GC (Agilent 7820 A) equipped with a TCD and FID detector. The experimental error was calculated ± 5 %. Fig. 1 shows the adsorption-desorption isotherm for the synthesized MgO sample that exhibits a typical IV(a) isotherm[45]. This isotherm has a H3 hysteresis loop, which may originate from the interparticle voids between aggregated nanoparticles of MgO. The BJH analysis of desorption isotherm for the pore size distribution (PSD) shows a broad peak at around 8 nm. In addition, Table S1. compares textural properties of the synthesized sample and bulk low surface area MgO, which reports considerably higher specific surface area (SBET) and total pore volume (Vtot) for the synthesized MgO sample. Therefore, the results confirm the successful synthesis of a support with high surface. The surface area and pore volume have been increased tenfold by the synthesis, see table S1, where the bulk MgO has a surface area of 22 m2/g and a pore volume of 0.06 cm3/g and the nano MgO has a surface area of 200 m2/g and pore volume of 0.06 cm3/g.Interestingly, it was demonstrated in previous reports[46,47] that size and shape of MgO crystals are important in the population of basic sites, which results in different activity and selectivity in catalysis. Specifically, Coluccia et al. [48] proposed that surface defects in MgO provide under-coordinated O2- atoms that act as strong basic sites. Therefore, we performed CO2-TPD to compare the CO2 adsorption capacity of synthesized and bulk MgO samples.As expected, the XRD analysis of the prepared support showed the typical diffraction pattern for pure FCC MgO[49], see Fig. 2. In the diffractogram for the catalysts containing 3 and 5 wt % Ru, no Ru could be observed due to the low loading. But at the 7 wt % Ru/MgO, metallic Ru could be observed, which goes along with the JCPDS card. These Ru peaks are located at 2θ of 38.4, 44.0, 58.2, and 69.6°, corresponding to planes (100), (101), (102), and (110), respectively[50]. For the corresponding 5Ni/MgO catalyst, the most intense peak from Ni(111) at around 2θ of 43.5° overlaps with the MgO(200) peak at 43.1°. The average particle size of MgO was calculated to be about 32 nm using the Scherrer equation[51] and FWHM for the MgO peak position at 2θ= 43.1°.We investigated the morphology of the high surface area MgO nanoparticles by electron microscopy. The SEM images in Fig. S1 show a rough surface of the relatively large MgO particles. At larger magnification, the SEM analysis confirms that these large MgO particles consist of small, agglomerated nanoparticles. Moreover, SEM-EDS results for 5Ru/MgO show a high distribution of Ru on the support ( Fig. 3e). SEM-EDS of 3Ru/MgO and 7Ru/MgO alongside with 5Ni/MgO samples are provided in Fig. S2 which also show a high dispersion of metals on these catalysts. As the active metals have been impregnated via incipient wetness impregnation, it was expected to have a high dispersion of metal particles. The small particle size calculated with the Scherrer equation supports the good dispersion of metal particles.The TEM images in Fig. 3a) and Fig. 3b) confirm the results from the SEM analysis and show that distinct MgO nanoparticles are agglomerated in larger particles. In addition, atomic fringes are present in Fig. 3b), which is another evidence and confirmation of the crystallinity of synthesized MgO material. The average particle size as measured from 150 particles in the TEM was around 35 nm, which is in good agreement with the particle size of MgO as estimated from the XRD analysis using the Scherrer equation (32 nm). Fig. 3c exhibits the high dispersion of Ru nanoparticles on 5Ru/MgO after reduction at 450 °C for 2 h. The histogram indicates that most Ru nanoparticles are between 2 and 4 nm in size (Fig. 3d). The average particle sizes of Ru in 3Ru/MgO and 7Ru/MgO are 2–4 and 4–6 nm, respectively. The corresponding particle size histograms are given in the supporting information (see Fig. S3).Results in Fig. 4a show that synthesized MgO has a larger CO2 adsorption capacity which could be correlated to its higher surface area and population of basic sites as reported previously[52]. Furthermore, different desorption temperatures indicate different strength of CO2 bond to MgO. This observation determines that CO2 is adsorbed partially at temperatures below 180 °C in the synthesized MgO which is likely favorable for the catalytic conversion of CO2 at low temperatures. Fig. 4b-e shows the H2-TPR results of the fresh catalysts after incipient wetness impregnation with the metal precursors and drying at 80 °C for > 24 h. The three samples with increasing Ru loadings (Fig. 4b-d) share two main peaks around 220 °C and 305 °C. We assign these peaks to the stepwise reduction of Ru3+ to Ru0 [53,54]. Some differences in the reduction profiles indicate some complex speciation related to the precursor loading. Nevertheless, all samples were fully reduced at temperatures above 425 °C. Based on these results, we decided to reduce all the Ru-based catalysts at 450 °C. Fig. 4e) shows that the complete reduction of the Ni-based catalyst occurs at around 366 °C corresponding to Ni2+ → Ni0 [55]. Fig. 5 shows the catalytic performance of different synthesized catalysts tested for conversion of CO2 to CH4 at temperature range of 200–500 °C and GHSV of 60000 h−1. As expected, the conversion in all catalysts increases with the temperature until it gets limited by thermodynamic equilibrium. Both 5Ni/MgO and 7Ru/MgO have some activity at 275 °C while 3Ru/MgO needs to be at 300 °C to activate. Impressively, the 5Ru/MgO catalyst already converts CO2 to CH4 at temperatures down to 250 °C. The catalysts also achieved the highest conversion and selectivity at different temperatures. Among the Ru-containing catalysts, the 5Ru/MgO catalyst achieved the highest yield of CH4 (54 % conversion and 98 % selectivity) at 375 °C. This may be explained by the higher metal loading compared to 3Ru/MgO and the higher metal dispersion compared to 7Ru/MgO. The catalytic tests also show that 5Ru/MgO is more active and selective than 5Ni/MgO. The 5Ni/MgO catalyst resulted in 45 % conversion and 95 % selectivity at 450 °C. Under the given reaction conditions, this corresponds to a site time yield (STY) of 263 molCH4 molNi −1 h−1. For comparison, the 5Ru/MgO catalysts resulted in a STY of 520 molCH4 molmetal −1 h−1.Further catalytic test results show that catalytic conversion of CO2 to CH4 is impossible over pure MgO supports (Fig. S4), which suggests that metal active sites are required to boost activity and selectivity. Furthermore, higher catalytic performance of 5Ru/MgO compared to 5 wt % Ru on bulk MgO with low surface area (Fig. S5) demonstrates the important role of high surface area in synthesized MgO from two perspectives. Firstly, the higher surface area of MgO provides higher metal dispersion and smaller metal particle size that result in improved catalytic performance as already discussed. Secondly, the population of basic sites is higher in high surface area MgO as confirmed from CO2-TPD, which persuades higher CO2 adsorption and activation. Table S2 in the supporting information compiles the performance of the synthesized catalysts in this work and recently reported Ni and Ru-based catalysts tested under similar catalytic conditions. These data show that the 5Ru/MgO catalyst presented here has higher STY compared to the recently reported catalysts [5,8,29,30]. In-situ DRIFTS studies were conducted to further study the difference between catalytic performances of Ru on bulk and Ru on synthesized nano MgO supports by mapping the species present at the surface of MgO supports under different gas compositions. Fig. 6a and Fig. 6c present results for Ru supported on bulk MgO and nano MgO, respectively, at 400 °C under 100 % CO2 over 60 min. Comparing the two spectra in Fig. 6a and Fig. 6c, there are certain similarities in the spectra range of 1025–1100 cm−1, which include the peaks that are assigned to monodentate carbonates, the peaks located at 1300–1600 cm−1 assigned to magnesium carbonate, MgCO3, and additionally the peak at 2078 cm−1 designated to the Ru-CO* carbonyl peak. The presence of the surface carbonyl peaks indicates dissociative addition mechanism of CO2 on CO. The peak at 2094 cm−1 has previously been attributed as COad which is at a ruthenium-oxide interface, or CO which is co-adsorbed with Oad. Interestingly, the peak located at 1980 cm−1, which is likely a COad species is either adsorbed to Ru with lower coordination, very small Ru nanoparticles, or MgO [56,57]. Furthermore, the concentration of the peak at 1980 cm−1 decreases slowly and steadily with time for the nano MgO (Fig. 6c), whereas in the case of bulk MgO (Fig. 6a) the peak only appears in the beginning[56,58–60]. The peak which solely appears on nano MgO is a shoulder at 1681 cm−1, which is assigned to a bidentate carbonate. A decrease in signal at 1862 cm−1 is seen when Ru on nano MgO is exposed to CO2, and this indicates potential catalyst oxidation, which may be a result of an interaction between the lattice O and the ruthenium nanoparticles [56].In Fig. 6b) and d), 10 % H2 in N2 is introduced and there is an immediate large decrease in the concentration for all the surface species and clear formation of methane (peak range of 2900–3100 cm−1). Methane formation stops after 4 min for both nano MgO and bulk MgO. Noticeably, the peak occurring at around 1980 cm−1 decreases at the same rate as the methane formation for both samples. Furthermore, the Ru-CO* peaks at 2094 cm−1 disappear immediately. On the other hand, the carbonate species located at 1300–1600 cm−1 decrease slowly and this diminish is faster in nano MgO sample (Fig. 6d) than in bulk MgO (Fig. 6b). This observation is likely an indication of more readily available carbonates due to the smaller MgO particle size [61,62]. In general, the IR-study shows that the main catalytic pathway follows the bifunctional mechanism as suggested by McFarland et al.[28]. We assign the peaks between 1300 and 1600 cm−1 to the formation of large amounts of MgCO3 on the catalyst's surface.Under H2, the decrease in MgCO3 and transient increase in adsorbed carbonyl species between 1980 and 2100 cm−1 indicate that the carbonates are involved in the reduction of CO2 to CO intermediates. Finally, these intermediates are hydrogenated into the methylene groups that appear as a small shoulder at 1300–1305 cm−1 before further reduction and methane desorption.Under reaction conditions with both CO2 and H2, Fig. S6, shows clear peaks from MgCO3 at 1300–1600 cm−1, the CO intermediates at 1980 cm−1, and the methyl intermediates at 1305 cm−1. In contrast, the peaks from Ru-CO* between 2078 and 2094 cm−1 disappear. This indicates that the Ru species are short-lived [24] and that the conversion of the carbonyl intermediate may be the rate-determining step[63].We reproduced the results by repeating the test on 5Ru/MgO fresh catalyst, which confirms the repeatability of obtained results for 5Ru/MgO (Fig. S7). In addition, we investigated the stability of 5Ru/MgO and 5Ni/MgO for 50 h at 375 °C and 450 °C, respectively, with GHSV of 60,000 h−1. Fig. 7 shows high stability of 5Ru/MgO catalyst over 50 h time on stream compared to 5Ni/MgO catalyst, which suffers from deactivation and a decrease of both the conversion and selectivity. Since 5Ru/MgO catalyst performs better at a lower temperature, the active sites are less prone to sintering and deactivation, while Ni containing catalyst requires higher performing temperature that leads to faster deactivation. Thus, we tested the 5Ni/MgO catalyst at low temperature of 375 °C for comparison. Moreover, the GHSV was decreased to 7500 h−1 to achieve almost as high conversion as for 5Ru/MgO at the same temperature. As Fig. S8 presents, both catalysts exhibit high stability and selectivity toward CH4 at 375 °C. However, the resulted STY for 5Ni/MgO is about 30 molCH4 molmetal −1 h−1 which is significantly lower than of for 5Ru/MgO (520 molCH4 molmetal −1 h−1) performed at the same temperature. Overall, 5Ru/MgO catalyst seems to be the best catalyst with the highest catalytic performance amongst the tested catalysts for CO2 hydrogenation to CH4. The samples were also characterized after the stability test. The XRD of the spent sample (Fig. S10), does not show any significant change, however, the isotherms and pore size distribution changed significantly (Fig. S11). The isotherm of 5Ru/MgO appears similar to the fresh sample. The pore volume has decreased due to the impregnation of Ru. However, the 5Ni/MgO has changed significantly, though the composition hasn’t changed. This could be a sign of the pore structure is not stable at the higher temperature, which is required for the Ni catalyzed methanation.This work introduced a simple method to synthesize high surface area nano MgO with high crystallinity. The high surface area MgO was used to support highly dispersed Ru and Ni nanoparticles and tested for CO2 methanation. TEM and SEM-EDS showed high dispersion of 5 wt % Ru on high surface area MgO support. The CO2-TPD results showed that high surface area MgO has higher CO2 adsorption capacity than low surface area MgO. The catalytic results show that the catalysts are highly active, and the 5 wt % loading of Ru on MgO catalyst has the highest activity at lower temperatures resulting in STY of 520 molCH4 molmetal −1 h−1 at 375 °C. This catalyst outperformed Ni-containing MgO catalyst (STY of 30 molCH4 molmetal −1 h−1 at 375 °C) even though both show high catalytic stability at 375 °C.Our FT-IR study shows that both MgO and Ru play a decisive role in the reaction mechanism and indicates that the catalyst facilitates the bifunctional reaction mechanism. Furthermore, the study suggests that the conversion of carbonyl intermediates is the rate-determining step.Based on this, we believe that the introduced synthesis method is a facile approach to make high surface area MgO, which is an efficient support to disperse active metal sites in addition to chemisorbing and activating CO2 for catalytic conversion of CO2 to CH4 process. Farnoosh Goodazi: Conceptualization, Methodology, Validation, Formal analysis, Investigation, Writing – original draft, Writing – review & editing, Visualization. Mikkel Kock: Conceptualization, Methodology, Validation, Formal analysis, Investigation, Writing – original draft, Writing – review & editing, Visualization Jerrik Mielby: Conceptualization, Methodology, Formal analysis, Investigation, Resources, Writing – original draft, Writing – review & editing. Visualization, Supervision. Søren Kegnæs: Conceptualization, Methodology, Formal analysis, Investigation, Resources, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration, Funding acquisition.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This work was supported by Independent Research Fund Denmark (grant no. 6111-00237 and 0217-00146B), Haldor Topsøe A/S, Villum fonden (Grant No. 13158) and the European Union’s Horizon 2020 research and innovation program under grant agreement No 872102. The authors gratefully acknowledge the Department of Chemistry, Technical University of Denmark (DTU) for the support of the project.Supplementary data associated with this article can be found in the online version at doi:10.1016/j.jcou.2023.102396. Supplementary material .

This work shows that Ru nanoparticles supported on high surface area nano MgO is a highly active and selective catalyst for CO2 methanation, which is a promising method to store renewable energy and limit the emission of greenhouse gasses. We studied the effect of the Ru loading on MgO supports with different surface areas and compared the results to the corresponding Ni-based catalyst. Our results show that high surface area MgO containing 5 wt % Ru has the highest activity. This catalyst was stable for more than 50 h and resulted in 54 % conversion at 375 °C, which, under the given reaction conditions, corresponds to a site time yield of 520 molCH4 molRu −1 h−1. For comparison, the Ni-based catalyst only resulted in 45 % conversion at 450 °C with a low selectivity to CH4 (STY=263 molCH4 molNi −1 h−1). Furthermore, Ru on high surface area MgO catalyst was already active at low temperature of 250 °C due to chemisorption and activation of CO2 on the MgO support, which is promising for low-temperature CO2 methanation.

Renewable energy sources, that minimize environmental impact, have been the basis of several studies in the last few years to find aiming an alternative to substitute the use of fossil fuels. Low-cost materials have been used for biofuel production, which contribute to reducing the damage caused by fossil fuels (Sammah and Ghiaci, 2018). Confirming this trend, the global biodiesel production in 2020 was estimated at around 36.9 million metric tons (Huang et al., 2020). Furthermore several studies have shown a market recovery in 2021, reaching the average production of 46 billion liters between 2023 and 2025 (International Energy Agency, 2020).Biodiesel is a sustainable alternative to the demand of diesel usage which enables of decrease pollutants emission such as CO2 (Mostafa and El-Gendy, 2017). This product can be classified as first-generation (derived from edible vegetable oils), second-generation (non-edible raw material or waste), and third-generation (algae biomass) biodiesel (Ramos et al., 2019). Some studies have been developed to optimize the synthesis process of biofuel, which normally proceeds by triacylglyceride transesterification with methanol or ethanol in the presence of catalysts that may have basic (NaOH, KOH, CH3ONa and CH3OK) or acid (H2SO4, H3PO4, HCl, and R ─ SO3H) characteristics. However, these catalysts are more difficult to be separated and reused, in addition to being able to cause secondary reactions such as saponification (Farias et al., 2020; Manríquez-Ramírez et al., 2013).In this context, heterogeneous catalysts have been applied for the biodiesel production, because they minimize the problems arising from the use of homogeneous catalysts, enabling a more efficient purification process (Luo et al., 2017; Niju et al., 2016). Several feedstock have been used in this catalysts production, such as plant and animal-derived compounds, metallurgical/mining industrial residues and natural clays. (Rizwanul Fattah et al., 2020). A few heterogeneous catalysts used to produce biodiesel are sodium silicate for soybean biodiesel (Guo et al., 2012), silica coated on Fe3O4 magnetic nanoparticles (Thangara et al., 2019), ground alkaline metals for palm biodiesel (Salamatinia et al., 2013), K2O/NaX and Na2O/NaX for Safflower biodiesel (Muciño et al., 2014), and alkaline metal catalyst (Li, Na and K) supported on rice husk silica for WCO biodiesel (Hindryawati et al., 2014).Search for greater economic viability and reduction in the environmental impact of the biodiesel production process has motivated the usage of waste biomass (Gollakota et al., 2019). Thereby, the WCO may be a good choice since it is cheaper than the refined oil (Vela et al., 2020), non edible and it can reduce the environmental impact caused by its inappropriate discarding (Gollakota et al., 2019), as around 5 million tons of refined oil are consumed per year around the world (Tan et al., 2019). Meanwhile, WCO normally presents relatively high free fatty acids that require pretreatment by esterification, before the transesterification, in order to avoid the soap formation (Ding et al., 2012).This work contributes to the development of novel heterogeneous catalysts (SPS, sodium potassium silicates) for producing biodiesel from WCO. Moreover, the alternative MPI silica was applied for the catalysts production after its modification with alkali hydroxides. Despite being a waste material, there was no need of WCO pre-treatment (acid esterification) and the direct transesterification was accomplished using SPS. In this process, it was possible to use the catalyst for five reaction cycles. The use of waste raw material (WCO) for biodiesel production and a natural source (MPI silica) for the catalyst preparation are relevant factors that ensure low-cost and environmentally friendly biodiesel production process.Waste cooking soybean oil was obtained from Brazilian restaurant. The other reagents were: Silica beach sand (MPI silica); Hydrochloric acid (HCl, Synth, 36.5%); Sodium hydroxide (NaOH, Vetec, 99%); Potassium hydroxide (KOH, Synth PA); Ethanol (C2H5OH, Dynamic PA, 96%); Sodium chloride (NaCl, Dynamic, 99%); Methanol (CH3OH, Vetec, 99.8%); Sodium sulfate (Na2SO4, Vetec, 99%); Benzoic acid (C7H6O2, Dynamic, 99.5%); Phenolphthalein (C20H14O4, Vetec PA ACS) and Distilled water. The chemicals obtained were used as received.Amorphous silica MPI was synthesized from beach sand using a methodology developed in this research group and described by de Carvalho et al. (2015). In this work, the SPS catalysts were obtained by calcination of mixtures containing alkaline hydroxides (NaOH and KOH) and MPI silica, at 450 °C in a muffle furnace for 4 h. Different molar ratios of NaOH:KOH:MPI silica were used (1:1:1; 2:1:1; 1:2:1, and 1:1:2), to produce the catalysts named SPS, SPS-1, SPS-2, and SPS-3 respectively. The obtained catalysts were applied in preliminary tests for the biodiesel production, using the following synthesis conditions: 3.5% (w/w, catalyst/WCO), 9:1 (molar ratio A/O), 2 h, and 70 °C.The catalysts were characterized by various techniques. X-ray diffraction analyses (XRD) were performed using a Bruker D2 Phaser device (Bruker AXS, Madison, WI, USA) with CuKα radiation (λ = 1.5406 Å), 30 kV filament, 10 mA current, Ni filter and a LYNXEYE detector in the range from (2θ) 5 to 50° for MPI silica, and 5 to 70° for the catalysts. Elemental analysis was conducted on a Bruker S2 Ranger X-ray fluorescence (XRF) Spectrometer (Bruker AXS, Madison, WI, USA) using Pd or Ag radiation (max. power 50 W, max. voltage 50 kV, max. current 2 mA, XFlash® Silicon Drift Detector). Fourier transform infrared spectroscopy (FTIR) was accomplished in a Shimadzu IRAffinity-1 spectrometer (Columbia, MD, USA) with attenuated total reflectance (ATR). Spectrum analysis variation was in the range of 600–4000 cm−1 with a resolution of 4 cm−1 and 32 scans. The thermal analysis was performed in a thermo microbalance (TG-209-F1-Libra, Netzsch, Selb, Germany) using an alumina crucible for measuring 10 mg of the samples with a continuous heating rate of 10 °C min−1 in nitrogen (N2(g)) purge gas at a flow rate of 20 mL min−1.The morphologies and chemical composition of catalysts were obtained using a field emission scanning electron microscope (FESEM, Auriga, Carl Zeiss, Oberkochen, BW, Germany) and energy dispersive X-ray spectroscopy (EDX, XFlash Detector 410-M, Madison, WI, USA), respectively.The analysis of CO2 desorption at programmed temperature (CO2-TPD) consisted of weighing a 200 mg of the SPS catalyst. Next, the pre-treatment was performed with heating at 200 °C under a N2 flow of 16 mL min−1 for 1 h. At the end of this period, the temperature was reduced to 60 °C and the CO2 flow (16 mL min−1) was inserted into the reaction line to start the adsorptive process for 30 min. The analysis was subsequently started in a He (g) atmosphere in the temperature range of 40 °C to 500 °C with a heating rate of 10 °C min−1 min under He (g) flow (16 mL min−1). The desorbed CO2 was then quantified by a thermal conductivity detector.The deconvolution methodology was applied to the CO2-TPD, and TG/DTG of catalyst and biodiesel using PeakFit 4.12 software (Systat Software, Inc., Berkshire, UK) by applying Gauss + Lorentz for better curve adjustment, Savitzky-Golay smoothing filter (<20%) and linear baseline.Hammett’s basicity test was performed using the titration method with acid-base indicators. A phenolphthalein indicator (H_= 9.3) and a 0.01 mol/L methanolic benzoic acid solution were used as the titrant for the experimental procedure. The Hammett basicity test consisted of stirring 0.15 g of the catalyst with 2 mL of methanolic indicator solution at a concentration of 0.1 mg mL−1 for 30 min at 300 rpm. The obtained data were then applied in Eq. (S1) and (S2) (all figures, tables and equations indicated with S are in the supplementary material) to calculate the number of basic sites from the basicity calculation.Raman spectra were obtained using a confocal Raman microscope (LabHAM HR Evolution, HORIBA Scientific), with laser wavelength of 532 nm, grade: 600 gr mm−1, laser power of 1% and scanning range 400–4000 cm−1. N2 adsorption–desorption isotherms at 77 K were used to determine the textural parameters of the SPS catalyst in a Micromeritics ASAP 2020 apparatus (Norcross, GA, USA). The specific surface area (SBET) was calculated using the Brunauer – Emmett – Teller (BET) equation.Biodiesel was synthesized in three consecutive stages: (i) previous filtration of the WCO in a separating funnel to remove impurities; (ii) transesterification reaction at 70 °C with methanol and catalyst in a reflux reactor; and (iii) purification according to the methodology described in the literature (Fernandes et al., 2012), with adaptations. The time, catalyst concentration and molar ratio parameters were varied for synthesis optimization.Hydrogen Nuclear Magnetic Resonance (NMR 1H) and Carbon-13 (NMR 13C) one-dimensional analyses were obtained by a Bruker Avance III HD NMR SPECT. 300 Spectrometer operating at frequencies of 300.13 MHz for hydrogen (1H) and 75.47 MHz for carbon (13C) respectively. The WCO and biodiesel were dissolved in deuterated chloroform (CDCl3) in the proportion of 20 mg of sample to 0.5 mL of solvent. The chemical shifts (δ) were expressed in parts per million (ppm) and Tetramethylsilane (TMS) was used as an internal standard. Eq. (S3) was used to calculate the conversion of esters (Gohain et al., 2017).The thermal analysis (TG/DTG) was accomplished in a thermo microbalance (TG-209-F1-Libra, Netzsch, Selb, Germany) using nitrogen (N2(g)) as purge gas at a flow rate of 20 mL min−1, alumina crucible, heating rate of 10 °C min−1, 10 mg of sample and final temperature was 600 °C. The deconvolution was performed using the Peakfit v.4.12 software.The physical–chemical properties were obtained under conditions for density at 20 °C and kinematic viscosity at 40 °C according to the American Society for Testing and Materials (ASTM) D4052 and D7042, respectively. The yields of the transesterification reactions were calculated using Eq. (S4) (Yang et al., 2016). The acidity index was obtained applying Eq. (S5) (AOCS, 2009).The regeneration was performed in three stages: (1) washing with mixture of hexane and ethanol (50 mL, 1:1 v/v) under stirring for 1 h, (2) soaking in a new mixture of hexane and ethanol (50 mL, 1:1 v/v) for 4 h to remove excess glycerol and waste cooking oil that could remain on its surface and (3) oven drying for 2 h at 150 °C. This procedure was repeated for five cycles.The overall methodology employed in this research is presented in a simplified flow diagram depicted in Fig. 1 .The catalysts syntheses were performed with different proportions of MPI silica, NaOH, and KOH in order to evaluate the reaction yield under the reaction conditions described in Table S1. (S indicates supplementary material). The selected SPS catalyst was obtained with lower proportion (1:1:1) of the reagents, since the others did not significantly affect the reaction yield for WCO biodiesel production, Table S2. However, the concentrations of the species obtained by XRF that correspond to the oxides of the components showed differences related to the feed molar ratio (Do Nascimento-Dias et al., 2017), which probably occurred due to the utilization of sodium and potassium hydroxides that reacted with silica forming the metal silicates and producing the catalyst, with the possibility of varying the concentrations. In addition, it was possible to verify that pure silica did not have the ability to catalyse the reaction, indicating the need to modify its structure and composition.The XRD pattern for MPI silica (Fig. 2 a) exhibited a broad peak centered at 2θ angle 22.8°, characteristic of an amorphous silica structure with the presence of short-range order in atomic clusters (Salakhum et al., 2018; Stanishevsky and Tchernov, 2019), the crystalline plane phases of the SPS catalyst corresponding to Na2O (2θ = 32.23°, ICSD 060435), KCl (2θ = 40.72°, ICSD 044281), K2(Si2O5) (2θ = 28.57°, 31.77° and 58.87°, ICSD 280480), K2O (2θ = 41.50° and 66.53°, ICSD 060489), SiO2 (2θ = 13.10° and 25.87°, ICSD 065497), K2CO3 (2θ = 29.87°, 31.75°, 31.77° and 66.47°, ICSD 000662) and Na2(Si2O5) (2θ = 32.80° and 50.47°, ICSD 080378) were shown in Fig. 2b indicating that the modification of the MPI silica to obtain catalytic sites was successful.The XRF results for the SPS catalyst and MPI silica were described in Table 1 and allowed to confirm the elements presence of found in the XRD phases, revealing the presence of alkali oxides as the main components of the produced SPS. These compounds have the ability to promote the catalytic activity of the transesterification reaction (Chouhan and Sarma, 2013), since mixed metal oxides are present as an interesting class of solid heterogeneous catalysts, allowing the association of the various oxide phases that promote appropriate characteristics for the reaction process (Lee et al., 2016).FTIR spectra of MPI silica and SPS were presented in Fig. S1 and Table S3. For both materials the band at 1417 cm−1 indicates the existence of Na2CO3 formed by the reaction of sodium hydroxide with atmospheric CO2 (Belmokhtar et al., 2016; Simanjuntak et al., 2014). The broadband in the region of 2500 and 3750 cm−1 may be attributed to OH groups from silanol and adsorbed water (Hindryawati et al., 2014). For MPI silica the band at 1314 cm−1 is associated with asymmetrical stretching of the siloxane (Si-O-Si) bonds (Hindryawati et al., 2014). Additional description of the SPS FTIR spectrum were provided in Table S3.A comparison of both spectra, allowed observing a decrease in the intensity of the bands in the range of 3526 to 2451 cm−1 in MPI silica, probably due to moisture loss during the silica calcination to produce the SPS. Additionally, the decrease of a signal at 1606 cm−1 and disappearance of the band at 1314 cm−1, in the SPS catalyst might have occurred due to a reaction of the silane’s groups polycondensation during SPS preparation. The increase in intensity of the band at 1427 cm−1 is attributed to the carbonates formed after the treatment of MPI silica to produce SPS (Peyne et al., 2017). Moreover an enlargement of the peak at 1000 cm−1 (Si-O-Si group in the MPI silica) in the range of 1000–800 cm−1 is noted, which could be probably due to the overlap of some peaks present in the SPS. The changes in the silicon bands would be caused by depolymerization of the silica network by the potassium ions, which are network-modifying agents and may affect the number of atoms in the first coordination sphere of the atomic silicon (Puligilla et al., 2018).The results of the thermogravimetric analysis for the MPI silica exhibited three main events characteristics of amorphous silica (Fig. 3 a). The first event of mass loss (4.8%) occurred in the range of 27.39 °C to 190 °C, due to the removal of physisorbed water on the silica surface. The second mass loss event occurs over a wide temperature range of 190 to about 632.86 °C (mass loss of 2.22%), probably due to the condensation of less stable silanols and of silane into siloxane. The third event (9.31% of mass loss) starting at 632.86 °C may be attributed to the more stable silanols were dehydroxylated and condensed (De Carvalho et al., 2015; Kin et al., 2009).The results of the thermal analysis for SPS, Fig. 3b, showed (I) a 5.75% of mass loss attributed to the water molecules adsorbed on the material (34–190 °C) and (II) a second mass loss event of 1.35% attributed to the silanols and silane into siloxane condensation, of the 190 °C to 604.96 °C, and (III) third mass loss event of 10.38%, release of water by condensation/polymerization of the Si-OH groups above 604.96 (He et al., 2010; Kin et al., 2009). The DTG deconvolution result of the SPS catalyst, Fig. 3c and Table S4 confirmed the three mass loss events verified in Fig. 3b. This shows the stability of the material after removing the water by heating for activating the catalyst.The FESEM micrograph for the MPI silica revealed the presence of irregularly distributed particles (Fig. 4 a). The EDX maps showed a homogeneous distribution of the elements Si, Al, O, and Na on the surface of the support (Fig. 4b-e). FESEM micrograph for SPS (Fig. 4f) indicated larger irregular structures in relation to the image obtained for the MPI silica (Fig. 4a). Upon modification of MPI silica with alkali treatment and calcination the surface heterogeneity (with steps and kinks) and particle agglomeration are enhanced. The mapping images (Fig. 4g-j) demonstrated that there is good dispersion of the elements Si, Na, K, and O over the catalyst. The morphological changes in the silica MPI which acts as a support presented formation of the SPS catalyst by adding alkali metals and the adopted synthesis methodology. The EDX maps and spectra for both materials, SPS (Fig. 4k) and MPI silica (Fig. S2) are in agreement with the XRD (Fig. 2) and XRF results (Table 1).The CO2-TPD results of SPS (Fig. 5 a) exhibited several desorption peaks, indicating the presence of with weak (100.67–149.71 °C), medium (149.71–273 °C) and strong (273–405 °C) basic sites, demonstrating the heterogeneity of the material (Eom et al., 2015).The deconvolution of the CO2-TPD (Fig. 5b and Table S5) enabled to verify that the largest proportion of basic sites are strong characteristic, composing about 51.9%, with 31.6% corresponding to sites of medium strength.The results of the Hammett basicity test were positive to identify that SPS has basic sites in the range of 9.3 < H_ <15 of phenolphthalein (pKb = 9.8), corresponding to 3.2 mmol g−1 of the indicator, which denote the presence of a substantial quantity of active sites at alkaline pH in the SPS (Okoye et al., 2019). According to the specialized literature, sodium and potassium silicates have a basic strength above the phenolphthalein range (range 15 < H_ <18.4) (Hindryawati et al., 2014), justifying the catalytic capacity of the SPS catalyst (this study). The obtained results of basicity are consistent with those of CO2-TPD.In the Raman spectrum of SPS catalyst, Fig. 6 , there were peaks in the range of 500–700 cm−1 that could be attributed to vibrational stretching and bend of the Si-O-Si (Santos et al., 2019; Partyka and Leśniak, 2016). At 965.20 cm−1 there was a peak that corresponds to the antisymmetric stretching of the Si-O bond (Zhu et al., 2019). At 1059.84 cm−1, a peak of strong intensity was attributed to the symmetrical stretching of the C-O of the K2CO3 group (Ma et al., 2021). A peak at 1074.96 cm−1 was assigned to stretching CO3 2– of the MgCO3 (Williams et al., 1992).The N2 adsorption–desorption isotherms of SPS (Fig. S3) were classified as type III at relative pressures of 0.1 < P/Po < 0.6, and type IV(a) at 0.6 < P/Po < 0.98, with an H3 hysteresis cycle (Thommes et al., 2015). The specific surface area (SBET) obtained for the SPS catalyst is 0.710 m2 g−1 and the pore volume is 0.00421 cm3 g−1. This demonstrates a sharp drop when compared to the MPI silica that presented SBET of 33.54 m2 g−1 and pore volume of 0.18 cm3 g−1, according to Carvalho et al. (2015). This decrease may be associated with the addition of K+ and Na+ and also due to the particle agglomeration upon calcination as seen in FESEM images. Strong basic sites that may occlude the catalyst pores (Farias et al., 2011).In accordance with the obtained XRD, XRF, and Raman results, three species make up the SPS catalyst: potassium carbonate, alkali metal oxides, and sodium and potassium silicates. In this work, a mechanism was proposed for biodiesel synthesis by transesterification reaction, adapted from Guo et al. (2012). This proposal is based on the silicate interactions (Fig. 7 ), in which the methanol approaches the catalyst surface favoring the ion exchange between the metal silicate (Na or K) and the hydrogen from the alcohol forming the methoxide. There is a subsequent nucleophilic attack of the methoxide on the carbonyl of the triglyceride forming the tetrahedral intermediate that after rearrangement results in the fatty acid methyl esters (FAME). Lastly, an intramolecular rearrangement of protons in the diglyceride occurs to stabilize the charge.The reaction yield (%) for the biodiesel synthesis (calculated by Eq. S4) was assessed at different conditions of time, molar ratio of alcohol to oil (A/O), SPS catalyst concentration and its reuse (Fig. 8 ). It was observed that the yield progressively increased with increase in time (Fig. 8a), catalyst concentration, and molar ratio of A/O (Fig. 8b-c), as determined through Eq. (S4). The tests for the SPS reuse revealed a good catalyst performance as the reaction yield was approximately the same during in the first four cycles, with a decrease only in the fifth cycle (Fig. 8d). The factors that most influenced the yield were time and A/O molar ratio from 12:1 to 15:1.The 1H NMR spectra for the waste cooking oil (WCO) (Fig. 9 a) and biodiesel (Fig. 9b) were obtained with the purpose of evaluating the biodiesel purity and the conversion to esters in the SPS-catalysed transesterification reaction. The conversion may be confirmed by the disappearance of the peaks at 4.1–4.3 ppm, attributed to hydrogen from triglycerides (WCO) (Ruschel et al., 2016), and the appearing of a signal at 3.7 ppm due to hydrogen from methoxy groups of methyl esters (biodiesel) (Gohain et al., 2017), resulting in a conversion of around 93.89% (Eq.(S3)). In addition, absence of contaminants is observed.In the 13C NMR spectra of WCO and biodiesel (Fig. S4), carboxyl ester group (signals at 174.3 and 54.43 ppm) and olefinic groups (unsaturated methyl ester, signals in the range of 130.18–129.73 ppm) respectively were observed (Fig. S4a). Furthermore, triglycerides peaks (CH-O and CH2-O) are found between 68.88 and 61.77 ppm (Fig. S4a) (Tariq et al., 2011). A peak referring to methyl carbon appears at 14.9 ppm, while carbons of methylene groups (hydrocarbon chain) were observed in the range of 34.9–22.48 ppm (Fig. S4b) (Gohain et al., 2017). Comparing the results for WCO and biodiesel, it is possible to observe the disappearance of triglyceride peaks and the appearance of ester carbonyl signal (C-O) at 51.43 ppm in the 13C NMR of biodiesel (Fig. S4b) (Tariq et al., 2011).Thermal analysis of WCO and biodiesel produced from WCO (Fig. 10 a and 10b) also confirmed the conversion and quality of biodiesel produced with the SPS catalyst, in agreement with the 1H NMR analysis (Fig. 9b). The DTG deconvolution method was performed on the DTG curve of biodiesel (Fig. S5 and Table S6) to evaluate its components. A mass loss event for WCO is observed in the range of 347–475 °C, (Fig. 10a) which corresponds to the decomposition of triglycerides while for the biodiesel the mass loss (Fig. 10b) occurs at lower temperatures in the region of 163 to 266 °C due to the decomposition of smaller molecules, corresponding to monoalkyl ester (Dantas et al., 2007; Misutsu et al., 2015). The SPS showed similar behaviour as other catalysts studied in the specific literature, which are described in Table 2 .It was possible to observe four components (Fig. S5) which corresponds to the observed mass loss in the TGA curve of biodiesel. Of these, three are attributed to monoalkyl ester, corresponding to 98% of the integrated area of DTG, and the other peak corresponds to oxidation products formed (about 2%) (Díaz-Ballote, 2018). It is possible to prove the synthesis occurrence via the thermal evaluation of biodiesel, which corroborates the 1H and 13C NMR results. The values calculated for the physicochemical properties such as density at 20 °C (0.890 g cm−3), kinematic viscosity at 40 °C (5.061 mm2 s−1) and acidity index (between 0.355 and 0.155 mg KOH g−1) were within the specifications of National Agency of Petroleum Natural Gas and Biofuels (ANP N° 45, 2014),. Characteristic bands of biodiesel were observed in the FTIR spectrum of biodiesel (Fig. S6). The band at 1736 cm−1, was attributed to the elongation of the carbonyl bond; absorptions in the 1425–1447 cm−1 correspond to the asymmetric CH3 flexion, and the bands in the range of 1188–1200 cm−1 correspond to O-CH3 stretching (Mumtaz et al., 2012, De Morais et al., 2013).The sodium potassium silicate (SPS) catalyst obtained in this work has a simple preparation method and may be considered of low-cost since it was produced from MPI silica derived from beach sand. In addition, exhibited high catalytic activity for biodiesel production for the conversion of waste oil (WCO) without previous of free fatty acids esterification. The efficiency of the catalyst synthesis method was verified by the XRD and XRF results, as well as the deconvolution method using the CO2-TPD result to evaluate the strength of basic catalyst sites. The biodiesel production process involving heterogenous catalysis showed some advantage as the purification steps are more efficient than the homogeneous one, reducing mainly the amount of waste water. On the other hand, the reuse cycles of the catalyst indicate the possibility of its application in industrial scale. The reaction time was the most decisive parameter for the biodiesel yield (92%) and its high conversion (93.89%) can be verified by the 1H NMR results. The spectroscopic, thermal analysis, and physicochemical data suggest the biodiesel from WCO was suitable for use as fuel.This research was funded by Higher Education Improvement of Coordination Personnel - Brazil (CAPES) - Funding Code 001. Keverson G. de Oliveira: Writing – original draft, Conceptualization, Investigation, Formal analysis. Ramoni R.S. de Lima: Conceptualization, Investigation, Writing – review & editing, Formal analysis. Clenildo de Longe: Investigation, Writing – review & editing. Tatiana de C. Bicudo: Investigation, Writing – review & editing. Rafael V. Sales: Investigation. Luciene S. de Carvalho: Supervision, Writing – review & editing.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.The authors acknowledge the support provided by the Post-Graduate Chemistry Program (PPGQ/UFRN), the Energetic Technologies Research Group (GPTEN), and the Central Analitica (IQ/UFRN). This study was partly financed by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior – Brasil (CAPES) – Finance Code 001.

Heterogeneous catalysts, named SPS (sodium potassium silicates), were synthesized with an alternative silica (MPI silica) obtained from beach sand. In this work, the MPI was modified with NaOH and KOH producing silicate-based catalyst for biodiesel synthesis from waste cooking oil (WCO). The obtained catalyst was characterized by XRD, CO2-TPD, the Hammett basicity test, XRF, FESEM, EDX, FTIR and TG/DTG. The results confirmed the presence of K2O/Na2O oxides and their silicates, the main active sites responsible for the catalytic action. CO2-TPD and the Hammett basicity data suggested the presence of weak, medium and strong basic sites. Biodiesel yield was about 92% and the SPS catalyst was reused for five cycles. The biodiesel conversion by NMR 1H was about 93.89%. The DTG deconvolution revealed the decomposition of four typical biodiesel compounds (R2 = 0.9987). The method applied for the WCO biodiesel production using SPS catalyst represents an environmentally friendly process, based on low-cost material and reuse of waste biomass.

Catalysts play a key role in today’s chemical industry. About 75% of chemical processes are based on catalytic reactions and over 90% of new processes developed in recent years use catalysts (Hagen, 2015). Catalysts may enhance reaction kinetics by orders of magnitude or provide the selectivity required for economic production of chemicals. Consequently, the selection of a suitable catalyst is a key step in chemical process design. Today, catalysts are usually selected based on experimental methods such as high-throughput experimental screenings or combinatorial chemistry. However, these methods are usually not target-oriented and may lead to a huge experimental effort. Moreover, the chemical design space of possible catalyst molecules is vast (Fink et al., 2005; Reymond, 2015). Therefore, it is not feasible to test all potential catalysts experimentally and the full potential of the chemical design space is thus likely not exploited. Consequently, it is highly desirable to develop in silico-methods to explore the chemical design space and to suggest the most promising candidates for experimental testing.To identify the most promising molecules in large design spaces efficiently, Computer-Aided Molecular Design (CAMD) (Austin et al., 2016) methods have been developed. CAMD methods explore the chemical design space in silico, typically based on optimization algorithms that are employed to find the best candidates. As described in our recent review article (Gertig et al., 2020c), CAMD methods comprise of 3 building blocks: 1. An algorithm is required to explore the chemical design space and to suggest and optimize molecular structures (Papadopoulos et al., 2018). Typically, the design spaces are defined by a set of functional groups or molecule fragments. Structures of candidate molecules are generated from these groups or fragments. The optimization of these structures can be based on deterministic or stochastic optimization (Papadopoulos et al., 2018). 2. CAMD requires a sound prediction of such unknown properties, e.g., molecular and thermodynamic quantities as well as chemical properties such as reaction kinetics. The reason for the need to predict properties in CAMD is that CAMD methods are supposed to not only examine already known molecular structures, but also to suggest new candidate molecules with unknown properties. Thermodynamic properties are commonly predicted in CAMD based on group-contribution (GC) methods (Papadopoulos et al., 2018; Gmehling, 2009). GC methods offer the advantage of a straightforward implementation and computational efficiency (Gani, 2019). However, the chemical design space accessible with these methods is limited to the functional groups a GC method was trained for. Moreover, the prediction of different properties often requires several GC methods. In contrast, the computationally more demanding quantum chemical (QC) (Atkins and Friedman, 2011) methods are not limited to certain functional groups. In conjunction with thermochemistry (Paulechka and Kazakov, 2017; Umer and Leonhard, 2013), QC methods provide consistent predictions of molecular and thermodynamic properties as well as kinetics of chemical reactions (Vereecken et al., 2015). Due to the availability of increased computational capacities, even the use of advanced QC methods in CAMD has become feasible in recent years (Gertig et al., 2020c). 3. The performance of the designed molecules needs to be evaluated during CAMD based on a chosen objective. Commonly, CAMD methods use simple performance indicators as objective function defined based on predicted molecular, thermodynamic or chemical properties (Papadopoulos et al., 2018). However, such indicators may not capture all aspects and trade-offs relevant for the intended use of the designed structures. Thus, CAMD using simple performance indicators likely results in the design of sub-optimal molecules (Adjiman et al., 2014). Preferentially, CAMD should evaluate all candidate molecules directly based on their intended application (Gertig et al., 2020c). In the context of process design, the designed molecules are applied in processes e.g., as working fluids, solvents or catalysts (Papadopoulos et al., 2018). Thus, each candidate molecule should be evaluated using process optimizations to determine the process performance that can be reached using the candidate. The integration of process optimizations into the design procedure corresponds to the extension of CAMD to integrated Computer-Aided Molecular and Process Design (CAMPD) (Papadopoulos et al., 2018). An algorithm is required to explore the chemical design space and to suggest and optimize molecular structures (Papadopoulos et al., 2018). Typically, the design spaces are defined by a set of functional groups or molecule fragments. Structures of candidate molecules are generated from these groups or fragments. The optimization of these structures can be based on deterministic or stochastic optimization (Papadopoulos et al., 2018).CAMD requires a sound prediction of such unknown properties, e.g., molecular and thermodynamic quantities as well as chemical properties such as reaction kinetics. The reason for the need to predict properties in CAMD is that CAMD methods are supposed to not only examine already known molecular structures, but also to suggest new candidate molecules with unknown properties. Thermodynamic properties are commonly predicted in CAMD based on group-contribution (GC) methods (Papadopoulos et al., 2018; Gmehling, 2009). GC methods offer the advantage of a straightforward implementation and computational efficiency (Gani, 2019). However, the chemical design space accessible with these methods is limited to the functional groups a GC method was trained for. Moreover, the prediction of different properties often requires several GC methods. In contrast, the computationally more demanding quantum chemical (QC) (Atkins and Friedman, 2011) methods are not limited to certain functional groups. In conjunction with thermochemistry (Paulechka and Kazakov, 2017; Umer and Leonhard, 2013), QC methods provide consistent predictions of molecular and thermodynamic properties as well as kinetics of chemical reactions (Vereecken et al., 2015). Due to the availability of increased computational capacities, even the use of advanced QC methods in CAMD has become feasible in recent years (Gertig et al., 2020c).The performance of the designed molecules needs to be evaluated during CAMD based on a chosen objective. Commonly, CAMD methods use simple performance indicators as objective function defined based on predicted molecular, thermodynamic or chemical properties (Papadopoulos et al., 2018). However, such indicators may not capture all aspects and trade-offs relevant for the intended use of the designed structures. Thus, CAMD using simple performance indicators likely results in the design of sub-optimal molecules (Adjiman et al., 2014). Preferentially, CAMD should evaluate all candidate molecules directly based on their intended application (Gertig et al., 2020c). In the context of process design, the designed molecules are applied in processes e.g., as working fluids, solvents or catalysts (Papadopoulos et al., 2018). Thus, each candidate molecule should be evaluated using process optimizations to determine the process performance that can be reached using the candidate. The integration of process optimizations into the design procedure corresponds to the extension of CAMD to integrated Computer-Aided Molecular and Process Design (CAMPD) (Papadopoulos et al., 2018).CAMPD methods have already been used extensively for integrated in silico design of molecules and processes, e.g., working fluids and Organic Rankine Cycles (Schilling et al., 2017; 2020; Linke et al., 2015) or extraction solvents and processes (Austin et al., 2017; Papadopoulos and Linke, 2009; Scheffczyk et al., 2018). A very comprehensive review of CAM(P)D applications was recently given by Papadopoulos et al. (2018). For reactive chemical processes, CAMD methods have been used to design reaction solvents that accelerate reaction kinetics (Gertig et al., 2019a; Struebing et al., 2013; 2017; Liu et al., 2019a; 2019b). Moreover, CAMPD methods have been developed for integrated design of reaction solvents and processes (Zhou et al., 2015; Gertig et al., 2020b; Zhang et al., 2020).The main difference of CAMPD methods for non-reactive and for reactive processes lies in the building block property prediction: CAMPD methods for non-reactive processes are commonly based on GC methods for property prediction (Papadopoulos et al., 2018). In contrast, the prediction of reaction kinetics usually requires quantum chemistry. Quantum chemical methods in conjunction with thermochemistry and transition state theory (TST) (Vereecken et al., 2015; Eyring, 1935) have proven to be suited to predict reaction kinetics in CAM(P)D of reaction solvents and processes (Gertig et al., 2020c).CAMPD methods for the integrated design of molecules and non-reactive processes have gained a high level of maturity and several CAM(P)D methods have also been developed for the design of reaction solvents. The use of computational methods has as well gained importance in the search for new catalysts during the past years as shown by several recent review articles. The review by Ahn et al. (2019) introduces organic and metalorganic catalysis as well as strategies for the use of computational methods in the search for new catalysts. Foscato and Jensen (2020) present an elaborated review of computational methods for the development of homogeneous catalysts including large-scale in silico screenings. Freeze et al. (2019) provide an extensive review of catalyst development strategies that especially includes heterogeneous catalysts. These reviews show that computational methods are nowadays used extensively to shed light on catalytic mechanisms, to explain experimental findings or to evaluate new catalysts before going to experiments. Moreover, various systematic approaches for the use of in silico methods in catalyst development have been proposed. Nevertheless, designing molecular catalysts in silico is still regarded as one of the “holy grails in chemistry” (Poree and Schoenebeck, 2017) and only few approaches to automated CAMD of catalyst molecules have been shown so far.For in silico studies of catalysis, some authors have proposed to employ an abstract catalytic environment. In an early approach called “Theozymes”, Tantillo et al. (1998) determine the transition state (TS) of chemical reactions using quantum chemical methods. Subsequently, functional groups are chosen to represent the catalyst and placed around the transition state structure. The spatial positions of these functional groups are optimized in order to stabilize the TS. The stabilization of the TS reduces the activation barrier of the chemical reaction and accelerates reaction kinetics. Thus, the best possible catalytic activity is determined for the chosen functional groups. Recent studies (Hare et al., 2017) are still based on the original work of Tantillo et al. In a recent approach, Dittner and Hartke (2018, 2020) employ an abstract environment that is optimized in order to maximize the catalytic effect of electrostatic interactions. Approaches based on abstract representation of catalysts are well suited to study catalytic effects and provide a theoretical optimum of catalytic activity. However, no real catalyst structures are designed directly.Few approaches have been proposed in literature to design catalyst molecular structures directly. Lin et al. (2005) design transition metal catalysts based on selected functional groups. To optimize the molecular structure of the catalysts, a suitable tabu search algorithm (Chavali et al., 2004) is employed. The properties of designed catalysts are predicted with Quantitative Structure-Property Relationships (QSPR) that are fitted to experimental data. Targets for properties such as density and toxicity are used as design objectives. Consequently, the method directly suggests catalyst structures, but does not optimize these structures based on the achieved catalytic performance. Chu et al. (2012) employ a quantitative structure-activity relationship (QSAR) model in catalyst design. This QSAR model relates the catalytic activity of Ruthenium catalyst complexes for olefin metathesis to descriptors such as bond distances, angles and partial charges (Occhipinti et al., 2006). New catalyst complexes are constructed from a fixed metal core, a list of ligand scaffolds and a variety of molecule fragments that can be attached to the ligand scaffolds. An evolutionary algorithm optimizes the structure of the catalyst. Krausbeck et al. (2017) propose a design method called “Molecular Scaffold Design” based on the idea that unstable, distorted structures occur during chemical reactions and that catalysts need to stabilize such structures to enhance reaction kinetics. The design starts with an unstable fragment with frozen geometry that includes distorted reactants, the core of the catalyst and additional binding sites. A list of atoms that may be added at binding sites is specified. Subsequently, a set of candidates is generated by enumerating the different combinations of atoms that can be added at the binding sites and saturating the resulting structures with hydrogen atoms. These additional hydrogen atoms are replaced by binding sites in the subsequent iteration. The structures are scored with a measure of forces on the nuclei of the unstable fragment computed with quantum mechanical density functional theory (DFT). In an iterative procedure, new “onion shells” (Krausbeck et al., 2017) of atoms are constructed around promising candidates from previous steps until a structure is obtained where the forces on the nuclei in the unstable fragment vanish. More recently, Chang et al. (2018) designed Ni catalyst complexes for a catalytic CO/CO 2 conversion. In their design approach, selected groups of the ligands of the Ni complexes are optimized with the objective to minimize the activation energy of the rate-limiting reaction step. During the design procedure, the activation energies are predicted using the tight binding linear combination of atomic potentials (TB-LCAP) (Xiao et al., 2008) method. Promising candidates from the design are subsequently investigated in more detail using DFT.The design approaches discussed above can be regarded as pioneering work towards in silico design of molecular catalysts. However, two important building blocks of a reliable, direct in silico design of molecular catalyst structures have not been completed, yet. First, a reliable prediction method is needed for the catalytic performance of candidate catalysts, i.e., the acceleration of the reaction kinetics by the designed catalysts. To be reliable, the prediction should employ high-level QC methods already during the design procedure. Second, the ultimate objective of chemical process design is maximum process performance rather than the acceleration of the chemical reactions. Thus, a process-based evaluation of each candidate catalyst is desired. For this purpose, process optimizations have to be integrated into the in silico design of molecular catalysts. In this work, we propose a CAMPD method called CAT-COSMO-CAMPD that integrates the discussed building blocks into the in silico design of molecular catalysts. The prediction of catalytic effects is based on TST and advanced QC methods such as DLPNO-CCSD(T) (Riplinger and Neese, 2013) and COSMO-RS (Klamt et al., 2010). This prediction is broadly applicable and not limited to certain functional groups. Thereby, large and diverse chemical design spaces can be explored. Optimal catalyst structures and process conditions are determined by a hybrid optimization scheme: The genetic optimization algorithm LEA3D (Douguet et al., 2005) generates and optimizes catalyst structures based on a library of 3D molecule fragments. Deterministic process optimizations maximize the performance of processes for each molecular catalyst considered during the design procedure. Thus, the desired process-based evaluation is ensured for all candidate catalysts. Currently, CAT-COSMO-CAMPD is applicable to the integrated design of homogeneous molecular catalysts and chemical processes involving gaseous and liquid phases. Potential extensions to other classes of catalysts are discussed in Section 4.The proposed CAT-COSMO-CAMPD method is explained in detail in Section 2 of this article. Next, the application of CAT-COSMO-CAMPD to the case study of a catalytic carbamate-cleavage process is presented (Section 3). Subsequently, current limitations and future prospects of CAT-COSMO-CAMPD are discussed and conclusions are drawn (Section 4).The integrated catalyst and process design problem is formulated as optimization problem specifying the generic CAMD problem discussed by Gani (2004): (1) max x , y f ( x , Θ , k ) process-based objective s . t . k = h 1 ( x , Θ ) kinetic model Θ = h 2 ( x , y ) thermodynamic property model 0 = h 3 ( x , Θ , k ) process model g 1 ( y ) = 0 chemical feasibility g 2 ( y ) ≤ 0 chemical feasibility c 1 ( Θ ) ≤ 0 constraints on thermodynamic properties c 2 ( y ) ≤ 0 constraints on molecular properties c 3 ( x , Θ , k ) ≤ 0 constraints on the process x ∈ X variable process conditions y ∈ Y molecular structure of catalyst In Problem (1), f ( x , Θ , k ) represents the process-based objective (e.g., conversion or yield) that may depend on the variable process conditions x , on thermodynamic equilibrium properties Θ and on the reaction kinetics determined by reaction rate constants k . The objective is maximized by optimizing the variable process conditions x and the molecular structure of the catalyst molecule y . Rate constants k themselves also depend on x and Θ and are calculated using a kinetic model h 1 ( x , Θ ) . The thermodynamic equilibrium properties Θ depend on the variable process conditions x as well as on the molecular structure of the catalyst molecule y and are calculated using a thermodynamic property model h 2 ( x , y ) . The equations of the process model are represented by h 3 ( x , Θ , k ) . Equality constraints g 1 ( y ) and inequality constraints g 2 ( y ) ensure chemical feasibility of the catalyst molecules, e.g., correct valency of all atoms in the molecule. Additionally, constraints on thermodynamic properties c 1 ( Θ ) (e.g., minimal boiling point of the catalyst molecule), on molecular properties c 2 ( y ) (e.g., number of atoms in the molecule or restrictions on the combination of functional groups) and on the process c 3 ( x , Θ , k ) may be used. The variable process conditions x contained in a range of allowed process conditions X represent the process-related degrees of freedom. Besides quantities like pressures and temperatures, these process-related degrees of freedom may e.g., also include vessel sizes or compositions of mixtures fed to the process. The molecular structure of the catalyst molecule y is contained in a chemical design space Y .It should be noted that the rate constants k = h 1 ( x , Θ ) in Problem (1) do not directly depend on the molecular structure of the catalyst molecule y . However, this missing direct dependence does not mean that catalysts do not impact the rate constants but rather reflects the way rate constants are calculated. Catalyst molecules influence rate constants by reducing so-called activation barriers that reactions need to overcome to take place. We regard these activation barriers as quantities associated with thermodynamic pseudo-equilibria and therefore include them in the thermodynamic equilibrium properties Θ . The calculation of reaction rate constants is explained in more detail in Section 2.1. Subsequently, the prediction of thermodynamic equilibrium properties (Section 2.2) and process modeling (Section 2.3) are described, before the solution approach to the optimization Problem (1) is presented (Section 2.4).The kinetics of catalytic reactions are described by reaction rate constants k that indirectly depend on the structure of the used catalyst molecule y as explained above. The methods we use to predict rate constants were described in detail in earlier work (Gertig et al., 2019, 2021; Kröger et al., 2017). An overview is given in the following without detailed derivations of all equations.The rate constants k of elementary reactions are calculated based on conventional transition state theory (TST) (Vereecken et al., 2015) and the so-called Eyring Equation (Eyring, 1935): (2) k = k B T h ( V m ) ( n − 1 ) exp ( − Δ G ‡ R T ) . In Eq. (2), k B is the Boltzmann constant, T is the temperature at which the reaction takes place, h is Planck’s constant and R is the gas constant. The molar volume of the reaction phase is represented by V m and n denotes the reaction order defined as the sum of the stoichiometric coefficients of all reactants. The activation barrier G ‡ is the difference in molar Gibbs free energy between the state of the reactants and a so-called transition state (TS). According to conventional TST, this transition state is a first-order saddle point in energy that is passed along the reaction path and can be determined using quantum chemical methods (Foresman and Frisch, 2015). The TS is an unstable state in pseudo-equilibrium with the reactant state. 1 1 The term “pseudo-equilibrium” implies that the reactants are not in equilibrium with the products at the same time, although the back-reaction might proceed via the same transition state. In case the rate constant k is calculated for a reaction taking place in a liquid phase, the activation barrier G ‡ is split in different contributions: (3) k = k B T h ( V m i . G . ) ( n − 1 ) exp ( − Δ G ‡ , i . G . R T ) * ∏ i γ i uN γ ‡ uN exp ( − Δ G ˜ ‡ solv − ∑ i Δ G ˜ i solv R T ) = k i . G . ∏ i γ i uN γ ‡ uN exp ( − Δ G ˜ ‡ solv − ∑ i Δ G ˜ i solv R T ) . This splitting into different contributions offers the advantage that the most appropriate methods can be chosen for the calculation of each contribution (Deglmann et al., 2009; Peters et al., 2008; Hellweg and Eckert, 2017; Coote, 2009). In Eq. (3), V m i . G . is the molar volume of the reaction phase in the used ideal gas reference state and may be calculated using the ideal gas law. The activation barrier in the ideal gas reference state is denoted by Δ G ‡ , i . G . . The first 3 terms shown in the upper line of Eq. (3) determine a reaction rate constant k i . G . in the ideal gas reference state: (4) k i . G . = k B T h ( V m i . G . ) ( n − 1 ) exp ( − Δ G ‡ , i . G . R T ) . Solvation effects represent the non-ideal effects that the environment of the reacting species has on the rate constant k . These solvation effects are accounted for by two terms in Eq. (3): First, the ratio of the product of the unsymmetrically normalized activity coefficients γ i uN of all reactants i to the unsymmetrically normalized activity coefficient γ ‡ uN of the transition state. Second, an exponential term containing the different Gibbs free energies of solvation G ˜ solv of the reactants i and the transition state ‡ , respectively. The exponential term accounts for the difference between the ideal gas reference state and a liquid reference state. The term with the unsymmetrically normalized activity coefficients γ uN in turn accounts for the difference between the liquid reference state and the actual reaction mixture composition. This latter term is sometimes neglected, which corresponds to approximating the rate constant k by the rate constant at infinite dilution. The Gibbs free energies of solvation G ˜ solv are computed based on molar reference states. A suitable choice of the ideal gas reference state is the reaction temperature T and a reference pressure of p 0 = 1 bar ≈ 1 atm. Details about the use of reference states as well as a derivation of Eq. (3) can be found in our earlier work (Gertig et al., 2019a; 2020b).To determine all quantities necessary to evaluate Eq. (3), the following computation scheme is applied: 1. Optimized geometries of reactants, catalysts and transition states are determined using the quantum-mechanical density functional theory (DFT) method B3LYP (Stephens et al., 1994) with empirical dispersion correction (Grimme et al., 2010) (B3LYP-D3) and TZVP basis set. Vibrational analysis is carried out subsequently. The B3LYP method is known for good accuracy in geometry optimization and frequency analysis in spite of the rather moderate computational resources required (Zheng et al., 2009; Gottschalk et al., 2018). The rigid rotor harmonic oscillator (RRHO) (Atkins and Friedman, 2011) model is used in the frequency analysis. The software Gaussian 09 (Frisch et al., 2013) is employed for both geometry optimization and frequency analysis. 2. To obtain accurate electronic energies, single point (SP) calculations are performed with the post-Hartree-Fock method DLPNO-CCSD(T) (Riplinger and Neese, 2013; Riplinger et al., 2013) with aug-cc-pVTZ basis set and TightPNO settings. The software ORCA (Neese, 2018) is used for these SP calculations. 3. Activation barriers Δ G ‡ , i . G . in the ideal gas reference state are determined by thermochemical calculations with GoodVibes (Funes-Ardoiz and Paton, 2018) based on the RRHO approximation. As the RRHO approximation can cause significant errors in calculated entropies in case of low frequencies, Grimme’s quasi-harmonic treatment (Grimme, 2012) is employed to reduce these errors. 4. Next, the reaction rate constants k i . G . in the ideal gas reference state are computed using Eq. (4). 5. Optionally, tunneling corrections to k i . G . may be computed in case significant impact of tunneling on the reaction rate is expected. Tunneling corrections based on Eckart (1930) computed with the TAMkin package (Ghysels et al., 2010) have been found to be a reasonable choice. 6. The advanced solvation model COSMO-RS (Klamt et al., 2010; Klamt, 1995; Klamt et al., 1998) is used to calculate Gibbs free energies of solvation G ˜ solv of reactants, catalysts and transition states. The software turbomole (Ahlrichs et al., 1989; TURBOMOLE, 2015) is employed for COSMO (Klamt and Schüürmann, 1993) calculations with BP86 (Becke, 1988; Perdew, 1986a; 1986b) and def2-TZVP basis set. For this purpose, geometries are optimized with BP86/def2-TZVP in vacuum and the actual COSMO calculations are performed as SP calculations. The COSMO-RS calculations (COSMOtherm; Eckert and Klamt, 2002) to obtain the G ˜ solv values are performed subsequently. 7. To calculate the required unsymmetrically normalized activity coefficients γ uN , we either employ COSMO-RS directly or a suitable surrogate model fitted to data from COSMO-RS. 8. Finally, reaction rate constants k in liquid phase are calculated using Eq. (3). Optimized geometries of reactants, catalysts and transition states are determined using the quantum-mechanical density functional theory (DFT) method B3LYP (Stephens et al., 1994) with empirical dispersion correction (Grimme et al., 2010) (B3LYP-D3) and TZVP basis set. Vibrational analysis is carried out subsequently. The B3LYP method is known for good accuracy in geometry optimization and frequency analysis in spite of the rather moderate computational resources required (Zheng et al., 2009; Gottschalk et al., 2018). The rigid rotor harmonic oscillator (RRHO) (Atkins and Friedman, 2011) model is used in the frequency analysis. The software Gaussian 09 (Frisch et al., 2013) is employed for both geometry optimization and frequency analysis.To obtain accurate electronic energies, single point (SP) calculations are performed with the post-Hartree-Fock method DLPNO-CCSD(T) (Riplinger and Neese, 2013; Riplinger et al., 2013) with aug-cc-pVTZ basis set and TightPNO settings. The software ORCA (Neese, 2018) is used for these SP calculations.Activation barriers Δ G ‡ , i . G . in the ideal gas reference state are determined by thermochemical calculations with GoodVibes (Funes-Ardoiz and Paton, 2018) based on the RRHO approximation. As the RRHO approximation can cause significant errors in calculated entropies in case of low frequencies, Grimme’s quasi-harmonic treatment (Grimme, 2012) is employed to reduce these errors.Next, the reaction rate constants k i . G . in the ideal gas reference state are computed using Eq. (4).Optionally, tunneling corrections to k i . G . may be computed in case significant impact of tunneling on the reaction rate is expected. Tunneling corrections based on Eckart (1930) computed with the TAMkin package (Ghysels et al., 2010) have been found to be a reasonable choice.The advanced solvation model COSMO-RS (Klamt et al., 2010; Klamt, 1995; Klamt et al., 1998) is used to calculate Gibbs free energies of solvation G ˜ solv of reactants, catalysts and transition states. The software turbomole (Ahlrichs et al., 1989; TURBOMOLE, 2015) is employed for COSMO (Klamt and Schüürmann, 1993) calculations with BP86 (Becke, 1988; Perdew, 1986a; 1986b) and def2-TZVP basis set. For this purpose, geometries are optimized with BP86/def2-TZVP in vacuum and the actual COSMO calculations are performed as SP calculations. The COSMO-RS calculations (COSMOtherm; Eckert and Klamt, 2002) to obtain the G ˜ solv values are performed subsequently.To calculate the required unsymmetrically normalized activity coefficients γ uN , we either employ COSMO-RS directly or a suitable surrogate model fitted to data from COSMO-RS.Finally, reaction rate constants k in liquid phase are calculated using Eq. (3).As discussed in our previous work (Gertig et al., 2021), it may be important to search for conformers of reacting species and transition states to obtain accurate reaction rate constants k . We perform this conformer search using rotor scans (Foresman and Frisch, 2015). For the integrated catalyst and process design with CAT-COSMO-CAMPD, we assume that it is sufficient to find the conformer with the lowest energy for all species by means of such scans. This assumption is considered a compromise between prediction accuracy and the effort and complexity of the computations performed during the integrated design. The rotor scans to search the most stable conformers are performed for reactants, products and solvents in advance of the actual design. As catalysts and transition states change during the design, selected automated rotor scans are performed as explained in Section 2.4.The expected uncertainty in the rate constants k computed with the methods described above was discussed in detail in previous work (Gertig et al., 2019, 2021; Kröger et al., 2017) and is thus only mentioned briefly here. Generally, we expect that the predicted rate constants k should agree with experimentally determined rate constants k exp within one order of magnitude at a temperature of 25 ∘ C. The uncertainty is expected to decrease with increasing temperature. Furthermore, some errors cancel if the prediction is used for comparison of different candidate catalysts, which is advantageous in CAM(P)D where absolute values are less important than rankings and trends.All thermodynamic equilibrium properties Θ required to solve Problem (1) are computed based on the COSMO-RS (Klamt et al., 2010) model. These quantities typically include Gibbs free energies of solvation G ˜ solv , pure component vapor pressures p i S as well as activity coefficients γ and unsymmetrically normalized activity coefficients γ uN .We compute the Gibbs free energies of solvation G ˜ solv at all temperatures of interest directly using COSMO-RS. This is practical because the G ˜ solv are calculated for defined reference states and thus do not change with changing reaction mixture composition during process simulations and optimizations.Pure component vapor pressures p i S are calculated using the Antoine equation (Pfennig, 2004). The required Antoine parameters for each component are computed with COSMO-RS.In contrast to G ˜ solv and p i S , the activity coefficients γ and unsymmetrically normalized activity coefficients γ uN depend on the reaction mixture composition. Therefore, these quantities have to be re-evaluated frequently in case mixture compositions change. These re-evaluations would consume too much time if COSMO-RS was used directly during process simulations and optimizations. This time-consumption is not only caused by the solution of the COSMO-RS equations themselves, but also by required software interfacing. Thus, the NRTL (Renon and Prausnitz, 1968) activity coefficient model is used as surrogate model. NRTL parameters are automatically fitted to activity coefficient data generated for every system under consideration using COSMO-RS.Process models are formulated based on balance equations and equations accounting for phase equilibria. Moreover, power laws (Levenspiel, 1999) are employed in conjunction with the predicted reaction rate constants k to describe the rates of elementary reactions. The predicted property data allows to formulate a wide range of process models. In the case study presented below, we consider semi-batch operation. In this case, the resulting set of equations is a differential-algebraic system of equations (DAE). The DAE system used in the case study is discussed in the supporting information. The process models are implemented in MATLAB (2019) and solved with the ode15s solver.In the following, the solution approach of the proposed CAT-COSMO-CAMPD method to the integrated catalyst and process design Problem (1) is explained. The solution approach of CAT-COSMO-CAMPD follows our quantum chemistry-based design methods for solvents (Scheffczyk et al., 2018; Gertig et al., 2019a; 2020b; Scheffczyk et al., 2017; Fleitmann et al., 2018). A preliminary and brief presentation of CAT-COSMO-CAMPD was already given at the conference corresponding to this special issue (Gertig et al., 2020a). CAT-COSMO-CAMPD employs a hybrid optimization scheme. The genetic optimization algorithm LEA3D (Douguet et al., 2005) is used to identify the structure y * of the optimal catalyst molecule, whereas gradient-based process optimization is used to determine optimal values of the variable process conditions x * for each considered catalyst.The LEA3D algorithm generates 3D molecular structures based on a pre-defined library of 3D molecule fragments. Fragments are combined randomly in the initial step of the optimization to obtain a first generation of candidate molecules. Based on the performance of these initial candidates, LEA3D uses genetic operations to alter the structures and to suggest the next generation of candidates. New generations are iteratively suggested until a specified maximum number of generations is reached. By this procedure, the space of candidate molecules is systematically explored to determine the optimal structure y * . Other convergence criteria suitable for genetic algorithms could be employed (Safe et al., 2004). Still, the used genetic algorithm is stochastic such that convergence to the global optimum cannot be guaranteed.Working with 3D molecular structures in the optimization of catalyst molecules offers the major advantage that quantum chemistry-based property prediction can be employed in a straightforward way: QC methods generally require 3D starting geometries as input. QC-based property prediction offers several advantages (see Section 1: Quantum chemical methods and thermochemistry consistently predict a broad range of molecular, thermodynamic and chemical properties including reaction rate constants. In contrast to e.g., group-contribution methods, QC methods are not limited to previously fitted functional groups. Thus, a wide variety of 3D molecule fragments may be chosen when setting up the chemical design space.Currently, CAT-COSMO-CAMPD requires the user to specify a so-called scaffold fragment among the other 3D fragments. This scaffold fragment contains the reactants as well as the catalytically active group contained in all catalysts designed in one CAT-COSMO-CAMPD run. Thus, during one design run, the mechanism of catalysis does not change. The scaffold fragment is employed to construct starting geometries for the search of transition states of the catalytic reactions with the designed catalysts. Further information about scaffold fragments is given in the subsequent description of the design procedure and an example is discussed in Section 3.1.The complete CAT-COSMO-CAMPD procedure to solve the integrated design Problem (1) is shown in the flowchart in Fig. 1 and explained in the following: 1. First, several specifications have to be made: • The reaction network and the process under consideration have to be specified. • The catalytically active group is chosen. The designed catalysts are based on this group. 2. Quantum chemical and thermochemical calculations are performed for reactants and products in the specified reaction network as well as for solvents used in the process. These QC and thermochemical calculations are discussed in Section 2.1 (Steps 1–3 and 6). Moreover, a so-called reference system is defined (Fig. 1) that corresponds to a reaction system including a typical molecule with the catalytically active group as catalyst. QC calculations are employed to determine the transition state geometry for the reference system. From this TS geometry, the scaffold fragment is constructed (for an example, see Fig. 4). As already described above, the scaffold fragment contains the reactants and the catalytically active group of the catalyst. All atoms of the reactants and the catalytically active group are positioned in space as in the determined TS geometry of the reference system. The scaffold fragment is used in Step 3 when the chemical design space Y is defined. 3. Design specifications have to be made for the integrated catalyst and process design: • A fragment library is provided that contains various 3D molecule fragments including the catalyst scaffold fragment. From these 3D fragments, catalyst structures are constructed. The choice of fragments in the library determines the chemical design space Y for CAMD of catalysts. It is important to note that the resulting design space does not correspond to a set of possible catalysts, but to a set of possible transition states of the catalytic reaction under consideration. The reason is that the scaffold fragment contains not only the catalytically active group of the catalyst, but also the reactants. Consequently, the transition states are designed directly instead of the catalyst molecules. The direct design of the transition states is advantageous because starting geometries for the optimization of the transition state structures have to be provided. Obtaining suitable starting geometries of transition states can be considered the critical aspect of the automated prediction of rate constants k of catalytic reactions. • The user may choose rotor scans and additional pre-optimizations performed in Step 5. • Settings are required for the genetic algorithm LEA3D used for the optimization of catalyst molecular structures. These settings include the maximum number of generations, the number of candidate catalyst molecules per generation as well as the probabilities for genetic operations such as mutation and cross-over of candidates. • The process model h 3 ( x , Θ , k ) is specified. • The process-based objective f ( x , Θ , k ) of the design is chosen. • The values of constant process parameters are assigned and process degrees of freedom x are selected. • The allowed range of operating conditions X is defined. • Constraints are set including constraints c 1 ( Θ ) on thermodynamic properties, c 2 ( y ) on molecular properties and c 3 ( x , Θ , k ) on the process. LEA3D inherently respects the constraints g 1 ( y ) and g 2 ( y ) to ensure chemical feasibility of designed molecules. Moreover, LEA3D ensures that each candidate catalyst molecule contains one scaffold fragment and thus one catalytically active group. 4. LEA3D suggests a generation of 3D molecular structures of transition states with candidate catalysts. The initial generation is created by random combination of the scaffold fragment with further molecule fragments from the fragment library. Candidates of subsequent generations are obtained from genetic operations such as mutation and cross-over that are applied to promising candidates of the preceding generation. The generated structures are passed to the automated property prediction and gradient-based process optimization. 5. Suitable starting geometries for the optimization of catalysts and transition states are determined. It is important to note that by the term “geometry” of a molecule or TS, we here understand the set of spatial positions of all atoms comprising the molecule or TS. In contrast, “molecular structures” include the connectivity of atoms as shown in structural formulas of molecules. In principle, 3D geometries of transition states are already provided by LEA3D in Step 4. However, it happens occasionally that these structures are not accurate enough for the subsequently used QC methods to work properly. Thus, a first pre-optimization improves the TS geometries employing the force field method MMFF94 (Halgren, 1996a; 1996b; 1996c) available in the chemical toolbox Open Babel (O’Boyle et al., 2011; Open Babel). As this first pre-optimization is not suited to handle transition states, the scaffold fragment is substituted by a dummy atom. Afterwards, the dummy is re-substituted by the scaffold fragment using translation and rotation operations in 3D cartesian coordinate space. Next, further pre-optimization is performed using Gaussian 09. Selected rotor scans are performed using the QC method AM1 (Dewar et al., 1985) to ensure that the minimum energy conformer of each TS is identified. The selection of rotor scans to perform is made by the user (Step 3). Optionally, a geometry optimization with the DFT method B3LYP that minimizes energy may follow the rotor scans. During rotor scans and energy minimization with Gaussian 09, it has to be ensured that no atomic distances are changed that correspond to bonds that break or form during the considered reaction. For this purpose, the “modredundant” option (Foresman and Frisch, 2015) is employed. The TS geometries obtained from the described pre-optimizations are used as starting geometries for subsequent QC calculations. Starting geometries of the catalyst molecules are extracted as a subset of the TS starting geometries. 6. To predict reaction kinetics of the catalytic reactions, reaction rate constants k are computed as explained in Section 2.1. Calculations for reactants, products and solvents performed in Step 2 are not repeated here. 7. Required thermodynamic equilibrium properties Θ are predicted as described in Section 2.2. 8. Gradient-based process optimizations are performed for processes with each candidate catalyst. For these processes, the process optimizations determine the optimal process performance f * ( x * , Θ , k ) as well as the corresponding optimal values of variable process conditions x * . The interior point algorithm available in the built-in function fmincon in MATLAB is employed for process optimizations. The evaluations of the process model h 3 ( x , Θ , k ) required during process optimization are performed with the solver ode15s as already mentioned in Section 2.3. First, several specifications have to be made: • The reaction network and the process under consideration have to be specified. • The catalytically active group is chosen. The designed catalysts are based on this group. The reaction network and the process under consideration have to be specified.The catalytically active group is chosen. The designed catalysts are based on this group.Quantum chemical and thermochemical calculations are performed for reactants and products in the specified reaction network as well as for solvents used in the process. These QC and thermochemical calculations are discussed in Section 2.1 (Steps 1–3 and 6). Moreover, a so-called reference system is defined (Fig. 1) that corresponds to a reaction system including a typical molecule with the catalytically active group as catalyst. QC calculations are employed to determine the transition state geometry for the reference system. From this TS geometry, the scaffold fragment is constructed (for an example, see Fig. 4). As already described above, the scaffold fragment contains the reactants and the catalytically active group of the catalyst. All atoms of the reactants and the catalytically active group are positioned in space as in the determined TS geometry of the reference system. The scaffold fragment is used in Step 3 when the chemical design space Y is defined.Design specifications have to be made for the integrated catalyst and process design: • A fragment library is provided that contains various 3D molecule fragments including the catalyst scaffold fragment. From these 3D fragments, catalyst structures are constructed. The choice of fragments in the library determines the chemical design space Y for CAMD of catalysts. It is important to note that the resulting design space does not correspond to a set of possible catalysts, but to a set of possible transition states of the catalytic reaction under consideration. The reason is that the scaffold fragment contains not only the catalytically active group of the catalyst, but also the reactants. Consequently, the transition states are designed directly instead of the catalyst molecules. The direct design of the transition states is advantageous because starting geometries for the optimization of the transition state structures have to be provided. Obtaining suitable starting geometries of transition states can be considered the critical aspect of the automated prediction of rate constants k of catalytic reactions. • The user may choose rotor scans and additional pre-optimizations performed in Step 5. • Settings are required for the genetic algorithm LEA3D used for the optimization of catalyst molecular structures. These settings include the maximum number of generations, the number of candidate catalyst molecules per generation as well as the probabilities for genetic operations such as mutation and cross-over of candidates. • The process model h 3 ( x , Θ , k ) is specified. • The process-based objective f ( x , Θ , k ) of the design is chosen. • The values of constant process parameters are assigned and process degrees of freedom x are selected. • The allowed range of operating conditions X is defined. • Constraints are set including constraints c 1 ( Θ ) on thermodynamic properties, c 2 ( y ) on molecular properties and c 3 ( x , Θ , k ) on the process. A fragment library is provided that contains various 3D molecule fragments including the catalyst scaffold fragment. From these 3D fragments, catalyst structures are constructed. The choice of fragments in the library determines the chemical design space Y for CAMD of catalysts. It is important to note that the resulting design space does not correspond to a set of possible catalysts, but to a set of possible transition states of the catalytic reaction under consideration. The reason is that the scaffold fragment contains not only the catalytically active group of the catalyst, but also the reactants. Consequently, the transition states are designed directly instead of the catalyst molecules. The direct design of the transition states is advantageous because starting geometries for the optimization of the transition state structures have to be provided. Obtaining suitable starting geometries of transition states can be considered the critical aspect of the automated prediction of rate constants k of catalytic reactions.The user may choose rotor scans and additional pre-optimizations performed in Step 5.Settings are required for the genetic algorithm LEA3D used for the optimization of catalyst molecular structures. These settings include the maximum number of generations, the number of candidate catalyst molecules per generation as well as the probabilities for genetic operations such as mutation and cross-over of candidates.The process model h 3 ( x , Θ , k ) is specified.The process-based objective f ( x , Θ , k ) of the design is chosen.The values of constant process parameters are assigned and process degrees of freedom x are selected.The allowed range of operating conditions X is defined.Constraints are set including constraints c 1 ( Θ ) on thermodynamic properties, c 2 ( y ) on molecular properties and c 3 ( x , Θ , k ) on the process.LEA3D inherently respects the constraints g 1 ( y ) and g 2 ( y ) to ensure chemical feasibility of designed molecules. Moreover, LEA3D ensures that each candidate catalyst molecule contains one scaffold fragment and thus one catalytically active group.LEA3D suggests a generation of 3D molecular structures of transition states with candidate catalysts. The initial generation is created by random combination of the scaffold fragment with further molecule fragments from the fragment library. Candidates of subsequent generations are obtained from genetic operations such as mutation and cross-over that are applied to promising candidates of the preceding generation. The generated structures are passed to the automated property prediction and gradient-based process optimization.Suitable starting geometries for the optimization of catalysts and transition states are determined. It is important to note that by the term “geometry” of a molecule or TS, we here understand the set of spatial positions of all atoms comprising the molecule or TS. In contrast, “molecular structures” include the connectivity of atoms as shown in structural formulas of molecules. In principle, 3D geometries of transition states are already provided by LEA3D in Step 4. However, it happens occasionally that these structures are not accurate enough for the subsequently used QC methods to work properly. Thus, a first pre-optimization improves the TS geometries employing the force field method MMFF94 (Halgren, 1996a; 1996b; 1996c) available in the chemical toolbox Open Babel (O’Boyle et al., 2011; Open Babel). As this first pre-optimization is not suited to handle transition states, the scaffold fragment is substituted by a dummy atom. Afterwards, the dummy is re-substituted by the scaffold fragment using translation and rotation operations in 3D cartesian coordinate space. Next, further pre-optimization is performed using Gaussian 09. Selected rotor scans are performed using the QC method AM1 (Dewar et al., 1985) to ensure that the minimum energy conformer of each TS is identified. The selection of rotor scans to perform is made by the user (Step 3). Optionally, a geometry optimization with the DFT method B3LYP that minimizes energy may follow the rotor scans. During rotor scans and energy minimization with Gaussian 09, it has to be ensured that no atomic distances are changed that correspond to bonds that break or form during the considered reaction. For this purpose, the “modredundant” option (Foresman and Frisch, 2015) is employed. The TS geometries obtained from the described pre-optimizations are used as starting geometries for subsequent QC calculations. Starting geometries of the catalyst molecules are extracted as a subset of the TS starting geometries.To predict reaction kinetics of the catalytic reactions, reaction rate constants k are computed as explained in Section 2.1. Calculations for reactants, products and solvents performed in Step 2 are not repeated here.Required thermodynamic equilibrium properties Θ are predicted as described in Section 2.2.Gradient-based process optimizations are performed for processes with each candidate catalyst. For these processes, the process optimizations determine the optimal process performance f * ( x * , Θ , k ) as well as the corresponding optimal values of variable process conditions x * . The interior point algorithm available in the built-in function fmincon in MATLAB is employed for process optimizations. The evaluations of the process model h 3 ( x , Θ , k ) required during process optimization are performed with the solver ode15s as already mentioned in Section 2.3.As the quantum chemical calculations performed in Steps 6 and 7 may be computationally demanding, the output files of these calculations are stored in dedicated QC file databases. If structures are suggested in Step 4 that were already considered previously, the demanding QC calculations can be skipped. It is important to note here that the prediction of reaction kinetics (Step 6) already requires some thermodynamic equilibrium properties Θ . Moreover, certain equilibrium properties need to be re-evaluated when conditions such as reaction mixture composition change during dynamic processes or when variable process conditions x are changed during process optimizations (Step 8). Thus, Steps 6 to 8 are not performed strictly in the sequence displayed in Section 2.4. 9. The candidate catalysts of the current generation are ranked based on the determined values of the optimal process performance f * ( x * , Θ , k ) . 10. A new generation of candidate structures is generated by LEA3D based on the previous generation using genetic operations (Step 4). The probabilities that certain candidates of the previous generation are selected as parents for candidates of the new generation are related to the process performance f * ( x * , Θ , k ) . Steps 4 to 10 are repeated until the maximum number of generations specified in Step 3 is reached. 11. As output of the integrated catalyst and process design with CAT-COSMO-CAMPD, a ranked list of catalyst structures y including the corresponding values of f * ( x * , Θ , k ) and x * is assembled. The candidate catalysts of the current generation are ranked based on the determined values of the optimal process performance f * ( x * , Θ , k ) .A new generation of candidate structures is generated by LEA3D based on the previous generation using genetic operations (Step 4). The probabilities that certain candidates of the previous generation are selected as parents for candidates of the new generation are related to the process performance f * ( x * , Θ , k ) . Steps 4 to 10 are repeated until the maximum number of generations specified in Step 3 is reached.As output of the integrated catalyst and process design with CAT-COSMO-CAMPD, a ranked list of catalyst structures y including the corresponding values of f * ( x * , Θ , k ) and x * is assembled.Obtaining a ranked list instead of a single optimal catalyst as output of the design offers an important advantage: The user may choose candidates for experimental testing among several near-optimal candidates from the design. Thus, further criteria such as ease of synthesis, commercial availability or toxicity that were not considered in the design may be taken into account in the final choice.To demonstrate the application of CAT-COSMO-CAMPD, an integrated catalyst and process design is performed for a catalytic carbamate-cleavage process (Wang et al., 2017) of methyl phenyl carbamate (MPC) to phenyl isocyanate and methanol (Fig. 2 ). Carbamate-cleavage reactions represent challenging steps in possible production routes to industrially important isocyanates (Six and Richter, 2003). One such production route that aims at CO 2 -based isocyanate production (Kaiser et al., 2018) has been investigated in the research project (Carbon2Chem). The design of carbamate-cleavage processes is challenging because the cleavage reactions are strongly endothermic. Moreover, reaction equilibria strongly favor carbamate formation and are thus very unfavorable for carbamate-cleavage processes (Leitner et al., 2018). Typically, reaction temperatures of T R > 200 ∘ C are required to drive the reaction. To avoid fast back-reactions, continuous removal of the alcohol produced as by-product is required during carbamate-cleavage. In case of volatile alcohols such as methanol, this continuous removal can be ensured using stripping with inert nitrogen gas (Cao et al., 2015). A suitable process flowsheet already introduced in previous studies (Gertig et al., 2020b; 2020a; 2019b) is shown in Fig. 3 . A semi-batch reactor is employed to carry out the catalytic cleavage reaction that takes place in the liquid phase using di-phenyl ether as solvent. Nitrogen is used for stripping to carry the formed volatile methanol out of the reactor. A flash is used to condense and recycle unintentionally removed isocyanate, carbamate, solvent and catalyst. Ideally, only nitrogen and methanol leave the semi-batch process with the top product stream of the flash. A process model h 3 ( x , Θ , k ) for this process was developed as explained in Section 2.3 and is discussed in more detail in the supporting information. Required rate constants k of the catalytic cleavage reaction as well as thermodynamic equilibrium properties are computed as described in Sections 2.1 and 2.2. It was shown in previous work (Gertig et al., 2021) that the employed methods are suited to carbamate-cleavage reactions. Due to the challenging nature of carbamate-cleavage, it is expected to be difficult to obtain satisfying isocyanate yields in the considered semi-batch cleavage process, even when using a catalyst. Thus, the isocyanate yield is a good choice for the objective of the integrated catalyst and carbamate-cleavage process design. The specifications of the according design with CAT-COSMO-CAMPD are given in the following section.The reaction and process under consideration are specified as described above. As catalytically active group, the carboxyl group is chosen that is known to have catalytic properties (Satchell and Satchell, 1975). The constructed scaffold fragment is shown in Fig. 4 . As can be seen, the scaffold fragment contains the reactant MPC in a partially cleaved state as well as the carboxyl group of the catalyst that is designed. The carboxyl group acts as both proton acceptor and proton donor in a concerted reaction: The proton originally bound to the nitrogen of the carbamate (atom 2 in Fig. 4) is accepted and simultaneously, the proton initially contained in the carboxyl group (atom 6 in Fig. 4) is donated to the methoxy group of the carbamate to form the by-product methanol. The violet dot marks the anchor point where the scaffold fragment is connected to other 3D molecule fragments in the catalyst design.No significant influence of tunneling is expected for the catalytic carbamate-cleavage. Thus, no tunneling corrections are computed. One rotor scan is performed for each candidate around the bond between the carboxyl group and the rest of the catalyst (atom 4 and violet dot in Fig. 4; see step 5 in Section 2.4).The yield used as objective f ( x , Θ , k ) of the integrated design is defined as the final moles of isocyanate present in the reactor divided by the initial moles of carbamate provided. This objective is maximized solving the integrated design Problem (1) in order to identify the optimal catalyst structure y * and corresponding optimal values of variable process conditions x * . The temperature T F in the flash and the volume flow V ˙ N 2 of nitrogen fed to the reactor are chosen as variable process conditions for process optimizations. The reaction temperature T R in the reactor is fixed to 473.15 K. This reaction temperature lies in the typical range of reaction conditions for carbamate-cleavage (Gertig et al., 2021; Wang et al., 2017). The reaction temperature is lower compared to our previous study of auto-catalytic carbamate-cleavage (Gertig et al., 2020b) to take into account that the designed catalysts accelerate the cleavage reaction. Higher temperatures are expected to accelerate the cleavage reaction, but may also lead to undesired side reactions. The pressure of the process is set to p set = 4 bar . The semi-batch process is allowed to run for 12 h. A reactor volume of V R = 1 m 3 as well as initial fractions of 15 mass-% MPC and 5 mass-% catalyst in the reaction mixture are chosen.The constraints g 1 ( y ) and g 2 ( y ) (see Problem (1)) to ensure chemical feasibility of designed catalyst molecules are inherently respected by the LEA3D algorithm as already mentioned in Section 2.4. Moreover, constraints c 2 ( y ) ensure that each designed catalyst contains exactly one scaffold fragment and limit the number of non-hydrogen atoms in the designed catalysts to a maximum of 13. The operating range X is set to allow flash temperatures of 280 K < T F < 380 K as well as nitrogen volume flows of 5 × 10 − 5 m 3 s − 1 < V ˙ N 2 < 1.5 × 10 − 1 m 3 s − 1 . The temperature range used for T F likely allows for the use of cooling water. The optimal nitrogen volume flow V ˙ N 2 * is sought between values close to zero and an upper bound that is expected to lie well above favorable values. To define the design space Y for catalyst design, various 3D alkyl, aryl, ether, ester, keto, nitrile, halide, sulfene, sulfide, and imine fragments are provided. The full list of fragments is given in the supporting information. After the initial generation of candidate catalyst molecules designed by random combination of fragments, the integrated catalyst and carbamate-cleavage process design is run for 6 further generations of candidates using a number of 12 candidates per generation.The integrated catalyst and carbamate-cleavage process design with CAT-COSMO-CAMPD results in the optimized catalyst molecular structure y * with SMILES (Weininger, 1988) code ClCOC(C(=O)O)C1CCCCC1. The corresponding 2D structural formula is shown in Fig. 5 . The achieved objective function value f * ( x * , Θ , k ) amounts to a yield of 21%. The corresponding optimal values x * of the variable process conditions are a flash temperature of T F = 281.1 K and a nitrogen volume flow of V ˙ N 2 = 5.4 × 10 − 2 m 3 s − 1 . The top catalyst molecule enables a considerably higher process performance compared to common carboxylic acids such as acetic acid (9% yield, also displayed in Fig. 5). In total, 33 candidate catalysts from the design shown in Fig. 5 meet all constraints. The full list of SMILES codes of these catalyst molecules can be found in the supporting information.The obtained results show that CAT-COSMO-CAMPD successfully identifies molecular structures of catalysts that considerably improve the predicted process performance compared to common reference molecules. However, it is not clear at this point whether the whole integrated catalyst and process design was required to achieve this result or whether it would be sufficient to perform a less complex catalyst design optimizing the rate constant k of the catalytic cleavage reaction. To shed light on this question, a catalyst design that is not integrated with process optimization is presented in the following.Performing a less complex catalyst design without integrated process optimization corresponds to reducing the Computer-Aided Molecular and Process Design (CAMPD) to a Computer-Aided Molecular Design (CAMD). Thus, Problem (1) (see Section 2) simplifies: The objective function f ( Θ , k ) does not depend on any variable process conditions x and no process model h 3 is required any more. Consequently, also the possibility to set process constraints c 3 as well as an operation range X are removed. Except any specifications that are not relevant for the resulting CAMD problem, the catalyst design without integrated process optimization is specified as the integrated design (Section 3.1).The catalyst design to maximize the rate constant k of the catalytic carbamate-cleavage reaction results in brominated formic acid with SMILES code OC(=O)Br as optimal catalyst molecule. The 2D structural formula is shown in Fig. 6 . The corresponding predicted rate constant at 200 ∘ C amounts to k = 4.43 × 10 − 6  m 6 mol − 2 s − 1 . In total, 26 catalysts from the design meet all constraints (Fig. 6). While we suspect brominated formic acid to be unstable at reaction conditions, the results of the design with k as objective demonstrate the problems associated with simple performance measures: Using a catalyst that enables a high rate constant k may still lead to a poor process performance. Brominated formic acid as catalyst leads to the highest predicted rate constant k , but the process using this catalyst achieves only a relatively low yield of 13% after 12 h (see Fig. 5). The underlying reason for the poor performance of simple design objectives is that besides a high reaction rate constant, further aspects are important to reach maximum process performance. For example, the volatility of the catalyst influences catalyst loss due to stripping with nitrogen. Highly polar catalysts increase the overall polarity of the reaction mixture, which has adverse effects on the reaction rate depending on the amount of catalyst used. Moreover, the catalyst molecule influences activities of other substances in the mixture and thus the vapor-liquid equilibria (VLE) in the process. Consequently, for example, the catalyst impacts the activity coefficient of methanol in the flash and can lead to unfavorable methanol recycling. In summary, several criteria need to be accounted for and trade-offs between these criteria have to be made in order to reach optimum process performance. Therefore, an integrated catalyst and process design is required that combines molecular design with process optimization. Still, evaluation criteria used to identify promising catalyst molecules might sometimes even go less far as to calculate reaction rate constants. For many reactions, simpler, heuristic criteria could be thought of. In the subsequent section, we discuss why such criteria may not only fail to identify catalysts enabling optimal process performance, but even fail to find catalysts that enable optimal rate constants.In many cases, heuristic criteria for catalyst design can be defined that allow assessing candidates with less effort than computing reaction kinetics or using process-based evaluation such as our design method CAT-COSMO-CAMPD. For the catalytic carbamate-cleavage reaction considered in the present case study, we discuss three possible examples of heuristic evaluation criteria: • The barrier in electronic energy Δ E el computed in vacuum often represents the largest contribution to the overall activation barrier Δ G ‡ in Gibbs free energy that the reaction has to overcome. Thus, a heuristic design objective for catalyst design could be to minimize Δ E el of the reaction. • The catalyst and the carbamate form a ring for concerted proton transport in the transition state of the catalytic carbamate-cleavage reaction (see Fig. 4, atoms 1–8). Thus, it could be supposed that the proton donor and/or acceptor properties of the carboxyl group determine catalytic activity. These donor/acceptor properties are related e.g., to the partial charges of the oxygen atoms of the carboxyl group of the catalyst. Therefore, the partial charges should be either strong or weak, depending on whether proton acceptance or donation is critical, or a certain value represents the optimal trade-off. Thus, another objective for catalyst design could be to maximize or minimize partial charges of the carboxyl oxygen atoms or to match a certain target value. • There are assumptions mentioned in literature (Satchell and Satchell, 1975) that a nucleophilic character of the oxygen atom of the reactant’s methoxy group stabilizes the transition state of the reaction. Thus, it is supposed that an important function of the catalyst is to increase this nucleophilic character. A high nucleophilicity should be associated with a high negative partial charge of the oxygen atom and should be observable in the transition state of the catalytic reaction. The barrier in electronic energy Δ E el computed in vacuum often represents the largest contribution to the overall activation barrier Δ G ‡ in Gibbs free energy that the reaction has to overcome. Thus, a heuristic design objective for catalyst design could be to minimize Δ E el of the reaction.The catalyst and the carbamate form a ring for concerted proton transport in the transition state of the catalytic carbamate-cleavage reaction (see Fig. 4, atoms 1–8). Thus, it could be supposed that the proton donor and/or acceptor properties of the carboxyl group determine catalytic activity. These donor/acceptor properties are related e.g., to the partial charges of the oxygen atoms of the carboxyl group of the catalyst. Therefore, the partial charges should be either strong or weak, depending on whether proton acceptance or donation is critical, or a certain value represents the optimal trade-off. Thus, another objective for catalyst design could be to maximize or minimize partial charges of the carboxyl oxygen atoms or to match a certain target value.There are assumptions mentioned in literature (Satchell and Satchell, 1975) that a nucleophilic character of the oxygen atom of the reactant’s methoxy group stabilizes the transition state of the reaction. Thus, it is supposed that an important function of the catalyst is to increase this nucleophilic character. A high nucleophilicity should be associated with a high negative partial charge of the oxygen atom and should be observable in the transition state of the catalytic reaction.The 3 criteria discussed above were evaluated for the catalysts designed in the design run presented in Section 3.4. All required quantities were extracted from the output of the DLPNO-CCSD(T) calculations (see Step 2 of the computation scheme explained in Section 2.1). Mulliken charges (Mulliken, 1955) are used to approximate the partial charges of atoms. Fig. 7 plots the criteria versus the logarithm of the reaction rate constant k achieved with the respective catalysts. For reasons discussed in the preceding sections, it cannot be expected that the heuristic criteria reflect the achievable process performance. Still, one could argue that using one of the heuristic criteria should at least result in the design of catalysts that optimize the reaction rate constant k of the catalytic carbamate-cleavage. However, if this was the case, the chosen criterion should correlate well with k .As can be seen in the upper part of Fig. 7, there is a general trend of increasing rate constants k with decreasing barriers in electronic energy Δ E el as expected. However, the correlation of k with Δ E el is clearly not good enough for Δ E el to serve as design objective. In particular, the catalyst leading to the lowest Δ E el enables only a moderate rate constant k . The reason for the insufficient correlation is that important effects on k are not reflected by Δ E el . Although representing a major contribution to the activation barrier Δ G ‡ that in turn determines the rate constant, evaluation of catalysts based on Δ E el completely neglects entropic and solvation effects.The partial charges of the oxygen atoms in the carboxyl group of the catalyst are shown in the middle part of Fig. 7. The term “donor O-atom” hereby refers to the oxygen atom initially bonded to a hydrogen atom. The correlation of these partial charges with k is not satisfactory. Indeed, there is a trend that the absolute values of the charges of the oxygen atoms decrease with increasing rate constant k , which could be related to proton donor and acceptor properties of the catalysts. In fact, most of the highly ranked catalysts contain electron-withdrawing groups like halogens or nitrile groups, which explains the small absolute charge of the oxygen atoms in the carboxyl group. However, it can be seen in Fig. 7 that the trend is only weak and there is strong scattering.The third suggested heuristic criterion, the nucleophilic character of the oxygen atom of the methoxy group, is evaluated based on the Mulliken charge of the O-atom in the TS. Interestingly, some of the catalysts that enable very high rate constants strongly increase the absolute charge of the O-atom, relating to an increased nucleophilicity. However, no general trend is observed over the whole range of k values. Thus, the third heuristic criterion is also no reliable design objective.The above discussion shows why choosing heuristic criteria as design objectives is not reliable: Important effects are likely neglected by such criteria. The correlation of heuristic criteria with the ultimate goal of a catalyst design may be weak and there may be strong scattering.In summary, two important conclusions are drawn from the presented results: First, a full computation of reaction kinetics is required in catalyst design. Heuristic criteria do not guarantee finding the catalyst structures that enable the largest catalytic effects. Second, various other effects besides accelerating reaction kinetics are important to achieve optimal process performance. Consequently, a process-based evaluation of candidate catalysts as used by CAT-COSMO-CAMPD is strongly recommended.Here, we present CAT-COSMO-CAMPD as a method for integrated in silico design of catalysts and processes. The integrated design is formulated as optimization problem and a hybrid optimization scheme is employed to solve the design problem: The genetic optimization algorithm LEA3D is used to explore the chemical design space and to identify the most promising candidate catalyst molecules. Process conditions are optimized in gradient-based process optimizations. The LEA3D algorithm works with 3D structures and designs molecules based on a library of 3D molecule fragments. Thus, 3D structures of all considered molecules are available throughout the design and the direct use of quantum chemical methods for property prediction is facilitated. Consequently, CAT-COSMO-CAMPD avoids simplified property prediction methods and the need for extensive experimental data. Reaction rate constants are predicted based on transition state theory and thermodynamic properties of mixtures are computed using the advanced solvation model COSMO-RS.The application of CAT-COSMO-CAMPD for integrated catalyst and process design is demonstrated for a catalytic carbamate-cleavage process. The results show that promising catalyst molecules are identified and processes are optimized successfully. Moreover, we show that the integration of molecular design with process optimization is required to achieve optimal process performance. In contrast, selection of catalysts based on predicted reaction rate constants or heuristic criteria likely fails to find the optimal catalyst structures.So far, we have applied CAT-COSMO-CAMPD only to organic molecular catalysts. More sophisticated homogeneous catalysts such as catalyst complexes may be designed with the same approach, although the QC-based prediction of the catalytic activity of the complexes will be more demanding. First, larger system sizes will either require more computer power or using less sophisticated QC methods. Second, converging geometry optimizations during automated design may become more challenging for larger catalyst structures. Third, different possible spin multiplicities will have to be considered for some catalyst complexes.The design of heterogeneous catalysts is not within the current scope of our method. For such catalysts, dedicated approaches are required to obtain candidate catalysts and to predict catalytic activity. For a review of such approaches, the reader is referred to Freeze et al. (2019). However, the integration of catalyst design with process optimization using a hybrid optimization scheme as introduced with CAT-COSMO-CAMPD may also be applied in the design of heterogeneous catalysts.Possible extensions to CAT-COSMO-CAMPD also include more sophisticated process simulation and optimization. For the demonstration of CAT-COSMO-CAMPD in the case study presented in the previous section, a process model was implemented that contains the 2-phase reactor and a flash. Considering additional separation steps or even the complete downstream processing is straightforward in the sense that the additional steps may be implemented in the same process model (Scheffczyk et al., 2018). Of course, the process optimization is expected to become increasingly challenging with an increasing number of degrees of freedom.In principle, it is also possible to call an external software for rigorous process simulation. In this case, the user could benefit from available libraries with detailed unit operation models and dedicated solution algorithms. In turn, the use of external programs requires suitable interfacing and thus needs to consider how data is transferred to and from the process simulation software. In addition, it must be ensured that no convergence problems occur that would require manual overrides and thus hinder the automation of the design. However, promising solutions for this approach have been presented (Scheffczyk et al., 2018; López et al., 2018; Navarro-Amorós et al., 2014). As an alternative to integrating external software into the design, sophisticated process simulations and optimizations could be used to examine promising candidates after the actual integrated design.Current limitations of CAT-COSMO-CAMPD in the design of homogeneous catalysts are the accuracy and computational requirements of QC methods used as well as the need to know the mechanism of catalysis and the active group in advance. However, in silico design methods such as CAT-COSMO-CAMPD will benefit from the growth of computational capacity and the ongoing development of efficient but accurate QC methods. Thus, we expect that the design of increasingly complex systems will become possible. There are also promising developments in the field of automated generation of reaction networks (Döntgen et al., 2015; 2018; Simm et al., 2018; Dewyer et al., 2018) that may serve to overcome the need of CAT-COSMO-CAMPD to know the mechanism of catalysis in advance. Moreover, such an automated determination of reaction mechanisms could ease the consideration of multiple catalytically active groups and undesired side reactions. At present, none of the available methods seems to be accurate and efficient enough for a broad range of systems. However, improving such methods is subject of ongoing research.Surrogate models (e.g., neural networks) may complement QC methods in property prediction during design. For the design of reaction solvents, a hybrid approach using both QC-based prediction of reaction rates and a surrogate model was already introduced by Struebing et al. (2013, 2017). In their work, only part of the candidate molecules are investigated using the full QC-based treatment. Most candidates are evaluated with a surrogate model that is constantly refined during the design procedure using the results from QC. However, one should be aware that the advantages of surrogates regarding computation time and computational capacity usually come with the disadvantage of a lower accuracy compared to the original model. Still, depending on the computational requirements and the number of candidates that need to be evaluated, also integrated catalyst and process design might benefit from such a hybrid approach. Exploiting the recent advances in machine learning for catalyst design (dos Passos Gomes et al., 2021) could therefore be promising.Despite the use of high-level QC methods, there is no guarantee that the prediction of catalytic activity does not overestimate the performance of candidates. Therefore, we recommend integrating in silico design with selected experiments in the future. A first step would be testing the top candidates obtained with CAT-COSMO-CAMPD experimentally as already shown for our approach to integrated design of solvents and separation processes (Scheffczyk et al., 2018). The experimental results could be used to confirm the catalytic activity and, if required, to modify the used combination of QC methods for a second design run with CAT-COSMO-CAMPD.In summary, this article shows the promise of methods for integrated in silico design of catalysts and processes. The proposed CAT-COSMO-CAMPD method was demonstrated successfully in the presented case study and is generally applicable to various other systems. As computational design of molecular catalysts is still a quite unexplored field of research, we expect major developments in the coming years. Christoph Gertig: Conceptualization, Methodology, Software, Formal analysis, Visualization, Writing – original draft, Writing – review & editing. Lorenz Fleitmann: Methodology, Software, Writing – review & editing. Carl Hemprich: Methodology, Software, Formal analysis, Visualization, Writing – review & editing. Janik Hense: Methodology, Formal analysis, Writing – review & editing. André Bardow: Conceptualization, Funding acquisition, Writing – review & editing, Supervision. Kai Leonhard: Conceptualization, Funding acquisition, Writing – review & editing, Supervision.The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.C.G., A.B. and K.L. gratefully acknowledge funding from the German Federal Ministry of Education and Research (BMBF) within the project Carbon2Polymers (03EK30442C). L.F. thanks the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) for funding under Germany’s Excellence Strategy - Cluster of Excellence 2186 "The Fuel Science Center” - ID: 390919832. Furthermore, the authors are grateful to J. Langanke, M. Leven and E. Erdkamp for valuable discussions. Simulations were performed with computing resources granted by RWTH Aachen University under projects rwth0284 and rwth0478.Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.compchemeng.2021.107438. Supplementary Data S1 Supplementary Raw Research Data. This is open data under the CC BY license http://creativecommons.org/licenses/by/4.0/ Supplementary Data S1

Catalysts are of paramount importance as most chemical processes would be uneconomical without suitable catalysts. Consequently, the identification of appropriate catalysts is a key step in chemical process design. However, the number of potential catalysts is usually vast. To suggest promising candidates for experimental testing, in silico catalyst design methods are highly desirable. Still, such computational methods are in their infancy. Moreover, simple performance indicators are commonly employed as design objective instead of evaluating the actual process performance enabled by considered catalysts. Here, we present the CAT-COSMO-CAMPD method for integrated in silico design of homogeneous molecular catalysts and processes. CAT-COSMO-CAMPD integrates design of molecular catalysts with process optimization, enabling a process-based evaluation of every designed candidate catalyst. Reaction kinetics of catalytic reactions are predicted by advanced quantum chemical methods. We demonstrate for a catalytic carbamate-cleavage process that CAT-COSMO-CAMPD successfully identifies catalyst molecules maximizing the predicted process performance.

The alarming increase of anthropogenic emissions of greenhouse gases (GHG) is encouraging extensive research to mitigate the impact of these emissions. The transition to low-carbon societies demands strategies to reduce GHG emissions, consisted mainly of CO2 and CH4. Most of the CO2 emissions come from the consumption of fossil fuel for energy (International Energy Agency, 2018). Nevertheless, it is less known the important contribution of intensive livestock and other organic waste to GHG emissions, which come from anaerobic digestion (IPCC Fourth Assessment Report, 2014). These emissions, mainly in the form of biogas (a mixture primarily formed by CO2 and CH4), were regarded a waste rather than a value, but this trend is changing. Biogas industry business model is moving to two different scenarios: (i) upgrading to biomethane in order to produce biofuels to generate energy and (ii) chemical valorisation via reforming to produce syngas, primarily formed by H2 and CO, which is an interesting platform chemical in chemical synthesis. Indeed, syngas is widely used as a precursor to synthetise fuels and other hydrocarbons via Fischer-Tropsch process, or other value-added products, such as methanol or acetic acid (Martín-Espejo et al., 2022). Biogas upgrading via chemical valorisation is deemed an emerging trend which may provide many opportunities to the chemical industry to tackle GHG emissions (Baena-Moreno et al., 2021). In this context, thermocatalytic biogas upgrading to syngas can be addressed via dry reforming of methane (DRM, Eq. (1)), representing an attractive way to convert CO2 and CH4. (1) C O 2 + C H 4 → 2 CO + 2 H 2 Typically, conventional catalysts for DRM are formed by a metal active phase which is dispersed over a support structure. Noble metals, like Ru, Rh or Pt, are reported to display great performance. Nevertheless, due their scarcity and high market prices, research focus is switching into the use of transition metals, such as Ni and Co (Jang et al., 2019; Sharifianjazi et al., 2021). Thermodynamically, DRM is a very endothermic reaction which requires high temperatures and energy inputs. At such temperatures, conventional catalysts are prone to deactivation due to sintering of the active phase and formation of coke deposits coming from side reactions such as Boudouard reaction (Eq. (2)), CH4 decomposition (Eq. (3)) and CO and CO2 reduction (Eqs. (4) and (5)) (Nikoo and Amin, 2011). Extensive investigation has been conducted in order to find cost-effective catalysts which avoid carbon deposition and sintering of the active phase while exhibiting acceptable catalytic performance. (2) 2 CO → C + C O 2 (3) C H 4 → C + 2 H 2 (4) CO + H 2 → C + H 2 O (5) C O 2 + 2 H 2 → C + 2 H 2 O Nickel-based catalysts are frequently chosen among other transition metals owing to their commendable activity, low price and availability. However, these catalysts suffer from coking and sintering, which leads to rapid deactivation. Multiple strategies have been applied to deal with the deactivation of these catalysts. Recent literature review reports highlight the most promising advances to design nickel-based DRM catalysts resistant to deactivation (Huang et al., 2022; Le saché and Reina, 2022; Yentekakis et al., 2021). The use of bimetallic formulations, the addition of promoters or the combination of different support structures are typical strategies used to enhance the performance of Ni-based catalysts, tuning the nature of the catalysts. In general, chemical and structural properties (e.g., redox, acid/base, oxygen mobility) are tuned using these strategies, making an impact on the stability and reaction mechanism. Besides, dispersion and particle size can be influenced by these parameters. As a promising alternative, the use of inorganic complex structures is aimed at stabilising the active phase in the structure while remaining active and accessible. Spinels, sandwich, tubular or mesoporous structures, hydroxyapatite, hexaaluminate, hydrotalcite, perovskites and pyrochlores have been investigated to improve the performance of DRM (Bhattar et al., 2021; Le Saché and Reina, 2022). Pyrochlores, with the formula A2B2O7, and perovskites, ABO3 and A2BO4, are mixed oxide materials in which A-site is typically substituted by a large rare-earth trivalent metal whereas B-site is substituted by a tetravalent transition metal of smaller diameter. These materials are highly crystalline, possess great thermal stability and oxygen mobility, which makes then suitable for high temperature and coke-prone processes, such as dry reforming of methane (Xu et al., 2020). For these very reasons, these mixed oxide materials have been previously studied for biogas reforming. For instance, Bhattar et al. (2020) studied the effect of the addition of Sr- and Ca- to Ni-substituted lanthanum zirconate catalysts. It was found that the addition of small amount of Sr improved the performance of the catalyst as well as an increase of the resistance of the catalyst to deactivation from carbon deposition. On the other hand, La2Ce2O7 and LaNiO3 were synthetised by Ramon et al. (2022) in order to elucidate the catalytic activity of this catalyst in DRM reaction, comparing two different synthesis methods. In another study by Ma et al. (2014), they investigated the effect of nickel-supported La2Zr2O7 pyrochlore-like materials for steam reforming of methane, showing an excellent catalytic behaviour since coking resistance was highly improved. Two studies by Le Saché et al. (2018a, 2020) successfully proved the incorporation of Ni into a La2Zr2-xNixO7-δ pyrochlore structure for dry and bi-reforming of methane. According to the XRD results, the formation of a perovskite-type La2NiZrO6 is responsible for the great catalytic performance. Bai et al. (2022) have studied the effect of the substitution of nickel over both cerium and zirconium on B-site La2(CeZrNi)2O7 for dry reforming of methane. In this study, it is believed that the exceptional oxygen vacancies and the interaction of the exsolved Ni with the support were key properties for the outstanding performance of the catalysts. Nevertheless, the incorporation of just cerium and zirconium to form a complex inorganic structure has not been tried for dry reforming despite cerium´s excellent oxygen storage/release ability (Teh et al., 2021). Attempts have been tried for Ce2Zr2O7 materials in photocatalysis for organic pollutants abatement, as Jayaraman and Mani (2020) have studied over a g-C3N4 support structure. Ce2Zr2O7 has also been studied on PbS in order to study the electrochemical properties as a supercapacitor electrode (Bibi et al., 2019). As of today, no attempts have been made to study the catalytic activity of Ni-substituted Ce2Zr2O7 for DRM.In this scenario, this work addresses the design of inorganic complex structures to stabilise nickel, leading to robust catalysts for DRM. Under this premise, the utilisation of a thermally stable cerium zirconate oxide structure is studied, inserting and stabilising nickel within the structure (Ce2Zr2-xNixO7-δ). This study is focused on the synthesis, catalytic activity, pre- and post-characterisation of Ni-substituted cerium zirconate for DRM. Specifically, different loading of Ni, from 0 to 15 wt.%, were incorporated to the structure, substituting on the B-site of the mixed oxide structure. Our work showcases an effective strategy to design robust and economically viable gas-phase CO2 conversion catalysts with potential applications in reforming units and biogas plants.The catalysts were prepared through a modified version of the original Pechini method (Pechini, 1967), which is described elsewhere (Gaur et al., 2011; Kumar et al., 2016; Tietz et al., 2001). This method was chosen since it is reported to produce uniform substituted and non-susbtituted catalyst crystals (Haynes et al., 2008). Cerium nitrate (Ce(NO3)3·6H2O), zirconyl nitrate (ZrO(NO3)2·6H2O), provided by Sigma-Aldrich, and nickel nitrate (Ni(NO3)2·6H2O), provided by Alfa Aesar, were used as precursors. The necessary amount of each precursor was separately dissolved in deionised water and then mixed together. Citric acid (CA) was dissolved in deionised water and incorporated to the mixture of precursors while stirring at room temperature. The amount of citric acid added was molar ratio of CA:metal of 1.2:1. The solution was heated while stirring to 90°C to ensure metal complexation and ethylene glycol (EG) was added to the solution drop-wise, using a molar ratio of EG:CA 1:1. The solution was then continuously stirred and concentrated due to evaporation of the water until the appearance of a viscous gel. The stirring was then stopped and the dense gel was left at 90-100°C to promote the polyesterification reaction between the citric acid and ethylene glycol. The decomposition of the nitrate precursors led to large plumes of NOx gas. Once the toxic gas is released, the solid was dried at 100°C overnight. The resulting compound was powdered manually in an agate mortar and then calcined in a crucible at 1000°C during 8 h, using a heating rate of 7.5°C min−1, to ensure phase transition. To simplify, a special notation is chosen and the catalysts will be referred as CZ, CZN5, CZN10, CZN12 and CZN15 for 0, 5, 10, 12.5 and 15 wt.% of Ni, respectively.The textural properties of the samples were characterised by nitrogen adsorption-desorption measurements at liquid nitrogen temperature (-195.8°C) in a Micromeritics Tristar II apparatus. Before analysis, the samples were out-gassed under vacuum conditions at 250 °C for 4 h. The specific surface area (SBET) was calculated using the Brunauer-Emmet-Teller (BET) method. The average pore volume was determined as the ratio of the pore volume and the specific surface area. This was then normalised using a coefficient which depends on the pores shape.The metal content of Ni was measured by inductively coupled plasma spectroscopy (ICP-MS) using iCAP 7200 ICP-OES Duo (ThermoFisher Scientific) spectrometer previous microwave digestion in an ETHOS EASY (Milestone) microwave digestion platform.X-Ray Diffraction (XRD) measurements were carried out on X'Pert Pro PANalytic diffractometer with Cu-Kα anode at room temperature, working at a voltage of 45 kV and a current of 40 mA. The diffractograms were registered between 20 and 90° (2θ) with a step size of 0.05° and a step time of 300 s. The structural determination was done by comparison with PDF2 ICDD2000 (Powder Diffraction File 2 International Center for Diffraction Data, 2000) database.Temperature-programmed reduction (TPR) with H2 was carried out on the calcined catalysts in a conventional U-shaped quartz reactor connected with a thermal conductivity detector (TCD) using a flow of 50 mL min−1 of 5% H2 (v/v) diluted in Ar. TPR measurements were performed using 100 mg of each catalyst and a heating rate of 10°C min−1 from room temperature to 900°C, using a CO2 (s)/acetone cold trap to condense the water formed during the process.Scanning electron microscopy (SEM) was carried out on the calcined catalysts under vacuum using a Hitachi S4800 SEM-FEG 0.5-30 kV voltage microscope using a cold cathode field emission gun of 1 nm resolution and equipped with a Bruker-X Flash-4010 EDS analyser.Transmission electron microscopy (TEM) of the samples were performed on a JEOL 2100Plus (200 kV) microscope. It was equipped with an Energy Dispersive X-Ray analysis system (EDX X-Max 80T, Oxford Instruments) and a CCD camera for image recording.The catalytic performance of the prepared samples for DRM reaction was evaluated under atmospheric pressure in a tubular, continuous flow fixed-bed reactor (Hastelloy reactor) with an internal diameter of 9 mm in an automatised Microactivity Reference apparatus from PID Eng&Tech. Stability tests were performed in a tubular fixed bed quartz reactor with an inside diameter of 10 mm.The catalysts were sieved and the 100-200 µm fraction was placed in the reactor over a quartz wool bed. Prior to the activity test, the catalysts were in situ reduced in a flow of 50 mL min−1 40% H2 (v/v) in He, at 800°C for 1 h using a heating rate of 7.5°C min−1. The reaction was performed passing a reactant feed flow of 100 mL min−1 and molar ratio of N2:CH4:CO2 2:1:1, every 50°C from 500 to 800°C until achieving the steady state on each step. The WHSV (Weight Hourly Space Velocity) was fixed at 30 L gcat −1 h−1. Furthermore, stability tests were conducted at the same conditions, at 600 °C and 800°C, during 100 h.The composition of the product gas stream was monitored using an on-line gas chromatography (Agilent Technologies) equipped with a HayeSep Q and Mol sieve 5A column. An ABB AO2020 on-line gas analyser was used to determine the composition of the product gas stream in the stability tests. The spent samples were recovered for post-reaction characterisation. In all the cases, carbon balance was closed ±5%.The conversion (Xi) of the reactants (Eqs. (6) and (7)) and the H2/CO molar ratio (Eq. (8)) was calculated in order to evaluate the catalytic behaviour. The conversion was calculated as follows: (6) X C H 4 ( % ) = F C H 4 , in − F C H 4 , out F C H 4 , in · 100 (7) X C O 2 ( % ) = F C O 2 , in − F C O 2 , out F C O 2 , in · 100 (8) H 2 / CO = F H 2 , out F CO , out where F is the molar flow of CH4, CO2, H2, and CO, respectively, and the subscripts in or out correspond to either the inlet and the outlet reactor flow.Chemical composition of nickel and textural properties of the prepared samples are listed in Table 1 . The metal loading of the catalysts is close to the nominal values of 5, 10, 12.5 and 15 wt.% of Ni, witnessing the successful preparation method to carefully adjust the desired active phase loading. Still, nickel amounts of the high-Ni containing samples (CZN12 and CZN15) are slightly lower than the intended values which might indicate a threshold on optimal Ni uptake.Regarding the textural properties of the samples, we observe the noteworthy low surface area of all of the synthesised materials in contrast to benchmark supported catalysts, which have much higher surface area (e.g., Ni-doped Al2O3-CeO2 with SBET = 208 m2 g−1 (Marinho et al., 2021)). Therefore, this factor is important to consider since the engineered materials herein are considered to be active for DRM despite their low surface area. In other words, the DRM reaction does not necessarily require high-surface area catalysts to reach high performances as we will evidence further down in this work. Furthermore, a slight reduction of the surface area is observed when the Ni loading is increased. This reduction is more evident in the pore volume which indicates a certain degree of Ni particles blockage of the catalysts´ pores.In order to confirm the formation of the different crystalline inorganic structures, XRD analysis of the calcined catalysts was performed. The resulting normalised diffractograms of the Ni-doped cerium zirconate oxide structures with 0, 5, 10, 12.5 and 15 wt.% are presented in Fig. 1 . No characteristic diffraction peaks of individual CeO2 oxide phases are observed, which may suggest the incorporation of Ce into the inorganic lattice structure. CZ undoped catalyst pattern presents the characteristic diffraction features of Ce0.5Zr0.5O2 tetragonal structure with space group P42/nmc (ICDD Card No. 00-038-1436) and Ce2Zr2O7 pyrochlore cubic structure Fm 3 ¯ m (ICDD Card No. 00-008-0221). Interestingly, when Ni loading is increased, a slight shift in the diffraction peaks towards lower angles is observed between the undoped material (CZ) and the doped catalysts (CZNX). Therefore, the partial substitution of Ni on the B-site affects the crystalline structure. This shift can be observed in 29.4, 33.9, 48.9 and 58.1° 2θ of the characteristic diffraction features of Ce0.5Zr0.5O2. The incorporation of Ni using the synthesis method produces a change in the final structure of the cerium zirconate oxide, incrementing the lattice parameter of the material, as reported when doping these structures (Haynes et al., 2008; Le Saché et al., 2018b; Pakhare et al., 2013). On the other hand, it is observed that Zr is not completely incorporated into the inorganic complex structure in CZN5 and CZN10, appearing a diffraction line around 30° 2θ which may correspond to ZrO2 tetragonal structure with space group P42/nmc (ICDD Card No. 00-024-1164).For CZN5, an interesting effect is observed. Diffraction lines abovementioned, associated to the inorganic mixed oxide, start to unfolding, distinctly observing two peaks with similar intensity instead of one, as emphasised in the inset of Fig. 1. It seems that two crystalline phases are clearly formed, appearing the characteristic features of Ce0.5Zr0.5O2 tetragonal structure (ICDD Card No. 00-038-1436) and the separate Ce2Zr2O7 pyrochlore cubic structure (ICDD Card No. 00-008-0221). For Ce2Zr2O7 pyrochlore cubic structure, the diffraction lines at 2θ values of 28.9, 33.6, 48.1 57.1, 59.9 and 70.4° are attributed to the (222), (400), (440), (622), (444) and (800) crystal planes. Considering the radius ratio rA/rB, which is relevant factor for A2B2O7-δ inorganic complex structures (Bai et al., 2022; Xu et al., 2020), the substitution of Ni on B-site within the structure produces a decrease of this parameter since Ni radius is smaller than that of Zr, which is associated with a less ordered structure. When the loading of Ni is further increased, the unfolding phenomenon vanished, appearing mainly one diffraction pattern corresponding again to Ce2Zr2O7 cubic structure (ICDD Card No. 00-008-0221). Therefore, it is proved that the incorporation of Ni produces variations in the inorganic oxides formed in this material.Regarding the active phase, it is observed that part of the Ni is inserted in the inorganic structure. Cubic perovskite CeNiO3 phase with space group Pm 3 ¯ m (Material project Card No. mp-866095) can be identified in all the doped materials at 33.5, 48.1 and 59.8°. Besides, the presence of some NiO domains can be observed in the diffractogram. Diffraction lines 37.3, 43.3, 62.9 and 75.4° 2θ are attributed to rhombohedral NiO (ICDD Card No. 00-022-1189). In general, the intensity of the diffraction line attributed to NiO species remains the same as the Ni content increases. Interestingly, in the most intense diffraction line of NiO, it is observed a slight displacement of the peak towards higher degrees as the nickel content is increased, reaching 43.5° 2θ for CZN15. This increase is associated to NiO1-x species, where x is higher as the increase of this diffraction line angle. Indeed, the ratio Ni/O is closer to orthorhombic Ni4O3 (mp-656887) than NiO in CZN15. This is closely related to the diffraction lines shift previously reported of the mixed oxide structure to lower angles. Therefore, it may be concluded that Ni is distributed in the lattice structure, interacting differently with the oxide species resulting of the formation of this material.Further insights on the redox behaviour of our mixed-oxide systems were gathered by temperature-programmed reduction (TPR) experiments. The resulting H2 consumption profiles are depicted in Fig. 2 . The composition of the samples, interactions among the active species of the calcined catalysts and the conditions necessary for the pre-treatment step were also analysed in light of the TPR data. The undoped material CZ has been found to have small reducibility. A small H2-consumption signal appears due to the possible interaction with the Ce species inserted in the catalyst. This signal appears between 400 and 540°C. Nevertheless, this reduction event is small. Overall, the observed H2 consumption in the Ni-based materials is mainly due to the reducibility of Ni species upon its incorporation in the catalysts. In the profiles herein presented of the doped materials, three main reduction regions are worth considering. The first region, with a temperature peak between 365 and 375°C, can be attributed to the reduction of NiO1-x species which are interacting with the Ce0.5Zr0.5O2 phase. The presence of this phase is presented mainly in CZN5 XRD in Fig. 1, which is consistent with the results. TPR signal is reduced for CZN10, CZN12 and CZN15, since this phase is disappearing when the Ni content is increased in the samples. A medium temperature region is observed between 400 and 435°C. This phase can be attributed to the interaction of NiO1-x species with the Ce2Zr2O7 pyrochlore phase. In this case, it is observed that the peak is displaced to higher temperatures as the Ni content is increased, which may indicate a stronger interaction with the phase. Besides, the intensity of the signal increases with the Ni loading, which may indicate that higher Ni loadings may lead to larger proportions of NiO1-x. Finally, the most intense signal, between 535 and 580°C, is attributed to the partial reduction of CeNiO3 phase, which is more difficult to reduce. This signal, in general, decreases as the Ni content is increased.XRD analysis was also performed on all the samples reduced at 800°C for 1 h. NiO1-x species were reduced to Ni0 since the characteristic peaks of NiO1-x shifted to higher diffraction line angles after the H2 treatment. Characteristic peaks of Ni0 at 44.6 and 50.0° 2θ are observed (ICDD Card No. 01-087-0712). Nevertheless, CeNiO3 phase still appears, which may indicate that this phase is just partially and/or superficially reduced but bulk crystalline domains remained upon the selected pre-reduction treatment.In order to calculate the size of Ni crystallite, Scherrer equation is used for all the samples in which Ni is incorporated, using the most intense peak, 44.6° 2θ. The results can be observed in Table 2 . In general, there is homogeneity in the size of Ni crystallite. No increments are observed as the Ni load is increased in the catalysts. Ni particle size is expected to be no bigger than 35 nm and well-dispersed on the surface of the inorganic structure despite the low surface area of these catalysts.To visualise the morphology of the catalysts, SEM analysis was performed in the reduced catalysts. As can be observed in Fig. 3 a, corresponding to the undoped inorganic structure CZ, a homogeneous, dense structure with small cavities is observed throughout the sample. No granular or spherical shape is observed. CZN5, on the other hand, presents two different structures. In Fig. 3b, a similar porous structure is observed, which may correspond to the Ce0.5Zr0.5O2 phase, whereas Fig. 3c structure revealed spherical shape of Ce2Zr2O7, as reported elsewhere (Bibi et al., 2019). Fig. 3d corresponds to CZN10, which still present zones with some spherical shape and other complex structures. Fig. 3e presents element mapping distribution from EDX, showcasing the homogenous distribution of our active phases.The reduced catalysts were tested under the above DRM reaction conditions. The effect of the temperature in DRM was studied at a temperature range from 500 to 800°C and atmospheric pressure, using a reactant gases molar ratio of CO2:CH4 1:1. As can be observed in Fig. 4 , CO2 conversion is greater than CH4 conversion for all the catalytic systems at all temperatures. This may be due to the higher activation energy of CH4 than CO2 (i.e., the energy barrier to active C-H cleave bond is higher than CO2 dissociation), requiring higher temperatures in agreement with DFT results reported elsewhere (Niu et al., 2020; Zhu et al., 2009). Besides, the possible occurrence of the reverse water-gas shift (RWGS, Eq. (9)) reaction, which competes with DRM, may contribute to the higher CO2 conversion due this parallel route consuming CO2 simultaneously. (9) C O 2 + H 2 → CO + H 2 O Due to the endothermic nature of the reaction, both CO2 and CH4 conversion increase with temperature. At low temperatures, CH4 conversion is low and far from the equilibrium values but, as the temperature rises, it gets better, reaching a conversion of 45% at 800°C for CZN10. As mentioned before, CH4 needs higher temperatures to overcome the energy barrier necessary for its activation. Like CH4, CO2 conversion is lower at low temperature range, but it becomes higher as the temperature increases. At 800°C, the conversion reached a value of 60% for CZN10, which is closer to equilibrium conditions. These commendable catalytic results are achieved despite of the low specific surface area reported. Focusing on the two best results, CZN5 and CZN10, the conversion gap between them become closer as the temperature is increased for both CH4 and CO2, even surpassing CO2 conversion offered by CZN10 catalysts if compared to CZN5 at 800°C.The un-doped CZ catalyst does not show activity in DRM, as can be predicted due to the lack of active metallic Ni phase in the solid. As the metal loading is increased to 10 wt.%, the conversion of both CH4 and CO2 rises due to the higher Ni concentration presented in the sample. Interestingly, CZN5 presents conversion values close but lower to CZN10 despite having half of Ni metal content. This can be related to the accessibility of Ni active sites to the reactant gases due to the interaction with the phases presented in the catalyst. The presence mixed phases in CZN5 leads to remarkable activity results which are close to the those exhibited by the CZN10, which contains twice the Ni content. Particularly, this might be related to the presence of Ce0.5Zr0.5O2 phase, which interacts more closely with Ni active centres, offering a more active catalyst according to TPR results. In any case, CZN10 shows the highest catalytic performance among the studied series. When the metal loading is further increased to 15 wt.% Ni, a decrement of the conversion can be observed despite having more Ni in the samples. This can be related again to the accessibility of Ni active centres. The lower presence of Ce0.5Zr0.5O2 phase as the Ni content is increased might be responsible for the decrease of conversion when compared to CZN5 and CZN10. Indeed, it appears to be an optimum amount of Ni which maximise the conversion in DRM, being close to 10 wt.%. Despite the more insertion of Ni in the structure on B-site, substituting Zr, the activity did not improve.In terms of H2/CO molar ratio, the tendency is to increase the ratio as the temperature increases, as observed in Fig. 5 . This increment might be related to a higher CH4 conversion, leading to better H2 production and thus higher ratio. The tendency for all the doped samples is similar, which may indicate the presence of parallel competing reactions. RWGS leads to the opposite effect in the ratio since it consumes H2, whereas methane decomposition generates H2. In fact, among the parallel reactions affecting DRM, RWGS and CH4 decomposition have a remarkable influence in the reactant conversion and H2/CO molar ratio, as previously reported in literature (Bradford and Vannice, 1999; Le Saché et al., 2018a, 2018b). These results may suggest that the contribution of CH4 decomposition side reaction is favoured at higher temperature, thus yielding a higher H2/CO ratio.In order to further explore the catalytic performance, stability tests were performed to study how efficient the catalyst is over long-term reaction run. CZN10 was chosen to conduct the stability tests to check its behaviour over a period of 100 h at 600°C and 800°C. The results at 600°C can be observed in Fig. 6 a. This catalyst displays good stability over time. It is observed a first period of stabilisation to reach the steady state and, after that, it shows signs of small deactivation starting with conversions of CO2 and CH4 of 24% and 18% and achieving conversions of 18% and 12% after 100 h on-stream, respectively. Again, the conversion of CO2 is slightly larger than that of CH4. Besides, a decrease in the molar H2/CO ratio from 0.6 to 0.45 is noted. It is estimated a declination rate of 0.0758 and 0.0701% h−1 for CH4 and CO2, respectively. CH4 is acknowledged to be activated by Ni active centres whereas CO2 by the support. Therefore, it is reasonable that, if deactivation is occurring due to coke deposition or sintering of the active sites, CH4 is reported to be slightly more sensitive to deactivation than CO2.At 800°C (Fig. 6b), results are slightly different and very promising. The conversion displays stable conversion levels over time. CO2 conversions sets around 49-51% whereas CH4 conversion is slightly lower. CH4 conversion is again reported to be more sensitive to a small deactivation than CO2 conversion. The molar H2/CO ratio is, in this case, quite stable and over 0.66 which be a useful syngas for some industrial applications such as hydroformylation reactions (Le Saché and Reina, 2022).For a broad picture to place our catalysts within the DRM scenario, Table 3 offers a comparison of our results with relevant studies available in literature. Herein, we must emphasise the low surface area of the catalysts presented in this work when compared reference catalysts. Very interestingly our work demonstrates that the DRM reaction does not necessarily require high-surface area catalysts to reach high performances. In other words, the DRM reaction is not a surface-area sensitive process. In addition, we shall remark that reaction time herein reported for our stability tests is considerably higher than standard experimental data from literature, providing stronger evidence of the stability and resistance of the materials. Actually, most of the reports included in the table test the catalysts for very short time to be considered realistic stability tests and we want to draw the readers attentions to this matter since long-term stability test of 100 h and beyond are needed as solid proof-of-concept for the catalyst´s resistance towards deactivation in a process like DRM.Overall, our CZN10 is deemed as a fairly stable catalyst when tested at 800°C delivering noticeable levels of conversion and a commendable H2/CO molar ratio in continuous operation during 100 h. As a result, our best formulation leads to a valuable syngas composition with potential interest for the chemical industry. Furthermore, it is interesting to remark that operational troubles with catalyst´s stability observed at 600°C can easily be overcome by rising the reaction temperature to 800°C. This is still a low temperature regime when it comes to industrial reformers which typically run on the 900-1000°C range, opening further opportunities for our catalytic formulation. Additionally, we shall highlight that the excellent catalytic performance displayed by our samples is achieved at high space velocities (i.e., 30 L g−1 h−1). Again, industrial reformers are operated at significantly lower space velocities which means that the implementation of our catalysts in a potential realistic application could lead to considerable reduction of the reforming reactor volume; or in other words, significant CAPEX savings.Deactivation of the catalyst is mainly caused by carbon deposition and/or sintering of the active phase. In order to elucidate this, the best performing catalyst was analysed after reaction by XRD in order to detect any structural changes after the different treatments it underwent. In Fig. 7 , the XRD pattern of CZN10 calcined, after the reduction treatment, after DRM reaction conditions at 600°C and 800°C for 100 h are shown. As it can be observed, the inorganic crystalline structure of the sample remains intact since there are no differences in the diffraction characteristics between the calcined and the post-reaction catalysts. This demonstrates the high thermal stability of the catalyst despite the reaction conditions. In addition, it can be observed a slight displacement of the characteristic peaks 44.4 and 51.7° 2θ, corresponding to the characteristic planes of (111) and (200) of metallic Ni0 to lower angles after 100 h at 600°C, indicating that part of the surface Ni0 is oxidising again to form NiO1-x species. On the other hand, the XRD pattern after 100 h at 800°C shows that NiO is appearing very likely due surface oxidation when transferring the sample from the reactor to the XRD chamber.Small growth of Ni particles is observed after reaction conditions. It was estimated a growth of the particle size from 33.7 to 39.8 and 37.3 nm at 600°C and 800°C, respectively. This small change occurred after 100 h of reaction, which affirmed the stability and robustness of the catalyst after the substitution of Ni on the complex oxide structure, preventing partially the sintering of the particles. It is thus estimated that a small proportion of the Ni inserted within the inorganic structure is exsolved, producing this increase in the particle size. Exsolution is actually considered a smart strategy to design efficient catalysts for energy applications when the metal has a good capacity to exsolve under reaction environments (Carrillo and Serra, 2021; Kousi et al., 2021; Kwon et al., 2017; Zhang et al., 2020).Carbon deposition is also studied as deactivation cause. The formation of carbon is observed in XRD pattern after 100 h at 600°C, where a peak attributed to graphitic carbon is detected. This peak corresponds to the graphite lattice plane (002) of carbon nanotubes at 26° 2θ. At 800°C, no carbon structures are observed. Carbon formation is hard to avoid due to the intricate reaction, since C-H activation of CH4 involves the formation of carbon species, widely studied elsewhere (Guharoy et al., 2019). Besides, the reaction temperature has an important influence on carbon deposition. It must be emphasised that carbon deposition is thermodynamically more favoured at lower temperatures, between 600 and 750 °C (Nikoo and Amin, 2011). Besides, carbon deposition is closely related to Ni particle size, since the larger the clusters the more favoured the carbon deposition is. This may be another reason of carbon deposition due to the increment of Ni crystallite size.TEM images of the (a) reduced, (b) after 48 h reaction and (c) after 100 h at 600 °C reaction are shown in Fig. 8 . Fig. 8a shows a particle of the reduced catalyst. After 48 h at 600 °C, some carbon nanotubes (CNT) are formed, as observed in Fig. 8b. Nevertheless, the amount is negligible. After 100 h at 600 °C, significant amount of CNTs appear. These CNTs start growing from the interface of the active metal phase and the support structure, “pulling out” part of the Ni particle from the surface. Nevertheless, despite the formation of carbon deposits, the catalyst remains acceptable since this carbon is partially covering the active sites of Ni, being the rest of the Ni atoms accessible for the reaction. A similar behaviour was reported for a Ni-substitute La pyrochlore (Le Saché et al., 2018a). The main causes of catalyst deactivation at 600 °C are the carbon deposits around Ni particles and the increase of Ni clusters. This situation is overcome when the reaction is run at 800 °C where our post-stability XRD pattern shows a carbon-free sample which is explain the excellent conversion levels at this reaction conditions for a continuous 100 h test.This work addresses the preparation, characterisation and testing of a series of Ni-promoted cerium zirconate oxide structures for gas-phase CO2 valorisation via DRM. Structural analysis revealed the presence of different inorganic mixed oxide structures depending on the amount of nickel incorporated. Nickel is believed to be incorporated within the lattice structure, remaining part of the nickel in the surface interacting with the support structure resulting in a complex structure with different types of Ni active species as evidenced by XRD and TPR experiments.CZN10 showed the best catalytic performance for DRM. Nevertheless, CZN5, having half of the nickel content, presented commendable conversion values, indicating that the crystalline Ce0.25Zr0.25O2 phase, which is only presented in CZN5, is a relevant specie that makes Ni more accessible enhancing the DRM behaviour. Stability test of CZN10 over 100 h demonstrated the long-term thermal stability of the catalysts, showing small deactivation in the low-temperature range (600 °C) and excellent stability at 800 °C. Such deactivation at 600 °C is ascribed to graphitic carbon deposition as evidenced by TEM. XRD analysis after 100 h hours revealed that the crystalline structure of the catalyst remained intact and part of the nickel was exsolved, slightly increasing the particle size of the Ni clusters, which are responsible for the activity of the catalyst. Potential regeneration of spent catalyst in the low-temperature operation range (600 °C) should be further investigated as smart strategy to reuse the catalysts in realistic applications.All in all, this work showcases a strategy to design thermally stable catalysts based on nickel promoted cerium-zirconium mixed oxides where nickel is incorporated within the structure, being able to withstand DRM conditions and deliver high quality syngas for long-term operations. Very importantly, the remarkable performance herein demonstrated by this Ni-engineered catalysts is achieved at relatively high space velocities when compared to industrial reformers which means that they can be instrumental for CAPEX savings in realistic applications while also paving the way to design compact DRM units that might fit very well in the biogas processing industry.Authors declare no conflict of interest.Financial support for this work was gathered from grant PID2019-108502RJ-I00 and grant IJC2019-040560-I both funded by MCIN/AEI/10.13039/501100011033 as well as RYC2018-024387-I funded by MCIN/AEI/10.13039/501100011033 and by ESF Investing in your future.

The increasing anthropogenic emissions of greenhouse gases (GHG) is encouraging extensive research in CO2 utilisation. Dry reforming of methane (DRM) depicts a viable strategy to convert both CO2 and CH4 into syngas, a worthwhile chemical intermediate. Among the different active phases for DRM, the use of nickel as catalyst is economically favourable, but typically deactivates due to sintering and carbon deposition. The stabilisation of Ni at different loadings in cerium zirconate inorganic complex structures is investigated in this work as strategy to develop robust Ni-based DRM catalysts. XRD and TPR-H2 analyses confirmed the existence of different phases according to the Ni loading in these materials. Besides, superficial Ni is observed as well as the existence of a CeNiO3 perovskite structure. The catalytic activity was tested, proving that 10 wt.% Ni loading is the optimum which maximises conversion. This catalyst was also tested in long-term stability experiments at 600 and 800°C in order to study the potential deactivation issues at two different temperatures. At 600°C, carbon formation is the main cause of catalytic deactivation, whereas a robust stability is shown at 800°C, observing no sintering of the active phase evidencing the success of this strategy rendering a new family of economically appealing CO2 and biogas mixtures upgrading catalysts.

Metal nanoparticles are commonly used to catalyze many chemical processes [1]. Since catalytic reactions occur at the metal surface, the high surface-area-to-volume-ratio of a nanoparticle provides an effective number of active sites per weight of metal in the overall catalyst. One of the biggest challenges for any catalytic system is to maintain this maximum amount of active sites throughout the lifespan of a solid catalyst [2]. Particularly in the case of metal nanoparticles, their growth is a prevailing phenomenon in which the efficient utilization of the metal in a catalyst is compromised, usually leading to a detriment in catalytic performance [3–5]. In order to stabilize the nanoparticles and prevent their growth, these are typically dispersed over a support material. The nature of the support is crucial in delivering this stability and offers an opportunity to develop improved solid catalysts.Reducible oxides used as support material display characteristic interactions with metal nanoparticles. During reductive conditions, one effect that arises is coverage of the nanoparticles by in-situ generated suboxides from the support [6,7], the so called Strong Metal-Support Interaction (SMSI) [8,9]. This effect can modify the available metal surface area, the electronic state of the metal and the particle shape [10–12]. An interesting example is the substantial change in reactivity of nickel during carbon monoxide hydrogenation: nickel supported on non-reducible oxides (e.g. Al2O3, SiO2) selectively hydrogenates carbon monoxide to methane, whereas when supported on reducible oxides like TiO2 or Nb2O5 the product distribution shifts towards heavier hydrocarbons [13–17]. Furthermore, previous reports have suggested that reducible supports can also deliver unique stability to nickel-based catalysts [13,18].The interest to achieve stable nickel-based systems in the presence of carbon monoxide arises from the extensive utilization of nickel catalysts in reactions involving carbon monoxide as reactant, intermediate or product [19–23] and the poor stability of nickel in the presence of carbon monoxide at low temperatures [24,25]. Deactivation of nickel-based catalysts during carbon monoxide hydrogenation proceeds most often through particle growth by the formation and diffusion of volatile nickel carbonyl [25–27]. This phenomenon is a classic example of Ostwald ripening, where species containing metal atoms, in this case nickel carbonyl, diffuse from smaller towards larger nanoparticles leading to metal sintering [28,29]. In order to prevent this, the reaction is typically operated at high temperatures and low CO pressures, since these conditions disfavor the formation of nickel carbonyl [24,26,30]. However, such conditions compromise the product selectivity mainly towards methane and therefore hamper the application of nickel catalysts for the synthesis of more commercially attractive products, such as long-chain hydrocarbons (C5+) or olefins [31,32]. Alternative strategies to inhibit the formation of nickel carbonyl in these catalysts have been explored in literature, for instance, by alloying nickel with copper [33,34] or by depositing nickel on titania, a reducible support [13].Here, we studied the effect of SMSI in nickel supported on niobia for the hydrogenation of carbon monoxide. For this, different reduction temperatures (250–450 °C) on NiO/Nb2O5 were used prior to H2-chemisorption, in order to determine the extent of SMSI, and prior to CO hydrogenation. H2-uptake suppression was observed when increasing the reduction temperature which is characteristic of the SMSI effect. Simultaneously, an increase in reduction temperature led to a decrease in nickel-based catalytic activity, however stable catalytic performance was gained in return with high selectivity for long-chain hydrocarbons. Ni/Nb2O5 showed higher turnover frequency and C5+ selectivity compared to nickel supported on a non-reducible support (α-Al2O3). The overall results obtained pointed out to an inhibition of nickel carbonyl formation by SMSI in Ni/Nb2O5, leading to a stable supported nickel catalyst for CO hydrogenation.Niobium oxide (Nb2O5) was used as support and obtained by crystallization of niobium oxide hydrate (Nb2O5•nH2O, HY-340, AD/4465), which was provided by Companhia Brasileira de Metalurgia e Mineração – CBMM. The crystallization was carried out in stagnant air at 600 °C during 4 h with a ramp of 5 °C min−1. The obtained Nb2O5 had a pseudo-hexagonal TT-phase, a specific surface area of 9 m2 g−1 and a specific mesopore volume of 0.05 cm3 g−1.A nickel supported on niobia catalyst was prepared using the incipient wetness impregnation method. Prior to impregnation the support (75–150 μm grains) was dried under vacuum at 80 °C for 1 h, thereafter the impregnation was performed at room temperature with a 4.2 M aqueous solution of Ni(NO3)2•6H2O (Acros, 99%) for a 6 wt.% Ni. In the next step, the catalyst was dried for 1 h at 60 °C in a fixed bed reactor under N2 flow and subsequently in the same reactor and gas flow calcined for 2 h at 350 °C (3 °C min−1). Nickel supported on α-alumina (BASF) was prepared in the same way. Metal loadings were defined as the mass of metallic Ni per gram of reduced catalyst.Temperature programmed reduction (TPR) analyses were performed using a Micromeritics Autochem 2990 instrument, where 100 mg sample was dried at 120 °C for 1 h in Ar flow followed by reduction from room temperature up to 700 or 1000 °C (5 °C min−1) in a 5 vol% H2/Ar flow. Powder X-ray diffractograms were measured using a Bruker-AXS D2 Phaser X-ray diffractometer, Co-Kα radiation (λ = 1.789 Å). Bright field transmission electron microscopy (TEM) and Scanning transmission electron microscopy (STEM-EDX) images were acquired with a Philips Tecnai-20 FEG (200 kV) microscope equipped with an energy dispersive X-ray (EDX) and high-angle annular dark-field (HAADF) detector. The reduced and subsequently passivated samples for the microscopy analysis were prepared by suspending the catalysts in 2-propanol (>99.9%, Sigma-Aldrich) using sonication and dropcasting the suspension on a carbon-coated Cu grid (200 mesh). The nickel particle size was determined using the iTEM software by analyzing at least 500 particles. Particle surface average diameters or Sauter mean (D[3,2]) were then calculated and corrected for a 2 nm NiO shell [35]. H2-chemisorption was measured on a Micromeritics ASAP 2020C using ∼100 mg of sample. Prior to the measurement, the calcined catalyst was reduced in H2 flow at different temperatures during 2 h (5 °C min−1). The sample was then evacuated, cooled to 150 °C and H2-chemisorption was measured at that temperature. Inductively coupled plasma-optical emission spectroscopy (ICP-OES) was performed on a SPECTRO ARCOS in order to establish the nickel content before and after catalysis; samples were extracted using aqua regia.Fourier-transform Infrared (FT-IR) spectroscopy measurements were carried out in a Specac “High Temperature High Pressure” transmission FT-IR cell. A self-supported catalyst wafer was prepared by applying on a sample a force of 4000 kg for 20 s, yielding a wafer of 16 mm diameter, and < 1 mm thickness. Catalyst wafers were reduced in-situ, each at different temperatures of 250, 350 and 450 °C (N2/H2 = 2 v/v; both Linde, 5.0). Subsequently, a sample was cooled down to 230 °C flushed with 5.0 purity N2 for 10 min, after which flowing CO (Linde, 5.0) was added with a ratio N2/CO = 2 v/v at 1 bar total pressure. Due to low photon-transmittance of the Nb2O5-supported Ni catalyst, 256 spectra were averaged to improve signal-to-noise. Spectra were recorded with a resolution of 4 cm−1.Catalytic performance was carried out in a quartz glass plug-flow reactor, loaded with 15–20 mg catalyst (38–150 μm) diluted with ∼200 mg SiC. Catalysts were reduced in situ at 250, 350 or 450 °C (5 °C min−1, 2 h) in an Ar/H2 = 2.0 v/v flow (GHSV = 190 000 h−1). After reduction, CO hydrogenation was performed at 230 °C, 1 bar, H2/CO = 2.0 v/v, GHSV = 28 000 h−1 and CO conversion < 5%. Reaction was carried out for 90 h. Finally, the CO flow was stopped and the H2 flow and temperature were kept for one hour in order to remove remaining hydrocarbons for further analysis of the catalyst. C1-C18 products were analyzed by online gas chromatography (Varian 430 GC, CP sil-5 column).A nickel supported on niobia catalyst was synthetized by incipient wetness impregnation method. After subsequent drying and calcination, the nickel content was determined by ICP-OES, being 5.6 ± 0.1 wt.%. Nickel supported on α-alumina was also synthetized as comparative system, with a nickel content of 5.5 ± 0.1 wt.% as determined by ICP-OES. Temperature programed reduction was carried out on the support (Nb2O5), the calcined NiO/Nb2O5 and NiO/α-Al2O3 samples. The corresponding reduction profiles for the niobia-based samples are shown in Fig. 1 . Nb2O5 showed a gradual consumption of hydrogen starting at 600 °C and a maximum consumption rate at 940 °C. The hydrogen consumption was assigned to the reduction of the Nb2O5 surface to NbO2, along with the change in color of the sample to deep indigo, characteristic of Nb4+ ions [36,37]. The reduction profile for the NiO/Nb2O5 sample showed a small hydrogen consumption signal at 200 °C which might be attributed to the reduction of Ni3+. The main hydrogen consumption between 230 and 430 °C was attributed to the reduction of NiO to metallic Ni [38,39]. Hydrogen consumption continued above 700 °C related to reduction of Nb2O5, catalyzed by the metallic nickel [40]. The consumption peak at 780 °C might correspond to the initial reduction of the support surface (Nb2O5 to NbO2) and further consumption above 810 °C to the reduction of bulk Nb2O5 and possibly subsequent reduction of NbO2 to Nb2O3. Nickel oxide supported on α-Al2O3 showed a similar reduction profile to the niobia-based sample (Figure S1), with a small hydrogen consumption signal at 220 °C ascribed to Ni3+ reduction and a main signal between 250 and 450 °C for NiO reduction to Ni. For both Nb2O5- and α-Al2O3-supported samples, the total hydrogen consumption below 450 °C corresponded to the complete reduction of all nickel oxide to metallic nickel.Based on the NiO/Nb2O5 TPR profile, four different reduction temperatures, namely 250, 350 and 450 °C, were chosen to study their effect on CO hydrogenation. The degree of reduction for the low temperatures (250 and 350 °C) was calculated by measuring TPR of NiO/Nb2O5 with an additional dwell step of 2 h at 250 or at 350 °C. The isothermal step at 250 °C resulted in two main distinctive signals in hydrogen uptake for the reduction of nickel oxide (Figure S2, A). The first one was observed by reaching 250 °C with a sharp increase in hydrogen uptake which gradually decreased back to the baseline throughout the 2 h at 250 °C. Based on the hydrogen uptake the degree of reduction at this temperature was 58%. The second main hydrogen uptake signal was observed after the isothermal step with a maximum at 360 °C (t = 200 min), this indicates that temperatures higher than 250 °C are necessary to completely reduce the nickel oxide to metallic nickel. The hydrogen uptake observed at 360 °C might relate to the observed shoulder at the same temperature in Fig. 1, which might correspond to the reduction of nickel oxide species with a stronger interaction with the support. The TPR profile with an isotherm step at 350 °C (Figure S2, B) showed a main hydrogen uptake signal which corresponded to a degree of reduction of 96%, indicating that most of the nickel oxide is reduced to metallic nickel at 350 °C. Table 1 shows the hydrogen uptake determined by H2-chemisorption for Ni/Nb2O5 and Ni/α-Al2O3 after reduction at different temperatures. An increase of the reduction temperature resulted in a decrease in hydrogen uptake for Ni/Nb2O5, resulting in an apparent increase in the derived particle size. This suppression of hydrogen chemisorption by reducible oxidic supports, the so called strong metal-support interaction (SMSI) effect, is a well-documented phenomenon attributed to coverage of the metal nanoparticles by suboxides from the support upon reductive conditions [6,8]. The degree of coverage by the suboxides is a temperature-dependent phenomenon, in which higher reduction temperatures enhance the mobility of these species and coverage of the nanoparticles [41]. Powder X-ray diffraction (Figure S3) neither showed the formation of new crystalline species (e.g. nickel niobates) nor provided indications of SMSI after reduction of Ni/Nb2O5. Reduction at 250 °C showed a substantial hydrogen uptake even though nickel oxide was not completely reduced at 250 °C as shown by the TPR results, indicating that most of the particles’ surface consisted of metallic nickel. The Ni/α-Al2O3 sample showed also a decreased in the hydrogen uptake upon increasing the reduction temperature, however this decrease was not as severe as the one observed for Ni/Nb2O5. After the chemisorption measurement and exposure to air at room temperature, the samples were analyzed by TEM (Fig. 2 and Figure S4). TEM images showed for all Ni/Nb2O5 samples a uniform distribution of nickel nanoparticles over the niobia. Furthermore, a similar nickel particle size (∼ 12 nm) was determined based on TEM as shown in Table 1, indicating no significant effect of the reduction temperature on the nickel particle size and confirming that the suppressed hydrogen chemisorption results related to the SMSI effect. In the case of the Ni/α-Al2O3 sample, TEM images (Figure S4) revealed a slight increase in particle size upon increasing the reduction temperature, in line with the results obtained from hydrogen chemisorption. The discrepancy observed here between the experimental and theoretical H2-uptake can be explained by the more significant impact of larger particles when determining the D[3,2] value, a surface-based diameter. Since a considerable amount of very small nanoparticles would not be detected by TEM.The catalytic performance of the Ni/Nb2O5 and Ni/α-Al2O3 catalysts was evaluated by varying the reduction temperatures similar to the H2-chemisorption experiments. The results are shown in Fig. 3 where nickel-normalized catalytic activity (Nickel Time Yield, NTY) is plotted against time-on-stream (TOS) up to 90 h, furthermore a summary of the catalytic performance is shown in Table 2 . The initial NTY (TOS = 0) showed consistency with the H2-chemisorption results, i.e. reduction at low temperatures for the niobia-supported sample displayed high H2-uptake along with markedly high initial NTY whereas an increase of the reduction temperature led to a suppression of the H2-uptake and a decrease in the initial NTY. However, the decrease in initial NTY is not proportional to the decrease in H2-uptake for unknown reasons. The stability throughout time significantly varied for each reduction temperature. Reduction at 250 °C led to severe deactivation, down to 70% loss in NTY at TOS = 90 h. A less pronounced deactivation was observed when the reduction temperature was increased to 350 °C with only 40% loss in NTY, however the catalyst did not reach steady state during the experiment due to continuous deactivation. In stark contrast, reduction at 450 °C showed a catalytic performance, with a low initial NTY which increased during the first 30 h of the reaction followed by a stable conversion until the end of the experiment. This might indicate a partial recover of the available metallic surface area during reaction conditions, which has been associated in literature to re-oxidation of the suboxides (e.g. NbOx) by water produced during reaction, hence modifying the SMSI effect [42]. Contrary to the Ni/Nb2O5 sample, the reduction temperature had a minor effect on the catalytic performance of the Ni/α-Al2O3 as shown in Fig. 3. Reduction at 350 °C led to a small increase in NTY than when reduced at 450 °C at the beginning of the experiment, this difference originated from their different initial particle size as revealed by their same initial turnover frequencies (TOF). Their prevalent decrease in NTY during the experiment resulted in almost similar NTY values at TOS = 90 h. The niobia-supported sample showed independently of the reduction temperature higher NTY values than the alumina-supported sample at the end of the experiment. The nickel content was determined after catalysis by ICP-OES showing no metal loss during the experiment for all samples.The promotional effect of niobia was maintained in all cases as shown by the TOFs compared to Ni/α-Al2O3, determined either by particle size distribution from TEM or H2-chemisorption (Table 2). Initial TOFs based on TEM particle size distributions showed the highest values for Ni/Nb2O5 reduced at low temperatures (250 and 350 °C) and decreased when increasing the reduction temperature to 450 °C. The inverse trend was observed for the apparent initial TOFs based on chemisorption results (TOFapp, Table 2); an increase in reduction temperature led to a substantial increase in TOFapp due to the hydrogen chemisorption suppression by SMSI (vide supra). However, the nickel surface under reaction conditions is expected to change and therefore these TOFs are indicated as ‘apparent’. Interestingly, Ni/Nb2O5 reduced at 250 °C shows consistent values for both initial TOFs indicating that coverage of the nickel nanoparticles by suboxides from the support has not taken place at this temperature. Consequently, the resulting high TOF might originate from the interphase of the nickel nanoparticles and the support. For Ni/α-Al2O3, the apparent TOFs and TOFs based on TEM have the same values and SMSI does not play a role in this catalyst system.The change in reduction temperature additionally influenced the selectivity of the nioba-supported catalyst, as shown with the Anderson–Schulz–Flory (ASF) product distribution plot in Fig. 4 and in Table 2. For all reduction temperatures, niobia-supported catalysts showed higher selectivity towards long-chain hydrocarbons when compared to the alumina-supported catalyst. At TOS = 90 h, the catalyst reduced at 350 °C had the highest α value, with the highest selectivity to C5+ products, followed by the reduction temperature at 250 °C. Reduction at 450 °C led to a shift in product distribution to shorter hydrocarbons and therefore a smaller α value. Suppressed C2H4/C2H6 was observed for the catalyst reduced at 250 or 350 °C, which has been attributed in the case of cobalt-based catalysts to re-adsorption of olefins to the metal surface to further increase chain-growth [43,44]. However, reduction at 450 °C did not show this behavior, instead a slight increase in the olefin selectivity was observed, as shown in Table 2. Re-adsorption of olefins could be hindered on the metal surface, shifting the selectivity to shorter hydrocarbons. On the other hand, Ni/α-Al2O3 showed the lowest α value, a high selectivity for methane and to a lesser extent for C2 to C10 products for both reduction temperatures. The formation of C2+ products in this case might be due to some small nickel metal nanoparticles (< 3 nm) found in this catalyst (Figure S4), which agrees with previous research reports [45].The Ni/Nb2O5 samples after catalysis were analyzed by TEM and the results are shown in Fig. 2. Significant changes for the nickel nanoparticles were observed for the spent catalyst reduced at 250 °C: broadened particle size distribution and increased average particle size (12 nm to 27 nm) were observed. Likewise, large particles were observed for the spent catalyst after reduction at 350 °C leading to a particle mean size of 18 nm. The observed nickel particle growth agreed with the stability of the catalyst during reaction; where reduction at 250 °C led to the severest deactivation and the most pronounced particle growth, increase of the reduction temperature to 350 °C attenuated the particle growth and diminished the deactivation rate. The resulting particle sizes and catalytic activity at TOS = 90 h led to similar TOFs for both reduction temperatures when compared to initial TOFs (Table 2). This is an indication that the decrease in NTY was mainly due to particle growth. The slight decrease in TOF might relate to carbon deposition over the nickel surface. In a similar way, TEM of the spent Ni/α-Al2O3 catalysts revealed a substantial increase in nickel particle size (Figure S5), indicating that the decrease in NTY originated mainly from particle growth. Particle growth for nickel-based catalysts under these reaction conditions most likely occurs via Ostwald ripening by the formation of Ni(CO)4 [25–27]. In contrast, the nickel particles remained well distributed over the support for the Ni/Nb2O5 sample reduced at 450 °C. No significant change in particle size was observed after catalysis, with a final mean particle size of 12 nm. Therefore, the increase in CO conversion during the first hours of the experiment means that sites more active became available and thus the TOF almost doubled (Table 2). These results suggest that SMSI inhibited the formation of Ni(CO)4 on a Nb2O5 support leading to a more stable catalyst. Ni(CO)4 formation rate has been reported to depend on the nickel surface morphology with particularly low coordinated Ni atoms readily reacting to form carbonyls [46,47], thus NbOx species might be responsible for blocking or modifying the electron density of these sites.Fourier-Transform Infrared (FT-IR) spectroscopy was used to study the differences for the sample reduced at 250, 350 and 450 °C in their tendencies to form nickel carbonyl. In-situ reduction of the wafer was carried out at these three different temperatures. Thereafter FT-IR spectra were recorded at 230 °C under atmospheric pressure of CO/N2 = 2 v/v flow (Fig. 5 ). Interestingly, a pronounced sharp band at 2080 cm−1 can be observed when the sample was reduced at 250 °C. This band is ascribed to subcarbonyl Ni(CO)x (x = 2, 3) species, precursors of Ni(CO)4, in accordance with literature [47–49]. These species were also detected for the catalyst reduced at 350 °C and in both cases the band disappeared after flushing with N2. In contrast, reduction at 450 °C did not give rise to this subcarbonyl Ni(CO)x band. These results show that there is a reduction temperature dependency in nickel carbonyl formation. Lower reduction temperatures thus most likely led to rapid deactivation during CO hydrogenation due to Ni particle growth via the formation and diffusion of Ni(CO)4 originated from the detected Ni(CO)x species. High reduction temperature showed stable catalytic activity (Fig. 3), suggesting that the SMSI effect suppressed the formation of Ni(CO)x species and hence Ni(CO)4, avoiding the diffusion of nickel over the support. Furthermore, the FT-IR spectra plotted in Fig. 5 show that the degree of CO activation in the adsorbed state is affected by the reduction temperature. That is, a reduction at 250 °C shows a large contribution of Ni(CO)x at 2090 cm−1 corresponding to the strongest carbon-oxygen bond based on the relatively high wavenumber of this band. A band at 1580 cm-1 is also observed which is attributed to carboxylate-type species, these may have originated from the oxidation of CO by the remaining NiO in this sample as shown by the TPR results. For a slightly higher reduction temperature (350 °C) besides the band at 2090 cm−1, a broad peak at around 1960 cm−1 is observed, which is ascribed to CO adsorbed in a 2-fold bridge position [50,51]. At the highest reduction temperature (i.e., 450 °C) a small peak at ∼ 1266 cm−1 appears, which can be ascribed to the weakest CO bond [52], or lowest wavenumber observed in this set of experiments.Two effects of the SMSI can explain the inhibition of nickel carbonyl formation. On one hand, NbOx suboxides might physically block the more reactive low-coordinated Ni atoms on the surface of the nanoparticles, preventing the formation of subcarbonyl Ni(CO)x species. A similar effect has been shown in literature by addition of alkali metals or sulfur to nickel-based catalysts [53,54]. On the other hand, the suboxides partially covering the nickel nanoparticle’s surface are capable of transferring electrons to the nickel [55–57]. In this case, Nb4+ or Nb3+ in the NbOx suboxides might transfer electron density to the metallic nickel, resulting in electron-rich Niδ− atoms at the surface. Upon CO chemisorption at the nickel surface, Niδ− increases the back-donation to the CO 2π* antibonding orbital weakening the C–O bond, as suggested by FT-IR, and thus avoiding the formation of Ni(CO)x species. Furthermore, an electron-rich metallic surface could hinder the re-adsorption of electron-rich molecules, explaining the increased olefin to paraffin ratios when the catalyst was reduced at high temperatures.The effect of different reduction temperatures was studied for nickel nanoparticles supported on niobia. An increase of the reduction temperature led to H2-chemisorption suppression, a typical phenomenon caused by reducible oxidic supports in which suboxides from the support cover partially the metal nanoparticles. The initial nickel-based catalytic activity was in line with the chemisorption results where high H2-uptake corresponded to high initial CO conversion. However, low reduction temperatures turned into a fast deactivation due to nickel particle growth as shown by TEM, whereas a high reduction temperature led to stable catalytic performance and no significant particle growth. Interestingly, reduction of the niobia-supported catalyst at high temperature brought about an activation period during the first hours under reaction conditions followed by stable nickel-based activity. FT-IR measurements of CO adsorbed on Ni/Nb2O5 showed that nickel subcarbonyls readily formed after low but not after high reduction temperature. This could explain the particle growth involving the formation and diffusion of nickel tetracarbonyl, which formed from the detected nickel subcarbonyls. The inhibition of nickel tetracarbonyl formation after high temperature reduction is associated to the presence of suboxide species over the nickel surface, by either physically blocking exposed low-coordination nickel atoms, or by enhancing the electron density on the nickel surface and facilitating C–O bond rupture instead of nickel tetracarbonyl formation. The reduction treatment had a strong influence in the product distribution, where the highest selectivity towards C5+ was obtained after reduction at 350 °C, while a further increase of the reduction temperature shifted the product distribution towards lighter products. Finally, the promotional effect of reducible oxides, such as niobia, in CO hydrogenation was clearly shown since independently of the reduction temperature nickel supported on niobia showed higher nickel-based activity, TOF and C5+ selectivity compared to nickel supported on α-alumina, a non-reducible support. We have shown that niobia used as support material offers the possibility to make stable nickel-based catalysts for CO hydrogenation with tunable product spectrum. Companhia Brasileira de Metalurgia e Mineração (CBMM) is thanked for financial support of this research. Dr. Robson Monteiro and Mr. Rogério Ribas (CBMM) are acknowledged for useful discussions and supplying the niobia support. Mr. Wouter Lamme (Utrecht University, UU), Mrs. Petra Keijzer (UU) and Mrs. Savannah Turner (UU) are acknowledged for performing TEM measurements. KPdJ acknowledges the European Research Council (ERC) for a EU FP7 ERC Advanced Grant no. 338,846.Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.cattod.2018.11.036.The following is Supplementary data to this article:

Stability of metal nanoparticles under reaction conditions is crucial in many catalytic processes. Nickel-based catalysts often encounter severe particle growth in the presence of carbon monoxide due to the formation and migration of nickel carbonyl. In this research, we showed that the reduction temperature of nickel oxide supported on niobia (Nb2O5) influenced the stability of the resulting nickel catalyst during subsequent carbon monoxide hydrogenation. Low reduction temperatures resulted in high initial nickel-normalized activity towards long-chain hydrocarbons (C5+), but fast deactivation throughout the experiment. High reduction temperatures led to a shift in product distribution towards shorter hydrocarbons and a decreased initial nickel-normalized activity, while during the first hours of the experiment an increase in turnover frequency and nickel-normalized activity was observed, resulting eventually in a stable catalytic performance. Electron microscopy analysis revealed extensive particle growth after catalysis when the catalyst had been reduced at low temperatures and no significant changes in particle size when reduced at high temperatures. By use of in-situ FT-IR spectroscopy, nickel subcarbonyl species which are precursors of volatile nickel tetracarbonyl were detected on Ni/Nb2O5 after low temperature reduction and exposure to CO, but not after high temperature reduction. Hence, particle growth is explained by the formation and diffusion of nickel carbonyl and subsequent Ostwald ripening, that leads to larger nickel particles with concomitant decrease in nickel-normalized activity. The stability of the catalyst reduced at high temperature was linked to the formation of niobium suboxides and their partial coverage of the nickel particles limiting the formation of nickel carbonyl and slowing down particle growth.

Owing to paucity of fossil fuels, it is critical to transform renewable sources into value added compounds and fuel additives [1]. Biodiesel manufacturing has sparked a lot of interest in this direction. Transesterification of vegetable oil and animal fat yields biodiesel. Glycerol is a 10% byproduct of the overall biodiesel synthesis process [2]. Biodiesel production is estimated to hit 41.4 billion litres in 2025, while glycerol production will hit 4.14 billion litres [3]. Glycerol conversion to various chemical such as glycerol carbonate, acrolein, glyceric acid, mono and di-glycerid, propylene glycol (1,2-PDO) and trimethylene glycol has been reported in recent literature [4]. Furthermore, selective transformation of biomass derived glycerine to 1,2-PDO by hydrogenolysis has gain noteworthy importance because of great marketable value of 1,2-PDO [5]. Also, plastics, polymers, agriculture adjuvants, solvents, the tobacco industry, detergents, and numerous functional fluids have all employed 1,2-PDO as a basic ingredient [6]. Presently, propylene glycol is formed from fossil fuel resources through hydration of 1,2-Epoxypropane via chlorohydrin, hydroperoxide processes and ethane epoxidation process [7,8]. As a result, making propylene glycol from bio-glycerol is an environmentally friendly procedure [9].Several noble metal and transition metal catalyst were synthesized, and their performance has been documented in literature [5,10,11]. Catalysts made of noble metals have been reported to be effective in this process [12]. Nevertheless, the generation of degradation products by means of excessive hydrogenolysis is one of the primary drawbacks of noble metal catalyst [13]. Because of their high activity and selectivity for the breakdown of the C–O bond, Cu-based catalysts supported on oxides with acidic sites were studied for hydrogenolysis of glycerol [4,14,15]. As previously stated, hydrogenolysis of glycerol to propylene glycol is a multistep reaction involving dehydration of glycerol to hydroxyacetone and then hydrogenation of hydroxyacetone to propylene glycol [16]. However, it is observed that all previous investigations reported in literature were mainly dedicated to catalyst synthesis and characterizations along with experimental reaction parameter optimization. Only a few studies have been conducted on kinetic model construction and kinetic parameter estimates for this reaction [15–21].Lahr et al. [17] investigated the hydrogenolysis of glycerol on a 5 ​mol% Ru/C catalyst and suggested a Langmuir-Hinshelwood-Hougen-Watson (LHHW) rate model that took into account the 1,2-PDO and EG formation in addition to the impact of competitive adsorption of above compound and glycerine over catalyst's site. Torres et al., described power-law model with activation energy of 54.19 ​kJ/mol [18]. The kinetic study over LDH catalyst showed energy of activation as 65.5 ​kJ/mol with power law model [9]. Sharma et al. [19] investigated the kinetics of glycerine hydrogenolysis in the presence of a Cu:Zn:Cr:Zr catalyst. The Langmuir-Hinshelwood model was modified to pseudo first order reaction by assuming the adsorption of glycerol and hydrogen did not inhibit the reaction. The calculated energy of activation was stated as 31.7 ​kcal ​mol−1. The kinetics of hydrogenolysis of glycerol using a Ni–Cu/Al2O3 material were described by Gandarias et al. [20], and LHHW model was built to define direct transformation of glycerine into 1,2- PDO, which then hydrogenolysis to 1-propanol. Kinetics of glycerol over Raney Ni catalyst achieved reaction activation energy of 60 ​kJ/mol using LHHW model [21]. A Cu–ZnO–Al2O3 catalyst with a 57.8 ​kJ/mol activation energy was also employed for the hydrogenolysis of glycerol [22]. Dossin el.al., reported a kinetic study of Eley-Rideal (ER) type model for transesterification reaction using batch slurry-reactor [23].The aim of present study was to develop a suitable reaction kinetic model for hydrogenolysis of glycerol in presence of bi-functional layered double hydroxide (LDH) catalyst. The modified power law and Eley-Rideal type model were developed to fit obtained experimental outcomes. Surface adsorption, reaction, and desorption stages of reactants and products molecules on numerous active sites present on surface of catalyst were studied in the model development. The obtained equations were solved in MATLAB by means of ode45 and ode23s. Finally, the best values for the reaction kinetic parameters were found by minimizing the residual sum of squares between the experimental and model predicted values for the experiments that were carried out. The results showed that the ER model was able to successfully relate the model anticipated and experimental reactant and product amounts.The data required for the kinetic study of glycerol to 1,2- PDO using bi-functional layered double hydroxide (LDH) catalyst are generated by performing the experiments using an autoclave reactor (Amar equipment). All the details such as catalyst preparation, catalyst activity test and changes with reaction conditions on selectivity and conversion are already published in our previous work [24,25].It is essential to grasp best promising reaction transformations for glycerol hydrogenolysis over the Cu0.45Zn0.15Mg5.4Al2O9 catalyst. The reaction mechanisms for glycerol conversion to 1,2-PDO was reported in previous studies [15,26–28]. There are two main reaction pathways for glycerine hydrogenolysis to propylene glycol over noble and transition metal catalysts in the literature [4]. In a two-step process, dehydration of glycerol over the acidic/basic sites of the catalyst produces metastable acetol as an intermediate product, which is followed by hydrogen addition to acetol over the active metallic sites of the catalyst to produce 1,2-PDO [9,13].The obtained predominant reaction product in this investigation was 1,2-PDO. EG was also been formed as a second major product obtained with traces of undesirable compounds such as methyl alcohol, ethyl alcohol, hydroxyacetone, 1- PO, 2- PO. So as to find out reaction path, intermediate and final reaction products i.e. 1,2-PDO, EG and acetol were taken as the feed instead of glycerol and reaction was carried out under optimized reaction conditions (210 ​°C temperature, 4.5 ​MPa pressure, 800 ​rpm, 20 ​wt% of 100 ​g glycerol solution, 1.6 ​g catalyst). The obtained experimental data for the reaction mechanism are shown in Table 1 .Glycerol as reactant achieved 100% conversion with ∼94% selectivity to1,2-PDO. While acetol achieved 99.7% conversion with ∼94% selectivity to 1,2- PDO. Some traces (<5%) of over hydrogenolysis products were also obtained. The conversion of 1,2-PDO was less than 4%, and the reaction products were methyl alcohol, ethyl alcohol, 1-PO, and 2-PO. In the case of EG as feed, the conversion was around 8%, and the end reaction products were methyl alcohol and ethyl alcohol. The obtained results from the experiments are shown in Table 1. On the basis of results obtained from reaction mechanism experiments, the plausible scheme of chemical transformations is shown in Fig. 1 . Similar kind of reaction pathway was reported for other non-noble metal catalyst [27].The experimental data utilized are shown in Fig. S1. The detailed information about influence of intra particle diffusion and external mass-transfer were explained in supporting information (Fig. S2). Reaction were elementary and reaction rate will depend upon the concentration of the reactant of the corresponding equation.The reaction can be written as: - (1) Glycerol ​ → catalyst  ​ 1 , 2 - PDO (2) Glycerol ​ → catalyst  ​EG (3) EG ​ → catalyst  ​Ethanol Rate equations for reaction can be written as: -For Glycerol, (4) - dC G dt  ​ =  ​ [ wk 1 H H 2 ] C G P H 2 + [ wk 2 H H 2 ] C G P H 2 For 1,2-Propanediol, (5) - dC 1 , 2 - PDO dt  ​ =  ​ [ wk 1 H H 2 ] C G P H 2 For Ethylene glycol, (6) - dC EG dt  ​ =  ​ [ wk 2 H H 2 ] C G P H 2 - [ wk 3 H H 2 ] C EG P H 2 where, CG, C1,2-PDO, CEG are glycerol concentration, 1,2-PDO and EG correspondingly at any time ‘t’, k1, k2, k3 are specific reaction rate constant, P H 2 is hydrogen's partial pressure and HH2 is henry's constant. Also, w stands for concentration (kg/m3(liquid)) of the catalyst.For the estimation of the parameters of the reaction kinetics and predicted time dependent concentration data, rate equations were elucidated mathematically in MATLAB via ode23s by fitting the data obtained experimentally. The optimum kinetic parameter values were found by reducing the residual sum of squares among simulated and experimental glycerine amounts across experiments. Parity plots are used to compare the observed and model estimated concentration of glycerine, 1,2- PDO, and EG, as illustrated in Fig. 2 . The excellent match between the experimental and modelled concentrations was established by these data. Using the Arrhenius equation, the impact of changing the temperature of reaction on rate constant was utilized to determine the activation energy. Fig. 3 shows the plot of ln k vs 1/T. Table 2 shows the pre-exponential factor and activation energy for the production of 1,2-PDO and EG using a modified power law model. Activation energy for formation of 1,2-PDO was found to be 52.6 ​kJ/mol with pre-exponential factor 7.1 ​× ​106 ​mol/gcat.h and 58.6 ​kJ/mol and 3.2 ​× ​106 ​mol/gcat.h for the formation of EG by using modified power-law model.To calculate the preliminary reaction rate parameters, a modified power law model was employed. The power law model, on the other hand, has a significant flaw: it does not account for all of the factors that affect heterogeneous reactions, such as adsorption, surface reaction and desorption on a catalyst site. While, for heterogeneous processes, the Eley-Rideal model is a regularly used realistic way to obtain the rate expression. For solid catalyzed reactions, the Eley-Rideal model is favoured because it incorporates a rate equation derived from reaction mechanism that includes actual surface phenomena throughout reaction. This approach considers adsorption, surface reaction and desorption steps on catalyst active site when calculating reaction rate. As a result, Eley-Rideal type model was developed to better explain reaction kinetics.In this model, initially the glycerol molecule adsorbed on catalyst site undergoes dehydration to form water and acetol. In next step, the adsorbed acetol reacts with hydrogen present in the reactor to form adsorbed propylene glycol. The desorption of 1,2-PDO, acetol and water from catalyst surface take place in the final step with regeneration of the active centers.The following reaction processes were used to get the rate equations:Step 1: Adsorption of glycerol (G) on catalyst's active site ($): (7) G + $ ⇄ k − 1 k 1 G . $ Step 2: Adsorbed glycerol is dehydrated to adsorbed acetol, and then the surface interaction between adsorbed acetol (A.$) and hydrogen molecule (H2) occurs: (8) G . $ ⇄ k − 2 k 2 A . $ + W (9) A . $ + H 2 ⇄ k − 3 k 3 P . $ 1,2- PDO, acetol, and water are represented by P, A, W, correspondingly.Step 3: Desorption of adsorbed acetol and 1,2- PDO from surface of catalyst, as well as active site regeneration: (10) A . $ ⇄ k − 4 k 4 A + $ (11) P . $ ⇄ k − 5 k 5 P + $ Individual rate equations can be expressed as follows using Eqs. (7) - (11): (12) ( − r 1 ) = k 1 ( C G C $ − C G . $ K 1 ) ; K 1 = k 1 k − 1 (13) ( − r 2 ) = k 2 ( C G . $ − C A . $ C W K 2 ) ; K 2 = k 2 k − 2 (14) ( − r 3 ) = k 3 ( C A . $ P H 2 − C P . $ K 3 ) ; K 3 = k 3 k − 3 (15) ( − r 4 ) = k 4 ( C A . $ − C A C $ K 4 ) ; K 4 = k 4 k − 4 (16) ( − r 5 ) = k 5 ( C P . $ − C P C $ K 5 ) ; K 5 = k 5 k − 5 The equilibrium constants for the corresponding reactions are K1, K2, K3, K4, K5.Adsorption, surface reaction, and desorption are all rate limiting stages, hence three alternative rate equations were generated. It was expected that rate limiting step would be surface reaction. Thus, rate of surface reactions can be written as follows:The resulting rate equations are reduced as follows:From Eq. (13), (17) ( − r 2 ) = k 2 C G . $ from Eq. (14), (18) ( − r 3 ) = k 3 C A . $ P H 2 Post considering the adsorption and desorption steps as significantly faster than surface reaction [29], we have.From Eq. (12), r 1 k 1 = 0 , then (19) C G . $ = K 1  ​ C G  ​ C $ Similarly, from Eq. (15), r 4 k 4 = 0 , then (20) C A . $ = C A  ​ C $ K 4 Similarly, from Eq. (16), r 5 k 5 = 0 , then (21) C P . $ = C P  ​ C $ K 5 Substitution of Eqs. (20) and (21) into the surface reaction rate Eqs. (17) and (18) subsequently results to: (22) ( − r 2 ) = k 2 K 1 C G C $ (23) ( − r 3 ) = k 3 C A P H 2 C $ K 4 From the catalyst site's overall balance, we have (24) CT$ ​= ​C$ ​+ ​CG.$ ​+ ​CA.$ ​+ ​CP.$ When the values of adsorbed species concentrations are substituted in total site balance (Eqn. (24)) the following results are obtained: (25) C T $ = C $ + K 1 C G C $ + C A  ​ C $ K 4 + C P  ​ C $ K 5 C $ = C T $ ( 1 + K 1 C G + C A K 4 + C P K 5 ) Substituting Eq. (25) into Eqs. (21) and (23) yields (26) ( − r 2 ) = k 2 K 1 C G C T $ [ 1 + K 1 C G + C A K 4 + C P K 5 ] = k 2 ′ K 1 C G [ 1 + K 1 C G + C A K 4 + C P K 5 ] where k 2 ′ = k 2 C T $ , glycerol to acetol apparent reaction rate constant. (27) ( − r 3 ) = k 3 C A P H 2 C T $ K 4 [ 1 + K 1 C G + C A K 4 + C P K 5 ] = k 3 ′ C A P H 2 K 4 [ 1 + K 1 C G + C A K 4 + C P K 5 ] here k 3 ′ = k 3 C T $ is apparent reaction rate constant for acetol to 1,2-PDO. Eqs. (26) and (27) reflect the concluding rate expressions for two step reaction, glycerol dehydration to acetol subsequently acetol hydrogenation to 1,2-PDO.The mole balance of individual species i at any instant in time t for the jth reaction may be stated as follows to develop the model: (28) d C i d t = ∑ j n i j r j The steady state concentrations of glycerol, acetol, and 1,2-PDO, respectively, are CG, CA, and CP and nij represents stociometry coefficient of ith species in jth reaction. rj: rate of jth reaction.The set of ordinary differential equation (Eqn. (28)) produced were numerically solved using MATLAB's ode23s function coupled with genetic algorithm (GA) optimization for stiff systems to estimate unknown parameters in the rate equations. The residual sum of squares, f (fitness function), among observed 1,2-PDO, glycerol, and acetol concentrations and estimated concentration values generated from model equations was minimized using GA.As follows is the expression of the Objective function: (29) f = ∑ i = 1 N [ ( C G , exp i − C G , s i m i ) 2 + ( C A , exp i − C A , s i m i ) 2 + ( C P , exp i − C P , s i m i ) 2 ] where, N represents experimental runs and C G , exp i , C A , exp i , and C P , exp i are experimental concentration of glycerine, hydroxyacetone and 1,2-PDO, while The corresponding anticipated concentrations derived by solving the model equations are C G , s i m i , C A , s i m i and C P , s i m i .For the estimate of kinetic parameters, a simple genetic algorithm code was constructed in this work. The best answer is determined by the size of the population, the probability of the genetic operators (crossover and mutation), and the seed values used. The objective function was minimized and the kinetic parameters were estimated using a population of 1000. For crossover and mutation probability, optimal values of genetic operators of 0.9 and 0.1 were used, respectively. Using equation below, pre-exponential factors and activation energy were calculated via assessed value for rate constants and adsorption constants acquired at various temperature [22]. (30) ki ​= ​ki o exp [-Ei/(RT)] Then all equilibrium constants are described by the equation given below [22]. (31) Kj ​= ​Kj o exp[Ej/(RT)] The obtained values are summarized in the Table 3 .Parity plots are used to compare the observed and model simulated concentrations of glycerol and propylene glycol, as illustrated in Fig. 4 . Arrhenius plot to evaluate the energy of activation of adsorption of glycerol, formation of 1,2-PDO and desorption of 1,2-PDO are shown in Fig. 5 . It was discovered that activation energy for production of propylene glycol is 41.42 ​kJ/mol. The value estimated using the modified power law model (52.6 ​kJ/mol) was found to be compatible with this finding. The obtained activation was compared with previously reported studies in Table S1. The results showed that the suggested Eley-Rideal model suited the experimental and simulated data quite well. The obtained activation energy differs for both the models owing to the assumptions utilized for model development.Kinetic study was performed over Bi-functional layered double hydroxide (LDH) catalyst. The effect of different reaction parameter was utilized to obtain the kinetic parameters of the hydrogenolysis of glycerol. Based on reaction products distribution plausible reaction mechanism was proposed. Two different type of kinetic models i.e. To suit the experimental data, modified power law and Eley-Rideal were tested. To simulate experimentally obtained concentration time data, a set of differential equations was created and numerically solved using ode23s in MATLAB in conjunction with the genetic algorithm optimization tool. The kinetic parameters were obtained by minimizing the residual sum of squares between the predicted and experimental concentrations of glycerol, propylene glycol, and EG. By using modified power law model, energy of activation and pre-exponential factor were estimated as 52.6 ​kJ ​mol−1 and 7.1 ​× ​106 ​mol/gcat.h for formation of 1,2-PDO and 58.6 ​kJ/mol and 3.2 ​× ​106 ​mol/gcat.h for the formation of EG, separately. Furthermore, the kinetic parameters were determined using a more realistic Eley-Rideal model, and the activation energy for the synthesis of 1,2-PDO was compared to the modified power law model. The activation energy obtained by using the Eley-Rideal model (41.42 ​kJ/mol) for the surface reaction step of glycerol was comparable with the activation energy obtained by using the modified power law model (52.6 ​kJ/mol). The results revealed that the suggested modified power law and Eley-Rideal model successfully linked the rate data, as well as the actual and predicted concentrations of glycerol and products.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.The following is the Supplementary data to this article: Multimedia component 1 Multimedia component 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.crgsc.2022.100289.

The performance of a highly efficient Cu0.45Zn0.15Mg5.4Al2O9 catalyst was investigated using a high-pressure autoclave reactor by adjusting several reaction conditions. The involvement of intermediates (Hydroxyacetone, Propylene glycol, Ethylene glycol) was explored to better understand the reaction pathways. The hydrogenolysis reaction was shown to be a multistep process, including the dehydration of glycerine to hydroxyacetone and then the hydrogenation of hydroxyacetone to 1,2-PDO. Finally, several kinetic models were established, and experimental outcomes were fitted to these models. The reaction kinetic parameters were calculated in MATLAB using ode45 and ode23s combined with optimization strategies to solve the resultant ordinary differential equations. The modified Power law model and the Eley-Rideal (E-R) models adequately correlated the experimental outcomes for two step hydrogenolysis reaction, according to the findings. The modified power law model revealed a pseudo-first order reaction in case of glycerol, and the energy of activation was calculated as 52.6 ​kJ/mol.

Metal-air batteries, such as Zn-air batteries, Mg-air batteries, Al-air batteries are a class of safe, reliable, and efficient energy storage devices have attracted increasing attention [1]. Research has proven they have much higher theoretical energy density than that of state-of-the-art Li-ion battery by producing electric energy through a redox reaction between metal and oxygen [2,3]. Among them, Al-air batteries possess great potential for large scale application due to the high specific capacity (2.98 Ah g−1) and energy density (8100 Wh kg−1), abundant resources of aluminum, environmentally friendly nature with high recyclability etc [4,5]. However, the sluggish kinetics of oxygen reduction reaction (ORR) normally resulting in serious cathode polarization and low energy efficiency is one of the serious issues hindering the wide commercialization of Al-air batteries [6,7].Platinum (Pt) nanoparticles (NP) dispersed on active carbon materials (Pt/C) has been commonly used to effectively prompt the ORR process, however, it suffers from high cost, low utilization efficiency and poor durability [8,9]. Tremendous efforts have been devoted on the development of low or non-Pt ORR catalysts [10], and the incorporation of other transition metals was reported to simultaneously enhance the ORR activity and durability [11,12]. It has been reported that transition metal (M) such as, Ag, Pd, Cu, Fe, Ni etc. are introduced to form Pt-M alloy or bimetallic catalysts to reduce Pt loading, increase utilization efficiency, high activity and stability. Among the transition metals, gold (Au) is a special candidate in view of its higher oxidation potential than Pt, which encourage the combination of Au and Pt to be a stable catalyst [13,14].The alloying of Pt with Au is a direct way of incorporation, which was reported to exert apparent effect on the electronic structure owing to the strong coupling between Pt and Au atoms, resulted in attractive ORR catalytic activity [12,15-17]. The problem is that Pt and Au are not always miscible in a whole range of concentrations and phase segregation can be expected, which influence the stability of catalyst seriously [18]. As an alternative choice, forming core (Au) -shell (Pt) structure has been reported to suppress the degradations of Pt nanoparticles (NPs) by up-shifting the dissolution potential of Pt and thereby pledging good long-term stability [19-24]. Shi et al. prepared Au-Pt core–shell catalyst in size of 30 ~ 75 nm aided by ionic liquid, which effectively improved the ORR catalytic activity and stability in comparison to the Pt/C catalyst, because of the high utilization of Pt and the protection of Pt active sites by Au [23]. Further increasing the stability of Au-Pt core–shell can be achieved by doped Au core with titanium oxide at vertex and edges, which restricted to much Au segregation on to the Pt at surface facets, as reported by Hu et al [21].In a reverse way to fabricate Au-Pt core–shell catalyst, decorating Pt surface with Au atoms to protect the vulnerable sites at edges and corners was also reported [25-27]. Kodama et al. deposited Au atoms on step sites of Pt single-crystal surface, which raised the ORR activity by 70% and also improved the durability of Pt [27]. Moreover, Takahashi et al., modified the edges and corners of Pt nanoparticles with arc-plasma deposition, the stability as well as the activity of Pt catalysts was improved significantly [25,26]. With this physical vapour deposition technique, the deposited amounts and deposition site of Au on Pt catalysts is easy to control.Recent attempts are addressed to develop bimetallic AuPt nanosize catalysts, which has been proved that Au clusters confer stability by raising the Pt oxidation potential and stabilizing Pt against dissolution under harsh work environment [28,29]. Zheng et al. synthesized smaller AuPt NPs (d ≈ 5 nm) in form of popcorn-like aggregates clusters (in size of ca. 36 nm), which only exhibited better ORR catalytic activity than 10 wt% Pt/C, poorer than 20 wt% Pt/C possibly due to the aggregated structure [29]. It is known that increase the particles size can improve the stability, but sacrifice the activate surface area therefore the catalytic activity. Further investigation to synthesis AuPt catalysts with low metal loading, high activity and stability still need further investigation.Following this context, we synthesized a series of AuxPt/MWNTs catalysts (x = 0.25, 0.67, 1.68 and 4.55 of atom ratio) by a simple one-pot reduction of chloroauric acid and chloroplatinic acid with the tris(hydroxylmethyl)phosphine oxide (THPO) in presence of MWNTs. The synthesized AuxPt NPs are highly dispersed with an average diameter of ca. 3.0 nm. The Au0.67Pt/MWNTs catalyst with metal loading of 10.2 wt% (Au:4.1 wt%, Pt:6.1 wt%) exhibited a competitive ORR catalytic activity and durability to 20 wt% Pt/C catalyst. The Au1.68Pt/MWNTs by properly increasing Au loading to 8.95 wt% (Pt:5.3 wt%) as the Al-air battery cathode showed larger capacity and power density, superior durability than 20 wt% Pt/C cathode.Commercial platinum catalyst (Pt/C, 20 wt% and 10 wt%, Alfa Aesar), chloroauric acid (HAuCl4, 99.999%, Shanghai Titan Scientific Co. Ltd., China), tetrakis(hydroxylmethyl) phosphonium chloride (THPC, 80% aqueous solution, Sigma-Aldrich), Nafion solution (5 wt%, Sigma-Aldrich) and multi-walled carbon nanotubes (MWNTs, diameter = 10 ~ 20 nm, length = 10 ~ 30 mm) were used directly without further treatment. Chloroplatinic acid (H2PtCl6, 37%), potassium hydroxide (KOH, AR), sodium hydroxide (NaOH, AR) and hydrogen peroxide (H2O2, AR, 30% aqueous solution) are purchased from Sinopharm Chemical Reagent Co., Ltd.The MWNTs was pretreated using a moderate surface oxidation to increase water affinity [13,30]. Typically, 200 mg of MWNTs was added in a gas-proof Erlenmeyer flask with a separating funnel in connection to a vacuum pump. The flask was vacuumed to a pressure of 0.01 MPa for 10 min, then 40 mL of deionized water and H2O2 mixture was added. The suspension was then sonicated for 10 min followed mixing for 2 h using magnetic stirrer, then kept still overnight. The pre-treated MWNTs were separated from the suspension by centrifuging at 8000 rpm, then dried in an oven at 80 °C overnight.43 mg of the treated MWNTs were added into 95 mL of deionized water at 75 °C and mixed using ultrasonic for 5 mins, and then 1 mL of 24.3 mM HAuCl4 and 1.25 mL of 20 mM H2PtCl6 were added, followed by addition of 600 μL of 1 M NaOH and 2 mL of 50 mM THPC. It was kept stirring for further 3 h at 75 °C to make uniform suspension, and then transferred to ice bath and stood overnight. The product was rinsed with deionized water till pH neutral, and using a freeze-dryer. the obtained hybrid was named as Au1.68Pt/MWNTs. Three other hybrids were synthesized using the same procedure with different volumes content of HAuCl4 and H2PtCl6 solution, specifically, with 0.333 mL of HAuCl4 and 2.083 mL of H2PtCl6, 0.5 mL of HAuCl4 and 1.875 mL of H2PtCl6, 1.5 mL of HAuCl4 and 0.625 mL of H2PtCl6, the catalysts are marked as Au0.25Pt/MWNTs, Au0.67Pt/MWNTs and Au4.55Pt/MWNTs, respectively.Catalyst morphology and elemental analyses were carried out using a spherical aberration corrected field emission transmission electron microscope (TEM, Titan G2 60–300) operated at 200 kV. The structure of the catalysts was characterized by an X-ray diffractometer (XRD, PANalytical) equipped with Cu kα radiation. The chemical component of the catalysts was investigated using an X-ray photoelectron spectroscope (XPS, ESCALAB 250Xi, Thermo Fisher Scientific) using Al kα radiation. The metal loading in the catalysts was examined by an inductively coupled plasma mass spectrometer (ICP-MS, X-Series Ⅱ, Thermo Fisher Scientific), where the hybrids were calcinated at 400 °C for 2 h and then 500 °C for 5 h in air to burn up the MWNTs substrate, after cooling to 200 °C the residuals were treated with aqua regia. The ICP-MS technique was used to determine the metal content in the solution, where each sample was tested for three times, taking the average value as the loading amount of Au and Pt for each sample.4 mg of a AuxPt/MWNTs catalyst, 100 μL of Nafion solution, 200 μL ethanol and 800 μL deionized water are used to prepare the catalyst ink, the slurry was mixed using an ultrasonic sound bath for 30 mins. Cyclic voltammetry (CV) analysis was performed by using an electrochemical workstation (CHI660E, CH Instruments) at a scan rate of 20 mV s−1 in N2 or O2 saturated 0.1 M KOH solutions. The working, counter and reference electrodes are glassy carbon electrode (GCE, d = 4 mm, S = 0.126 cm2), platinum wire and Ag/AgCl electrode, respectively. 8 μL of the catalyst slurry was dropped on the GCE, it was dried in ambient temperature to obtain a smooth coverage on the electrode with catalyst loading of 0.23 mg cm−2. All potential values were given with the respective to reversible hydrogen electrode (RHE) scale, the potentials were converted from Ag/AgCl electrode by using φ testvsRHE = φ testvsAg / A g C l + 0.209 + 0.059 p H ,where φ testvsRHE and φ testvsAg / A g C l is the testing potential verse RHE and Ag/AgCl reference electrode, respectively, 0.209 is the standard potential of Ag/AgCl electrode. The relationship between φ RHE and pH are showed in Fig.S1.The ORR kinetics of the AuxPt/MWNTs hybrid was examined using linear scan voltammetry (LSV) method. ORR kinetics was investigated using rotating disk electrode (RDE) and rotating ring-disk electrode (RRDE) technologies in O2 saturated 0.1 M KOH. The rotating speed were 400 rpm, 625 rpm, 900 rpm, 1225 rpm and 1600 rpm with the scan rate at 5 mV s−1 for RDE. The RRDE equipped with a glassy carbon disk electrode (d = 4 mm, S = 0.126 cm2) and a Pt ring electrode (S = 0.189 cm2) and it was performed at scan rate of 1600 rpm only. In the experiment, the disk potential scanned from 1.0 to 0.2 V at a rate of 5 mV s−1, and the ring potential was fixed at 1.8 V. Prior to testing, 5 μL and 8 μL of the catalyst slurry were coated and dried on the RDE and RRDE with catalyst loading of 0.25 and 0.23 mg cm−2, respectively.Al-air battery performance were measured in a homemade testing cell fabricated with Al foil anode (99.99%, 4.5 cm2), 4 M KOH electrolyte and air cathode. The air cathode comprises of a current collector (Ni foam) and a carbon paper (1 cm2) coated with 2 mg catalyst layer. The discharge polarization curves were carried out at 1 mV s−1 between the potential widow of 1.8–0 V vs. Al. the specific capacity was recorded at 100 mA cm−2, the dynamic galvanostatic measurement were performed between 1 mA cm−2 and 200 mA cm−2, the durability of air electrode was tested by discharging five cycles by replacing Al anode and electrolyte after each discharge.As shown in Table 1 , the overall Au and Pt loading amounts of the Au4.55Pt/MWNTs, Au1.68Pt/MWNTs, Au0.67Pt/MWNTs and Au0.25Pt/MWNTs catalysts are measured as 15.05 wt%, 14.25 wt%, 10.2 wt% and 9.5 wt% by the ICP-MS analysis, from which the exact Au/Pt ratios of the catalyst are determined as 4.55, 1.68, 0.67 and 0.25, respectively. Fig. 1 a1, b1 and c1 illustrate the TEM images of Au4.55Pt/MWNTs, Au1.68Pt/MWNTs and Au0.67Pt/MWNTs catalysts, respectively, along with the corresponding ones in higher magnification shown in Fig.1 a2, b2 and c2. All AuxPt NPs are deposited uniformly on the MWNTs substrates. The average particle size of the Au4.55Pt on MWNTs is measured in the picture as 3.02 nm, the Au1.68Pt as 2.98 nm, and the Au0.67Pt/MWNTs as 2.96 nm, suggesting that the variation of Au/Pt ratios did not influence much on the size of the bimetallic AuxPt NPs. In the high-resolution TEM (HRTEM) images in Fig. 1 a3, b3 and c3, two d-spacing values of 0.236 nm and 0.225 nm are measured, assigned to the Au (1 1 1) and Pt (1 1 1) facets, respectively [31]. The high-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) images in the Fig. 1 a4, b4 and c4 display that the Au and Pt NPs stay overlapped but do not grow together. The energy dispersive X-ray (EDX) analyses in Fig. 1 a5, b5 and c5 also display that Au (green color) and Pt (red color) NPs are very close to each other, implying the possible interaction between the Au and Pt NPs.XRD patterns of three AuxPt/MWNTs catalysts in Fig. 2 present the same feature. The peak at 26.6° can be assigned to the (0 0 2) facet of MWNTs (JCPDF No. 25–0284), and the peaks at 38.1°, 44.3°, 64.5° and 77.5° are corresponding to (1 1 1), (2 0 0), (2 2 0) and (3 1 1) facets of Au (JCPDF No. 04–0784). The peaks at 39.5°, 45.9°, 67.0° are attributed to (1 1 1), (2 0 0) and (2 2 0) facets of Pt (JCPDF No. 87–0636), where the diffractions of Pt are enhanced with the decrease of the Au/Pt ratio. No signal assigned to AuPt alloy is seen in the XRD patterns. Fig. 3 displays the XPS analysis results for AuxPt/MWNTs catalysts. The high-resolution signals of C1s for the three hybrids (column a) are fitted with three peaks at 284.8 eV, 285.2 eV and 286 eV in correspondence to C-H/C-H, C-P-O and C-OH groups, respectively[32]. The P2p signals (column b) present two fitting peaks at 133.8 eV and 134.7 eV in each curve, assigned to P2p3/2 and P2p1/2 groups [32]. The resolved Au4f signals (column c) manifest a doublet at 84.3 eV and 87.9 eV, attributed to the 4f7/2 and 4f5/2 of metallic Au [13]. The Pt4f signal (column d) can be fitted into a doublet at 71.4 eV and 74.7 eV associated with metallic Pt, and the peaks at 72.5 eV and 75.8 eV are attributed to the divalent state of Pt (Pt2+) [23,24]. The results indicate of triphenylphosphine oxide (THPO) as the capping molecule on the AuxPt NPs, which is normally generated from the cleavage of THPC in alkaline solutions [30].The CV plots of the AuxPt/MWNTs catalysts and 20 wt% Pt/C catalyst are recorded in both O2 and N2 saturated 0.1 M KOH ranging from 1.2 V to 0 V at a scanning rate of 20 mV s−1. In hydrogen underpotential deposition (HUPD) region, peaks observed between 0 V and 0.4 V attributed to hydrogen adsorption and desorption. For the Au4.55Pt/MWNTs hybrid, Fig. 4 a exhibits an oxygen reduction peak at 0.86 V with the current density of 0.86 mA cm−2 in O2 saturated 0.1 M KOH solution, in contrast to the curve in N2 saturated electrolyte. With the decrease of the Au/Pt ratio, the reduction peak potential of the Au1.68Pt/MWNTs and Au0.67Pt/MWNTs catalysts shifts positively to 0.866 V and 0.87 V, respectively, along with larger peak current densities of 0.93 mA cm−2 and 0.96 mA cm−2. Further decreasing the Au/Pt ratio, the reduction peak potential of Au0.25Pt/MWNTs shift positively to 0.876 V, however, the peak current decays to 0.64 mA cm−2 as showed in Fig.S2. In comparison, the 20 wt% Pt/C catalyst exhibits an oxygen reduction peak at 0.886 V with the current density of 0.44 mA cm−2. The CV plots of MWNTs were also measured (see Fig.S3), which presented a reduction peak at 0.736 V with the current density of 0.241 mA cm−2 in O2 saturated 0.1 M KOH solution, demonstrating that MWNTs has weak catalytic activity towards oxygen reduction reaction, decorating with AuPt NPs improve the ORR activity significantly. It is seen that all the AuxPt/MWNTs catalysts present larger reaction current densities than the 20 wt% Pt/C catalyst for oxygen reduction.The RDE experiment was also used to characterize the ORR performance of the AuxPt/MWNTs catalysts in O2-saturated 0.1 M KOH solutions. The onset potential (E onset) and half-wave potential (E 1/2) are used to characterize the ORR catalytic activity, defined as the potentials at 5% and 50% of the diffusion-limited current density, respectively. The RDE polarization curves of the AuxPt/MWNTs catalysts, MWNTs and commercial Pt/C can be found in Fig. 5 a and Fig.S4 and Fig.S5. Compare to MWNTs, the diffusion-limited current density of AuxPt/MWNTs increase significantly. It is found that with Au/Pt ratio changes from 4.55 to 1.68 and 0.67, the diffusion-limited current density increases, it then decays when the Au/Pt ratio further decrease to 0.25, thus the Au1.68Pt/MWNTs and Au0.67Pt/MWNTs exhibits the largest diffusion-limited current density. The loading mass of Pt for all AuxPt/MWNTs is less than 10 wt%, but the diffusion-limited current density is larger than that of 10 wt% Pt/C.The mechanism of ORR process can be studied by using the Koutecky-Levich (K-L) plots (Fig. 5b and Fig.S4 and S5), with indication of the relationship between the inverse square root of the rotating rate (ω -1/2) and the reciprocal of current density (J −1), and the following equations are used to calculated the overall electron transfer number (n): (1) J - 1 = J k - 1 + J L - 1 = J k - 1 + ( B ω 1 / 2 ) - 1 (2) B = 0.2 n F C O 2 ( D O 2 ) 2 / 3 υ - 1 / 6 where J k is the kinetic current density, J L is the diffusion-limited current density, ω is the angular velocity (rpm), B -1 is the slope of K-L plot, F is the Faraday constant, C O2 (1.2 × 10-6 mol cm−3) and D O2 (1.9 × 10-6 cm s−1) are the bulk concentration and diffusion coefficient of dissolved oxygen, and ν (0.01 cm2 s−1) is the viscosity coefficient. The n values are determined from the K-L plots as 3.9 ~ 4.1 for the Au4.55Pt/MWNTs (Fig.S4b), 3.8 ~ 4.0 for the Au1.68Pt/MWNTs(Fig. 5b), 3.9 ~ 4.1 for the Au0.67Pt/MWNTs (Fig.S4d), 3.8 ~ 4.1 for the Au0.25Pt/MWNTs (Fig.S4f), 3.6 ~ 3.9 for the 10 wt% Pt/C (Fig.S5b) and 4.1 ~ 4.3 for the 20 wt% Pt/C (Fig.S5d), demonstrating the four-electron pathway towards ORR in alkaline medium. However, the n value for the MWNTs is determined as 1.6 ~ 1.9 (Fig.S4h), suggesting a two-electron ORR process in connection with the generation of H2O2. Hence, the MWNTs substrate can catalyze oxygen reduction in alkaline medium but show little effect on the ORR performance of the AuxPt/MWNTs catalysts.The RDE polarization curves at 1600 rpm of the AuxPt/MWNTs catalysts are further studied. As displayed in Fig. 5c, the E onset and E 1/2 values are measured as 1.158 V and 0.890 V for the Au4.55Pt/MWNTs, 1.159 V and 0.892 V for the Au1.68Pt/MWNTs, 1.150 V and 0.894 V for the Au0.67Pt/MWNTs, 1.159 V and 0.895 V for Au0.25Pt/MWNTs, 1.151 V and 0.895 V for the 20 wt% Pt/C catalyst, respectively. Compared with the 20 wt% Pt/C, the Au0.67Pt/MWNTs and Au1.68Pt/MWNTs catalysts manifest comparable values of E 1/2 and diffusion-limited current. Tafel plots converted from polarization curves shown in Fig. 5d are also used to analyze the ORR kinetics, where the slopes for the Au4.55Pt/MWNTs, Au1.68Pt/MWNTs, Au0.67Pt/MWNTs and Au0.25Pt/MWNTs catalysts are determined as 77 mV dec-1, 74 mV dec-1 , 72 mV dec-1 and 94 mV dec-1, the one for the 20 wt% Pt/C catalyst is 73 mV dec-1, demonstrating that the ORR kinetic of Au4.55Pt/MWNTs, Au1.68Pt/MWNTs, Au0.67Pt/MWNTs is similar to the one of 20 wt% Pt/C catalyst.Notably, AuxPt/MWNTs catalysts show significant advantages when compared with the Pt/C catalyst in view of specific activity and mass activity. The electrochemical active surface areas (ECSA) of Pt was measured according to a method reported by Shao-Horn and co-workers [33,34](See ref. Fig. S6), the resulting ECSA of Pt for Au4.55Pt/MWNTs, Au1.68Pt/MWNTs, Au0.67Pt/MWNTs, Au0.25Pt/MWNTs and 20 wt% Pt/C is 184.5 m2 g-1 Pt, 204.8 m2 g-1 Pt, 87.9 m2 g-1 Pt, 33.8 m2 g-1 Pt, 90.9 m2 g-1 Pt, and the values of specific activity based on Pt are determined as 0.16 mA cm−2, 0.08 mA cm−2, 0.174 mA cm−2, 0.128 mA cm−2, 0.054 mA cm−2 at 0.9 V(Fig. 5e), respectively. The mass loading of Pt was measured using an ICP-MS, results in Table 1, show the Pt contents are 2.7 wt%, 5.3 wt%, 6.1 wt% and 7.6 wt% for Au4.55Pt/MWNTs, Au1.68Pt/MWNTs, Au0.67Pt/MWNTs, Au0.25Pt/MWNTs, respectively. The Pt mass activity values are 295 mA mg−1, 164 mA mg-1and 153 mA mg−1 and 112 mA mg−1 for Au4.55Pt/MWNTs, Au1.68Pt/MWNTs, Au0.67Pt/MWNTs and Au0.25Pt/MWNTs, which are generally two to six times higher than that of 20 wt% Pt/C (50 mA mg−1) as displayed in Fig. 5f. the metal mass activity while both Au and Pt included are 55 mA mg−1, 61 mA mg−1, 91 mA mg−1, and 89 mA mg−1 for Au4.55Pt/MWNTs, Au1.68Pt/MWNTs, Au0.67Pt/MWNTs and Au0.25Pt/MWNTs, respectively, higher than the value for the 20 wt% Pt/C catalyst (50 mA mg−1), indicating that the ORR activity increased by decorating Pt with Au cluster. possibly because the interaction between Pt and Au.RRDE measurement was employed to further investigate the oxygen reduction mechanism of the AuxPt/MWNTs catalysts in 0.1 M KOH. O2 is reduced on the glassy carbon disk electrode, and the ORR intermediate H2O2 is oxidized on the Pt ring electrode. The following equations are used to calculated the overall electron transfer number (n) and the corresponding H2O2: (3) n = 4 × I d I d + I r / N (4) H 2 O 2 % = 200 × I r / N I d + I r / N % where I d represents the disk current (A), I r is the ring current (A), N (collection efficiency) is taken as 44% according to our previous literature [13,30]. Fig. 6 a, 6c and 6e illustrate the disk currents of the AuxPt/MWNTs catalysts increase with the potential scan of disk electrode in the range from 1.2 V to 0.2 V, but the ring current approaches to zero for the AuxPt/MWNTs catalysts. Fig. 6b, 6d and 6f further illustrate that the H2O2 yields are close to zero and the total electron transfer numbers are determined as about 4 for all AuxPt/MWNTs catalysts, similar as 20wt.%Pt/C (Fig. S8), which again identifies the little contribution of the MWNTs substrate to the catalytic performance of the AuxPt/MWNTs catalysts.Apart from the ORR performance, the durability and methanol tolerance are also measured, which are carried out by the current versus time (i-t) chonoamperometry. In the durability tests, the potential was fixed at half-wave potential making the oxygen reduction to take place on the catalysts continuously, and the current was recorded. As shown in Fig. 7 a, the reaction currents for all catalysts decrease at the initial stage and then reached plateau. After 30000 s, about 84.6%, 87.5% and 87.8% of initial reaction current is observed for Au4.55Pt/MWNTs, Au0.67Pt/MWNTs and 20 wt% Pt/C, respectively. In contrast, Au1.68Pt/MWNTs exhibits higher stability with 91.6% of current retention, suggesting the superiority in practical application.In order to investigate the methanol tolerance of AuxPt/MWNTs catalysts, the I-t curves at 0.85 V vs RHE in O2-saturated 0.1 M KOH solution with the addition of 3 M methanol were recorded. As shown in Fig. 7b, apparent current decay is seen for the 20 wt% Pt/C catalyst soon after a current fluctuation in response to the addition of methanol, leaving only 44% of the initial value at 1400 s. In contrast, the current retention values are about 91%, 88% and 80% for the Au4.55Pt/MWNTs, Au1.68Pt/MWNTs and Au0.67Pt/MWNTs catalysts. It is noteworthy that the superior methanol tolerance of AuxPt/MWNTs catalysts to the Pt/C catalyst associates with the incorporation of Au NPs, in other words, the Au NPs serve to protect the Pt NPs from poisoning to some extent via particular interaction. The superior methanol tolerance of AuxPt/MWNTs catalyst also endow the application in direct methanol fuel cell.The AuxPt/MWNTs catalysts are then investigated as cathode in a home-made cell Al-air cell illustrated in Fig. 8 a. The cell consists an Al anode, air cathode and 4 M KOH electrolyte. For comparison propose, the battery performance of 20 wt% Pt/C was also tested. As showed in Fig. 8b, the cell with Au4.55Pt/MWNTs presents the lowest open circuit potential (OCP), which starts at 1.69 V but decays quickly to 1.43 V in half an hour followed by further gradual decrease to 1.36 V during 5 h. The OCP of the cell with Au1.68Pt/MWNTs starts at 1.78 V, then declines slightly to 1.67 V and maintained stable till the end of the testing. The cell with Au0.67Pt/MWNTs also has a starting OCP of 1.78 V, it declines to 1.55 V slowly during the 5 h testing. Although the OCP of the battery with 20 wt% Pt/C starts at 1.88 V, it decreases to below 1.63 V after 5 h. Fig. 8c exhibits the discharge behavior of Al-air batteries at the current density of 100 mA cm−2. The battery with Au4.55Pt/MWNTs has a discharge capacity as large as 939 mAh g−1, however, exhibits the discharge potential lower than 0.8 V. The other three batteries present the discharge potential above 0.9 V, the discharge capacity is 921, 898 and 886 mAh g−1 for Au1.68Pt/MWNTs, Au0.67Pt/MWNTs and 20 wt% Pt/C, respectively. In order to further investigate the performance of hybrid catalysts, the discharge polarization curves and the corresponding powder density curves are recorded as shown in Fig. 8d. The potential decrease sharply for the battery with Au4.55Pt/MWNTs, resulting in a maximum power density (Pmax ) of 72.7 mW cm−2. In contrast, the potential of the batteries with Au1.68Pt/MWNTs, Au0.67Pt/MWNTs and 20 wt% Pt/C cathodes decrease much slower, with the corresponding Pmax of 146.8 mW cm−2, 143.1 mW cm−2 and 144.2 mW cm−2, respectively. The assembled Al-air battery with as Au1.68Pt/MWNTs cathode can drive a fun working for at least 6 h, (Fig. S9a), and also can drive fan and hygrometer in series running for at least 3 h due to the high powder density (Fig. S9b).Apart from the above performance, the durability of catalysts is another critical factor to determine the service life of Al-air batteries. Fig. 8e displays the dynamic galvanostatic measurements of the Al-air batteries with Au1.68Pt/MWNTs, Au0.67Pt/MWNTs and 20 wt% Pt/C cathode, which are tested at the constant current density between 1 and 200 mA cm−2(60 min for each discharge plateau), accordingly, the discharge potential plateau decreases with increasing current density. There is no potential drops observed at each potential plateau with Au1.68Pt/MWNTs as cathode. In contrast, with Au0.67Pt/MWNTs and 20 wt% Pt/C cathode have potential drop happened often especially at higher current densities. To investigate the long-term stability of the catalysts, potential variations of Al-air batteries are recorded for five cycles which is operated by replacing the Al foil and electrolyte at the end of each cycle, the cathode is reused during these cycles. The potential of all cathodes is dropping at the initial stage and then the discharge plateau occurs. It is obvious that the discharge potential with Au1.68Pt/MWNTs cathode is stable for the five cycles, but it drops at the fourth cycle for the ones with Au0.67Pt/MWNTs and 20 wt% Pt/C cathode. In order to study the stability of Au1.68Pt/MWNTs after long-term operation, the morphology was characterized using TEM and HAADF as showed in Fig.S10. TEM image shows that the nanoparticles aggregated slightly, the HAADF and elemental mapping reveal that Au and Pt nanoparticles are still existed in the formation of bimetal particles, which endows the high stability of Au1.68Pt/MWNTs. These above results demonstrate that compared with 20 wt.%Pt/C, the AuxPt/MWNTs catalysts combines the advantages of high catalytic activity, superior durability, and low cost.AuxPt/MWNTs catalysts were synthesized by a facile one-pot method, where the ultrafine AuxPt NPs capped with THPO were uniformly deposited on the MWNTs substrate in an average size of ~ 3.0 nm. The AuxPt/MWNTs catalysts perform four-electron pathway towards ORR, and exhibit superior catalytic activity in terms of specific activity and mass activity. Amount which, the Au1.68Pt/MWNTs catalyst exhibits higher powder density, higher specific capacity and better durability than 20 wt% Pt/C when used as Al-air cathode. The above results demonstrate the incorporation of Pt and Au NPs enhanced the catalytic performance towards ORR. The excellent catalytic performance and stability of the bimetallic AuxPt/MWNTs catalysts allow prospective applications as efficient and stable catalysts on Al-air battery and fuel cells at lower Pt usage.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.The work was supported by the National Natural Science Foundation of China (No. 51874051), Guangxi Natural Science Foundation (No. 2018GXNSFAA281184, 2019GXNSFAA245046), Guangxi Key Laboratory of Optical and Electronic Materials and Devices (No. 20KF-4, 20AA-18) and Bagui Scholar Program of Guangxi Province. The authors are also grateful for the assistance from Mr. Xiaobin Zhou from Shiyanjia Lab (www.shiyanjia.com) on materials characterizations.Supplementary data to this article can be found online at https://doi.org/10.1016/j.apsusc.2021.150474.The following are the Supplementary data to this article: Supplementary material 1

A series of AuPt nanoparticles supported on multi-walled carbon nanotubes (AuxPt/MWNTs) catalysts with ultrafine distribution (d ≈ 3.0 nm) were synthesized for Al-air battery cathode to enhance the oxygen reduction reaction. Among them, Au0.67Pt/MWNTs catalyst with metal loading of 10.2 wt% (Au:4.1 wt%, Pt:6.1 wt%) exhibited a superior ORR catalytic activity and competitive durability to 20 wt% Pt/C catalyst. When applied as Al-air battery, appropriate increasing Au loading encourage better battery performance. Au1.68Pt/MWNTs with 8.95 wt% of Au and as little as 5.3 wt% Pt content exhibit larger specific capacity (921 mAh g−1) and power density (146.8 mW cm−2) as well as better durability than 20 wt% Pt/C catalyst when it is assembled as cathode in Al-air battery.

The Ni–25%X (X=Fe, Co, Cu, molar fraction) solid solutions were prepared and then doped into MgH2 through high-energy ball milling. The initial dehydrogenation temperatures of MgH2/Ni–25%X composites are all decreased by about 90 °C relative to the as-milled pristine MgH2. The Ni–25%Co solid solution exhibits the most excellent catalytic effect, and the milled MgH2/Ni–25%Co composite can release 5.19 wt.% hydrogen within 10 min at 300 °C, while the as-milled pristine MgH2 can only release 1.78 wt.% hydrogen. More importantly, the dehydrogenated MgH2/Ni–25%Co composite can absorb 5.39 wt.% hydrogen at 275 °C within 3 min. The superior hydrogen sorption kinetics of MgH2/Ni–25%Co can be ascribed to the actual catalytic effect of in-situ formed Mg2Ni(Co) compounds. First-principles calculations show that the hydrogen absorption/desorption energy barriers of Mg/MgH2 systems decrease significantly after doping with transition metal atoms, which interprets well the improved hydrogen sorption properties of MgH2 catalyzed by Ni-based solid solutions.

Oxidation of alcohols to their corresponding aldehydes or acids is a fundamentally important reaction both academically and industrially [1]. Although the stoichiometric oxidant, such as KMnO4, Cr-based reagents, iodine compounds [2], are efficient and versatile in catalytic oxidation of alcohols, they are neither environmentally friendly nor economically scalable. The oxidation of alcohols over heterogeneous catalysts by using O2 as an oxidant offers a safer, greener, and cheaper alternative to the stoichiometric oxidizing [1,3]. However, a base additive is required to eliminate β-H in most cases of heterogeneous catalysis, which is widely regarded as costly and non-environmentally friendly, also complicating product purification [1,3-5]. Therefore, from the standpoint of green and sustainable chemistry, solid base supports, which could facilitate alkali-free alcohols oxidation, are very attractive.Among various solid base supports, hydrotalcite (HTC), also known as layer double hydroxide, has been considered as an excellent catalyst support due to its tunable basicity /acidity properties [5–9]. By taking advantage of Brønsted-base sites arising from HTC support, several groups have synthesized HTC-supported metal nanoparticles/nanoclusters catalysts with good catalytic activity in alkali-free alcohol oxidation [5–14]. Mitsudome et al. [6] synthesized Au/Mg-Al HTC, which could catalyze the oxidation of various alcohols like benzylic alcohols and cyclohexanols without extra additives. Wang et al. [10] prepared Au/Mg-Al HTC with smaller particle size (1–5 nm) of Au nanoparticles, which showed superior catalytic properties in aerobic alcohol oxidation. Li et al. [11] synthesized Au nanoclusters (1.5 nm) deposited on Mg-Al HTC (Au NCs/Mg-Al HTC), and its catalytic performance was higher than previously reported Au/Mg-Al HTC catalysts. HT-supported Pd nanoparticles catalysts also exhibited excellent catalytic activity in oxidation of alcohols [8,9,14]. The TOF for benzyl alcohol oxidation of Pd/Mg-Al HT [9] were higher than most of Au NPs/HT or Au NCs/HT catalysts [6,10,11], which could achieve up to 11,590 h − 1 under higher reaction temperature (140 ºC). In our previous study [14], we observed that the flower-like structure of Ni-Al HTC increased the catalytic activity of Pd/HTC catalyst under base-free condition, presumably due to improved support basicity and synergetic interactions between Ni and Pd atoms in individual nanoparticles.Numerous works have demonstrated that the Pd precursors could affect the catalytic performance of supported Pd catalysts [15–18]. For example, Ali et al. [18] investigated the influence of various supports and Pd precursors on their activity in the CO hydrogenation, and showed that the catalyst prepared by PdCl2 presented higher activity than Pd(NO3)2 as precursor due to the Cl−from the PdCl2 increasing the number of intermediates/sites of CO hydrogenation. Zhang et al. [17] found that the activity of Pd/C catalysts in phenol hydrogenation was strongly affected by Pd precursors. In addition, the reduction reagents type and the presence of another transition metal also affect activity of supported Pd catalysts [15].Thus, this work is built on the earlier observations that as-synthesized flower-like Pd/HTC (10 wt.% Pd loading) catalyst could efficiently catalyze alkali-free oxidation of benzyl alcohols [5,14], and we wished to dig further into whether control synthesis of the flower-like Pd/HTC catalysts from various Pd precursors, reduction reagents, and Pd loading amount would result in further enhanced activities and product selectivity for alcohol oxidation. Herein, a series of flower-like Pd/HTC catalysts with three of palladium precursors (Na2PdCl4, K2PdCl4, PdCl2), two of reducing reagents (NaBH4, N2H4), and different Pd loading dose (1, 2, 3, 5, 7, 10 wt.%) were synthesized and characterized to determine possible effects of above parameters on structural properties and catalytic performance. Benzyl alcohol was chosen as a typical substrate in the reaction to investigate the structure-performance relationship. Moreover, the XPS and FTIR studies were used to unveil the possible mechanism reasons for the different catalytic behaviors over synthesized Pd/HTC catalysts.Potassium tetrachloropalladate (II) (K2PdCl4), sodium tetrachloropalladate (II) (Na2PdCl4), palladium (II) chloride (PdCl2), hydrazine hydrate (N2H4) and sodium borohydride (NaBH4), acetonitrile-d3 (CD3CN) were purchased from Sigma-Aldrich Denmark A/S (Søborg, Denmark). All chemicals were of analytical grade and used without further purification.Ni-Al hydrotalcite (HTC) was synthesized by urea decomposition method with the assistance of F + to control the morphology [14]. In a typical procedure, 0.24 mmol of Ni(NO3)2•6H2O, 0.08 mmol of Al(NO3)3•9H2O, 1 mmol of urea and 0.64 mmol of NH4F were dispersed in 200 mL of deionized water, and homogeneously mixed. The resulting solution was then transferred into a sealed Teflon autoclave at 130 °C for 24 h, followed by centrifugation and washing for several times with distilled water. The green powder of Ni-Al HTC was obtained by drying in the oven at 70 °C for overnight.0.2 g of Ni-Al HTC was dispersed in 50 mL of distilled water (solution A), at the same time, 0.031 g of K2PdCl4 (0.094 mmol) was dissolved in 5 ml of aqueous solution (solution B). And then two solutions were mixed together, followed by stirring at room temperature under a nitrogen atmosphere for 3 h. 0.056 g of hydrazine hydrate (1.76 mmol) was introduced to the solution for reduction of Pd2+. The mixtures were continuously agitated for another 2 h under a nitrogen atmosphere. The black powders were obtained by centrifugation, and then subjected to washing with distilled water for several times. The as-obtained powder sample was dried at 70 °C overnight and denoted as K-H. The samples of Na-H and Pd-H were obtained by following a similar process except for replacing K2PdCl4 with 0.028 g of Na2PdCl4 (0.094 mmol), and 0.017 g of PdCl2 (0.094 mmol), respectively. When PdCl2 was used as Pd precursor, the aqueous solution was mildly acidified with diluted HCl to pH 3.To examine the effect of thermal treatment, the synthesized Pd/HTC catalysts were pretreated at 200 °C for 1 h in air and the samples were named as K-H-200, Na-H-200 and Pd-H-200, respectively.In a typical reaction, 0.028 g of Na2PdCl4 (0.094 mmol) was stirred with PVA solution (1 wt%, PVA/Pd = 2 wt./1 wt.), and the solution was denoted as solution C. The solutions A and C were mixing together under N2 atmosphere at room temperature for 3 h, and then a freshly prepared NaBH4 solution (0.1 M, NaBH4/Pd = 5 mol/mol) was introduced into the mixtures to allow reaction proceed for another 2 h. The resulting sample was obtained by centrifugation, washing and drying in the oven at 70 °C overnight and named as Na-Na. The K-Na and Pd-Na were synthesized using a similar procedure except for changing the Pd precursors from Na2PdCl4 to K2PdCl4 and PdCl2, respectively.The morphologies and chemical elements were characterized by Hitachi TM3030 SEM (Krefeld, Germany) equipped with energy-dispersive spectrometer (EDS) of Quantax 70 system (Bruker, Berlin, Germany) and FEI Tecnai F20TEM (Hillsboro, OR, USA). Powder X-ray diffraction (XRD) patterns were collected in standard programs of 20 min on an Aeris Panalytical XRD with a Cu Kα1 source over a range from 5 to 90º, operating at 40 kV and 15 mA, at room temperature. IR spectra were recorded on a FTIR spectrometer (PIKE, Madison, WI; Bruker, Ettlingen, Germany) with a resolution of 4 cm −1. The powder samples were heated in an oven at 80 ºC for 4 h before the IR investigations, and 2 mg of powder was used for IR measurement. Acetonitrile-d3 was adsorbed on the samples at room temperature and pressure. The difference spectra in all figures were obtained by subtracting the spectra of the samples before the admission of the adsorbate. The chemical states of metal in samples were characterized by X-ray photoelectron spectroscopy (Specs XR 50 X-ray source +PHOIBOS 100 Analyzer) with an Al Kα1 source. The use of 1H and 13C NMR spectroscopy on a Bruker Avance III spectrometer at 400 MHz identified the structures of products from alcohols oxidation reaction.All oxidation reactions of alcohols were carried out in 15 ml glass tubes placed in a shaking incubator (IKA Incubator shaker KS 4000 i control) with a temperature control at 70 °C and a rotation speed of 250 rpm. Typically, 1 mmol of benzyl alcohol (BA) and 20 mg of as-synthesized Pd/HTC catalysts were added in 5 ml of xylene, which is acting as the reaction medium. All experiments were conducted at least in triplicates and the results are reported as means ± standard deviations. The aliquots from the reaction mixture were analyzed by using a gas chromatography (Bruker, Billerica, MA, USA) system equipped with a flame ionization detector (FID) and a ZB-FFAP column (30 m length, 0.25 mm I.D., 0.25 µm film thickness; Phenomenex, Værløse, Denmark). The temperature of both injector and detector of GC-FID were set at 250 ºC. The temperature program for oven was as follows: holding at initial temperature of 90 ºC for 1 min, followed by increasing up to 170 ºC at the rate of 20 ºC /min, and then increasing further to 220 ºC at the rate of 10 ºC /min, and finally keeping at 220 ºC for 10 min. The column flow was maintained at a constant pressure of 22 Psi throughout the analysis with helium as the carrier gas. The conversion and yield were quantified using an area normalization method. The standard reference compounds were used to identify product and side product based their respective retention time. Based on GC analysis, conversion of BA, and yield of benzaldehyde (BD), and selectivity of BD were estimated as following equations: T h e c o n v e r s i o n = ( 1 − T h e a r e a o f s u b s t r a t e p e a k T h e t o t a l p e a k a r e a s o f b o t h s u b s t r a t e a n d p r o d u c t s ) * 100 % T h e y i e l d o f B D = ( T h e a r e a o f B D p e a k T h e t o t a l p e a k a r e a s o f b o t h s u b s t r a t e a n d p r o d u c t s ) * 100 % T h e S e l e c t i v i t y o f B D = ( T h e a r e a o f B D p e a k T h e t o t a l p e a k a r e a s o f a l l p r o d u c t s ) * 100 % After the reaction was completed, the solid catalyst was separated through precipitation by centrifugation. Then, the liquid reaction mixture was dried over anhydrous Na2SO4, filtered, and evaporated by a rotary evaporator to 0.5 mL for purification of aldehyde products on silica gel by preparative TLC plates (L × W, 20 cm×20 cm, Merck). The solvent path was 19 cm and the development system was chloroform /methanol (3 : 1, v/v). A UV light (265 nm) was used to identify the products on TLC plates. The bands containing aldehyde product, were scraped and extracted with ethyl acetate and evaporated by the rotary evaporator. The weight of dried products was measured for calculation of isolated yield. The quantification of the isolated yield was based on the ratio of the purified product to the theoretical yield of expected product based on the conversion of substrate. Chemical structure identification of aldehyde products by NMR was recorded on a Bruker Avance III spectrometer (400 MHz).Data processing was applied in IBM SPSS statistics 21. One-way analysis of variance (one-way ANOVA) and independent-Samples T Test were performed to identify significant differences between groups (P<0.05).The formation of Pd nanoparticles requires the ligand dissociation of Pd salts followed by reduction of Pd 2+ cation. Hydrazine hydrate and sodium borohydride are commonly used as reducing reagents; and their reduction mechanisms are suggested as following equations [19,20]: (1) Hydrazine Hydrate : [ PdC l 4 ] 2 − + N 2 H 4 + 2 O H − → Pd + N 2 + 4 C l − + 2 H 2 O (2) Sodium borohydride : [ PdC l 4 ] 2 − + NaB H 4 → Pd + B H 3 + 4 C l − + N a + + H + It has been proved that preparation and pretreatment of catalysts affect their catalytic behavior [5,15-17]. To investigate the effect of precursor and reducing reagents on structure-activity relationship of Pd/HTC catalysts, three kinds of Pd salts (namely Na2PdCl4, K2PdCl4, and PdCl2) and two types of reduction reagents (NaBH4 and N2H4), were used to for the syntheses of Pd/HTC catalysts.The crystal structures of as-synthesized Pd/HTC catalysts were characterized by XRD (Fig. 1 ). It could be seen that the diffraction peaks of all samples exhibited the pure hydrotalcites-like phase without other detectable impurities [5]. No additional peaks assignable to any Pd phase could be observed in the Fig. 1 due to the detection limit of the instrument. A closer inspection and comparison of reflections corresponding to d 003 (the inset figure in Fig. 1) showed a reduced peak intensity in order of Na-Na < K-Na < Na-H < K-H. The real reason for this remains unknown. However, Han et al. [21] and Hao et al. [22] suggested that the introduction of metal species on the surface of support materials resulted in the decrease of XRD peak intensity of support materials. In our case, the amount of Pd loading is same for all prepared Pd/hydrotalcite samples, the XRD peak intensity of hydrotalcite can reflect the dispersion of metal particles to some extent. The lower XRD intensity of HTC support means that the more even dispersion of Pd nanoparticles on the surface of supports.To observe the dispersion of Pd nanoparticles more intuitively, typical TEM micrographs for as-obtained samples by using three kinds of Pd salts as precursors and reduced with NaBH4/N2H4 are exhibited in Fig. 2 , and their particle size distributions derived from the TEM images by counting at least 100 particles in each case are also presented. The representative high-resolution transmission electron microscopy (HR-TEM) image of as-obtained samples in Figure S1 showed that the d-spacing of d111 = 0.23 nm corresponded well with that of the (111) plane of Pd [23], which confirmed successful formation of Pd nanoparticles on the surface of Ni-Al HTC.As shown in Fig. 2, using N2H4 as reducing reagent (Eq. (1)), the average size of resulting Pd nanoparticles in the samples (Pd-H, Na-H, K-H) were generally bigger than that in the samples (Pd-Na, Na-Na, K-Na) using NaBH4 as reducing reagent (Eq. (2)]. The shape and distribution of Pd nanoparticles are more regular and uniform in the Na-Na, and K-Na than those in the corresponding Na-H, and K-H, which is consistent with XRD analysis results. The phenomenon is in line with expected results because in the process of NaBH4 reduction, PVA worked as stabilizers to protect the Pd nanoparticles from agglomeration and crystal growth, while the lacking of stabilizers in the N2H4 reduction leads to a disorder gathering of Pd nanoparticles.The distribution of Pd nanoparticles in the samples of Pd-Na, K-Na and Na-Na are carefully inspected and compared with NaBH4 as reducing reagent, as shown in Fig. 3 (a, c, e). It could be seen that the distribution of Pd nanoparticles is more even in K-Na and Na-Na than that in Pd-Na. The average size of Pd nanoparticles obeyed the order that Pd-Na (3.86 nm) > Na-Na (3.14 nm) > K-Na (2.76 nm). It seems that the cations Na+ or K + in precursors affected the Pd nanoparticle size and distribution. The possible explanation could be that the alkali metal ions of Na+ and K + in the sol-gel precursor enforced the interaction between Pd species and Ni-Al HTC support, which decreased the mobility of Pd species, therefore the rate of nuclei growth would be slower, resulting in the formation of smaller Pd nanoparticles. The K + has larger ionic radii that engenders much stronger interaction than Na+, thus induces stronger transport inhibition of Pd species on the surface of Ni-Al HTC, leading to smaller Pd nanoparticles. This is an interesting phenomenon which is similar to the Hofmeister effect [24–26].When N2H4 is used as a reduction reagent, there is no significant difference in Pd nanoparticles size and distribution for K-H (3.92 nm) and Na-H (3.82 nm), which were produced by K2PdCl4 and Na2PdCl4 as precursors, respectively. This is because by using N2H4 as a reducing reagent, the precursor is not colloid, so the Hofmeister effect is not existing. However, by using the PdCl2 as Pd source in precursor still cause bigger Pd nanoparticles (Pd-H, 4.66 nm), indicating that the presence of alkali metal ions in precursors can help to reduce the aggregate of Pd nanoparticles when N2H4 is used as reducing reagent.To further investigate the structure-activity relationship resulted from different Pd precursors and reducing reagents, benzyl alcohol oxidation was used as a typical reaction to examine the catalytic activity of the as-synthesized catalysts, and the results are presented in Fig. 3 (a-c). In Fig. 3c, ANOVA analysis showed that there has no significant difference in BA conversion among those Pd/HTC catalysts that synthesized from different precursors and reducing reagents. However, the product selectivity was affected by the reduction reagents (Fig. 3b).For a better comparison, the conversion of BA and the selectivity of BD against the mean size of Pd nanoparticles are plotted in Fig. 3a and 3b, respectively. As shown in Fig. 3a, the average BA conversion displayed an increasing mode for NaBH4-reduced Pd nanoparticles but a declining mode for N2H4 as reduced catalyst against the increasing mean size of resulting Pd nanoparticles. However, the ANOVA analysis showed that the difference of BA conversion for each sample is not significant. As depicted in Fig. 3b, the selectivity of BD for those samples which were obtained from by using N2H4 as reduction reagent (94–96%) is obviously higher than corresponding catalysts produced from by using NaBH4 as a reduction reagent (88–93%). It suggested that N2H4 reduced Pd nanoparticles might have a favorable catalytic structure and property towards selective oxidation of benzyl alcohol into benzaldehyde over NaBH4 reduced product. The reasons accounting for that will be exploited in the following section.Our previous work has disclosed the possible catalytic mechanism of Pd/HTC catalysts [14]. In this work, we also attempted to reveal the mechanism reasons accounting for sorts of similarity (e.g. in activity) or difference (e.g. in selectivity) in catalytic behaviors of Pd/HTC catalysts, which were obtained from different Pd salts and reduction reagents. Chemical changes of the surface caused by different Pd precursors and reduction agents was further studied by the XPS technique. The XPS spectra of synthesized samples were recorded (Fig. 4 ), and the relative fractions of species in Ni 2p3/2 and Pd 3d5/2 are listed in Table S1. As shown in Fig. 4, the peaks of Pd 0, Pd 2+ in Pd 3d5/2 are distinguished among samples of Na-H, Na-Na, K-H and K-Na. The deconvolution of the Pd 3d spectra for all samples showed two components with binding energies (BE) of the Pd 3d5/2 electrons at about 335.3 and 336.7 eV, which can be assigned to Pd 0 and Pd 2+ species, respectively (Fig. 4) [27]. From the Table S1, the deconvolution results showed that the oxidized state of Pd species in those samples is not significantly different, reflected in catalytic activity, the BA conversions over those Pd/HTC catalysts are similar (Fig. 3a). Furthermore, the binding energy (BE) value of the Pd 3d5/2 electron in Na-Na and K-Na are slightly higher than that in the Na-H and K-H, indicating that there is a slight electron transfer from Pd NPs to surrounding species when using NaBH4 as reduction reagents.The acid properties of the catalysts can affect their catalytic behavior, especially for product selectivity [28]. The comparison of the FTIR spectra of acetonitrile-d3 adsorbed at room temperature on the catalysts was employed for the determination of the acidity of the samples [29,30]. Kotrla et al. [29] reported that if CN in CD3CN is combined with Brønsted-acid (B-acid) sites in solid catalysts, the stretching vibration of the CN bond will shift to the lower wavenumber direction, furthermore, the stronger the acidity gives a greater displacement of the vibration of CN bond. Therefore, we examined the acidities of some typical samples by FTIR with alkaline probe molecule (CD3CN). Al2O3 is known as a solid acid, whereas MgO is a typical basic oxide. We choose those three chemicals as the references to comparison the shift of CN bond vibration when acetonitrile-d3 adsorbed at room temperature on the catalysts (Fig. 5 ).As shown in Fig. 5, all as-synthesized Pd/HTC catalysts show similar FTIR spectra. A broad band that centered at around 3450 cm−1 is attributed to the stretching mode of OH− group from the brucite-like layers, and the strong adsorption at around 1357 cm−1 is associated with the carbonate anions in the interlayers. In addition, an intense peak at around 2262 cm−1 was displayed in all FTIR spectra, which is assigned to the stretching vibration of the CN bond for acetonitrile-d3. The red dotted line in Fig. 5b represents the position of the CN bond vibration when acetonitrile-d3 adsorbed on SiO2 and MgO (the wavenumber at 2263 cm−1), while the black dotted line denotes the position for CN bond combined with the B-acid sites in Al2O3 (the wavenumber at 2261 cm−1). As shown in Fig. 5b, the blue shift of the CN bond for acetonitrile-d3 was observed from all Pd/HTC catalysts, which indicates that acidic sites exist on the surface of the as-obtained samples. From the shifting distance of the CN bond vibration peak (Fig. 5c), it can be seen that Pd-H possesses relatively strong acidity (the wavenumber at 2260 cm−1). Indeed, the Pd-H exhibited the highest BD selectivity among those as-synthesized Pd/HTC catalysts (Fig. 3b). This is similar to a previous observation, Fang et al. [28] proved that the acidity of support affected the product selectivity in alcohol dehydrogenation, the stronger acidity of support gave higher product selectivity.From the XPS and FTIR studies, we plausibly postulate that the use the NaBH4 as reduction reagent, the stabilizer protected Pd nanoparticles from aggregate, so the Pd nanoparticles evenly distributed, however, the stabilizer also takes up some B-acid site, leading to a lower product selectivity.The XRD patterns of Pd/HTC catalysts with varied Pd loading amounts were displayed in Figure S5. All diffraction peaks of the XRD patterns (Figure S5) could be assigned to characteristic peaks of hydrotalcite without any impurities, and the typical peaks of Pd were not observed in all cases, likely due to the detection limit of the instrument. The Pd nanoparticles distribution of those Pd/HTC catalysts were investigated by TEM images (Fig. 6 and Figure S6). With the increasing of Pd loading amount, the mean sizes of Pd nanoparticles increased from 2.48 nm (1 wt.%) to 3.65 nm (3 wt.%) and then slightly decreased to 3.14 nm (5 wt.%), finally kept at around 3 nm even the Pd loading increase to 10 wt.%. Du et al. [5] reported that metal nanoparticles are more inclined to load on the edge of pedals because the edges of the “flower-pedals” in Ni-Al HTC possess abundant coordinatively unsaturated metal sites (CUS) which result in higher cohesive energy, so it is more conducive to the deposit and grow metal nanoparticles. At low amount of Pd loading, there is relatively sufficient CUS on the edges of pedals, therefore, the Pd particle size climbed from 2.48 nm to 3.65 nm with the increase of Pd loading due to the aggregation (Figure S6(a) and S6(b)). While under the situation of high Pd loading amount, because of all CUS were occupied, the Pd nanoparticles were gradually loaded onto other area of pedals, leading to the slight decreasing of average particle size (Figure S6(c) and S6(d)).The catalytic performances of those Pd/HTC catalysts with varied Pd loading amount were investigated by the BA oxidation and the results are presented in Fig. 7 . As shown in Fig. 7, the BD selectivity continuously declined with the increase of Pd loading amount because the side reaction like disproportion and over-oxidation (Scheme 1 ) are more readily occurred along with high Pd amount. Furthermore, when the load of Pd exceeds 5%, the BA conversion and BD yield are descended with the increase of Pd loading amount (Fig. 7). It had been proved that the basic sites on the surface of Ni-Al HTC also plays an important role in the oxidation of alcohol [14]. Therefore, although Pd nanoparticles are actually active components, the high amount of Pd loading might occupy the basic sites on the surface of Ni-Al HTC, leading to a decreased catalytic activity. A balance point between Pd loading and basic sites on the surface of Ni-Al HTC support could be found. It can be seen that the selectivity of BD and conversion of BA for the Pd/HTC catalysts with 3 wt.% and 5 wt.% Pd loading are obviously higher than others with the low loading or high loading (Fig. 7).The pretreated process of catalyst was also investigated in this work, as shown in Figure S2, T-test analysis of the catalytic activities of Pd/HTC samples and its corresponding pretreated samples demonstrated that the pretreated Pd/HTC catalysts were not exhibiting enhanced activity comparing to the original ones, and the catalytic performances of the Pd-Na and Na-H were even dropping after pretreatment. This indicated that the ameliorating effect of pretreatment for Au/HT catalysts [5] cannot be generalized to Pd/HTC catalysts likely due to the different redox potentials of each metal.The general applicability of Pd/HTC catalyst for aerobic oxidation of alcohols was further evaluated by extending the substrate scope (Table 1 ). The product structure was confirmed by NMR analysis of isolated product, and the data are in the supporting information. As shown in Table 1, the synthesized Pd/HTC catalyst shows a higher catalytic activity for primary aromatic alcohols excluding those having highly electron-withdrawing substituents such as (2-(trifluoromethyl)phenyl)methanol, p-bromobenzyl alcohol and p-nitrobenzyl alcohol (Entries 1–10). Notably, the Pd/HTC catalyst could catalyze the benzyl alcohol oxidation in the solvent-free conditions, the TOF can reached 1182 h − 1 (Table 1, Entry 2), which is higher than previously reported Pd/HTC catalyst [8]. Primary allylic alcohols like cinnamyl alcohol also exhibited good reactivity (Entries 11), however, the side product of hydrogenation compound could be generated simultaneously. The selectivity is significantly lower for allylic alcohols (Entries 11 and 12). In addition, the synthesized Pd/HTC catalyst has excellent regioselectivity and preferred to catalyze the terminal hydroxyl group in benzylic alcohols. For examples, in the oxidation of 1-phenylethane-1,2-diol, only the product of 2‑hydroxy-2-phenylacetaldehyde was found in the reaction mixture after 7 h (Entry 13); and the secondary benzylic alcohols give very low yield of carbonyl product compared to primary alcohols (Entries 14–17). Moreover, the synthesized Pd/HTC catalyst also could catalyze the oxidation of cyclohexanols under mild reaction conditions, although it might need longer reaction time to achieve a good yield (Entry 18).In this work, we prepared a series of flower-like HTC-supported Pd nanoparticles catalysts using three kinds of Pd salts (PdCl2, Na2PdCl4 and K2PdCl4), two types of reduction reagents (NaHB4 and N2H4), and various Pd loading amount ((1, 2, 3, 5, 7 and 10 wt.%). These catalysts were characterized by SEM-EDX, TEM, and XRD. It was shown that the presence of alkali metal ions in precursors assists to reduce the aggregation of Pd nanoparticles; the Pd nanoparticles distributed more evenly when NaBH4 was used as reduction reagent compared to using N2H4 as reducing reagent, and Hofmeister effect was observed in the former case. The benzyl alcohol oxidation was chosen as a model reaction to elucidate the effect of Pd salts and reduction reagents on structure-activity relationship of Pd/HTC catalysts. Although the mean size of Pd nanoparticles synthesized by using N2H4 as reduction reagent was bigger than corresponding Pd precursor reduced by NaHB4, the benzyl alcohol conversion was neither affected by the Pd precursor nor the reduction reagents. However, higher benzaldehyde selectivity was achieved when N2H4 is used as reduction reagent. The XPS and FTIR studies elucidated that the oxidation states of Pd were not significantly changed by using different Pd salts and reduction reagents, but the Brønsted-acid sites on the surface support was affected by reduction reagents, thereby affecting product selectivity. Furthermore, although the Pd nanoparticles are real active component in Pd/HTC catalysts, higher Pd loading amount over the optimal dosage resulted in a decreasing catalytic performance due to the decrease of basic sites. The optimal Pd loading was in the range of 3 wt.% - 5 wt.%. In the last, the synthesized flower-like Pd/HTC catalyst exhibited general applicability to extended substrate scope of aromatic and aliphatic alcohols. However, high catalytic activity is not extendable to the aromatic alcohols with electron-withdrawing substituents and second alcohols; and regioselectivity is also largely limited to oxidation of primary alcohols when both types of alcohols are present simultaneously. Rongrong Dai: Methodology, Data curation, Writing – original draft. Zheng Guo: Conceptualization, Funding acquisition, Supervision, Writing – review & editing.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.Financial supports from Novo Nordisk Foundation (Grant no. NNF16OC0021740), AUFF-NOVA, Aarhus Universitets Forskningsfond (AUFF-E-2015-FLS-9–12) and Danmarks Frie Forskningsfond | Teknologi og Produktion (0136–00206B) are gratefully acknowledged. R. Dai thanks for the financial support from the China Scholarship Council (CSC).Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.mcat.2022.112403. Image, application 1 Image, application 2

Benzyl alcohol was chosen as a model substrate for Pd/hydrotalcite mediated selective oxidation of alcohols to investigate the roles of palladium precursors (Na2PdCl4, K2PdCl4, PdCl2), reduction reagents (N2H4 and NaBH4), and Pd loading amount in determining structure-activity relationship of flower-like nano-catalyst. XRD analyses show typical characteristic peaks of all hydrotalcites regardless of palladium precursors; whereas TEM studies show a larger mean size of Pd nanoparticles obtained with reducing reagent N2H4 over NaBH4. However, the difference in particle sizes didn't result in a significant difference in benzyl alcohol conversion, but extended to catalytic selectivity where N2H4-reduced Pd/hydrotalcite achieved 94%-96% selectivity over 88%-93% by NaBH4. FTIR-CD3CN surface acidity study suggested that a stronger acidity of N2H4 reduced Pd/hydrotalcite may account for its better selectivity. 3% Pd/hydrotalcite with N2H4 as reducing reagent demonstrated the best catalytic performance; and was successfully extended to oxidation of 17 aromatic and aliphatic alcohols. The conversion of alcohols is strongly dependent on individual alcohol molecular structure and electronic effect; but all reactions showed medium-to-excellent selectivity.

CO2 conversion to high value-added chemicals such as methanol and ethylene glycol (EG), which are both important bulk chemicals, had been paid great efforts since it could solve the problem of reliance on the non-renewable fossil fuels [1–6]. However, direct hydrogenation of CO2 is limited by the high activation energy for cleavage of the CO bonds of CO2. Therefore, studying the indirect transformation route of CO2 can effectively avoid the limitation of thermodynamics and kinetics [7]. Owing to the exothermic reaction of methanol synthesis, low reaction temperature is favorable. Ding and his coworkers [1] proposed the selective hydrogenation of ethylene carbonate (EC) derived from CO2 to simultaneously synthesize methanol and EG with 100% atom economy over the homogeneous Ru-based catalyst. Whereas, the difficult separation processes and the high-cost of Ru-based homogeneous catalysts imposed the high manufacture cost. Therefore, to overcome these limitations, researchers have paid much attention to develop relatively inexpensive [8,9] and heterogeneous catalysts such as Cu–SiO2-PG and Cu/MCM-41 [10,11] with high product selectivity in EC hydrogenation in recent years. Obviously, using renewable carbon source CO2 or CO2-derived EC to synthesize methanol and EG is an important research topic related to the sustainable development of resources, energy, and economic society [12–14]. Especially methanol, which is a perfect carrier of hydrogen energy, could be used as basic material to produce olefins that is in great domestic demand via methanol-to-olefins (MTO). Therefore the production of methanol from CO2 could realize carbon fixation and the emission reduction, and is of great significance for realizing carbon neutrality [15]. At present, Cu-based catalysts have good catalytic performance in catalyzing ester hydrogenation, but they are prone to agglomeration, and the loss of copper species and change of valence state during use could reduce the activity. However, the size and morphology of the active species of the carbon modified Cu-based catalyst were altered, hence the catalytic activity and stability of the catalyst were improved. For example, Cu@C catalyst prepared by Xiao et al. [16] using Cu-BTC as the precursor system has smaller metal particle size and larger specific surface area than catalyst prepared from traditional precipitation method, and the porous carbon matrix can inhibit the agglomeration of metal particles in the catalyst. Li et al. [17] showed that after modification with glucose, the highest Cu+/(Cu0 ​+ ​Cu+) molar ratio of Cu8G1/SiO2-AE catalyst was detected. In addition, the glucose modification of the Cu/SiO2-AE catalyst can also alleviate the deactivation of the catalyst distinctly. Li et al. [18] modified the Cu@SiO2 catalysts with β-cyclodextrin. The results revealed that the involvement of β-cyclodextrin improved the Cu dispersion and facilitated the exposure of more copper active sites, which also indicated that the confined catalyst inhibited the sintering of copper particles. Meanwhile, the stability of the attained carbon modified catalyst was superior. Therefore, to improve the potential application of Cu-based catalysts in industry for EC hydrogenation, it is highly desired to promote the catalytic activity and stability of Cu-based catalysts, as well as study the stabilization mechanism [18].Recently, metal-organic framework derived (MOF-derived) catalysts attract increasing attention due to their tunable structures and designable pore surfaces, and the core-shell structure was found to be able to prevent the migration and aggregation of metal particles. In general, the organic ligands in MOFs facilitates the dispersion of metal copper and clusters, which could form highly dispersed nanoparticles in MOF-derived materials [19–23]. In comparison with the conventional catalysts, MOF-derived Co2P/CN x [24], Ni/C [25], Pd@Co/CNT [26] catalysts exhibited superior performances and stability in hydrogenation reactions. Moreover, it is known that the nitrogen-doped carbon-based catalyst could lead to undesirable transesterification of EC and methanol, resulting in the decrease of methanol productivity [27]. Therein, the new class of porous regularly crystalline materials Cu3(BTC) with large ordered pores and remarkable surface area was considered as the precursors to prepare highly dispersed Cu-based catalysts with core-shell structures (Scheme 1 ). Moreover, the as-prepared catalysts were extensively characterized by N2 physisorption, TGA measurement, FT-IR, XRD, TEM, SAED, XPS, and XAES to analyze the microstructure and the physicochemical properties. The catalytic performances, the influence of reaction conditions and the durability of the as-prepared catalyst were studied. Furthermore, the synergistic mechanism to catalytic performance between different valence of copper, and the influence of graphite oxide (GO) on stability of copper species was also investigated. Synthesis of Cu@GO. Cu@GO catalyst was synthesized by ultrasonic precipitation (UP) method derived from metal-organic framework. Certain amount of Cu(NO3)2·3H2O, NaOH, and 1,3,5-benzenetricarboxylic acid were dissolved in 100 mL deionized water, 110 mL deionized water, and 250 mL ethanol with stirring, respectively. Afterwards, the attained aqueous alkali was added into the copper ion liquid dropwise around 40 min under stirring. Then, 1,3,5-benzenetricarboxylic acid solution was dropped at the same drop speed. Thereby the attained bluish violet turbid liquid was dealt by ultrasonic at 313 K for 3 h to obtain copper MOFs. Subsequently, the suspension was separated by centrifuge and washed by ethanol and deionized water for several times. Then, the as-prepared solid was dried at 373 K overnight. After drying the solid was calcined under nitrogen atmosphere at 723 K for 4 h. Before being used for catalysis, the solid was reduced by 10 vol% H2/N2 at 623 K for 4 h, and the finally obtained catalysts were named as Cu@GO. Synthesis of Cu-catalyst. The Cu-catalyst was synthesized by the following procedure. The desired amounts of Cu(NO3)2·3H2O and NaOH were weighted and dissolved in 100 mL and 110 mL deionized water, respectively. After stirring for 10 min, the aqueous alkali was added into the solution of Cu2+ dropwise in around 60 min to attain Cu complex precipitation. After aging for 3 h at ambient temperature, the attained suspension was separated via filtration and washed carefully by deionized water for several times. The subsequent step was the same as the Cu@GO catalyst and the finally obtained catalyst was named as Cu-catalyst. Characterization. On Quantachrome Autosorb-1, N2 physisorption analysis was performed. The measurement started at liquid nitrogen temperature (77 K), then the sample was outgassed in vacuum at 573 K for 3 h. Through Brunauer-Emmett-Teller (BET) method, the special surface area (S BET) was obtained. Meanwhile, the total pore volume (V pore) was determined from the absorbed volume of nitrogen with a relative pressure of 0.99. Based on the desorption branch of the isotherm, pore size distribution was estimated by the Barrett, Joyner, and Halenda (BJH) method. Inductively coupled plasma-atomic emission spectroscopy (ICP-AES) on PerkinElmer optima 5300DV was used to determine the Cu content of catalysts. TGA measurement was realized on Diamond TG-DTA6300 thermo-gravimetry analyzer in the purity N2 atmosphere with a heating rate of 5 K min-1. X-ray diffraction (XRD) patterns of samples were obtained through a PANalytical Empyrean diffractometer with Cu Kα radiation (λ = 0.15406 nm) over 2θ range of 10o–90o. Fourier-transform infrared spectra (FT-IR) was measured on a Bruker Tensor 27 spectrometer. The catalysts were ground and uniformly dispersed in KBr to obtain pellets suitable for FT-IR characterization. Raman spectra was attained on an apparatus pfLabRam HR800. The powder catalyst was dispersed on a glass slide for Raman characterization. H2 temperature-programmed reduction (H2-TPR) was executed on a Quantachrome Chembet pulsar TPR/TPD instrument. Before reducing it by hydrogen, He gas flow was used to pretreat the calcined catalysts at 473 K for around 1 hour. Then it was cooling down to ambient temperature. Subsequently, the TPR program was executed with an increasing temperature program from 323 K to 803 K at the rate of 10 K min-1 under 10 vol% H2/N2. Transmission electron microscopy (TEM) and high-resolution transmission electron microscopy (HRTEM) images were attained by field-emission transmission electron microscopy (JEOL, JEM-2100F), which was operated at an acceleration voltage of 200 kV to characterize the morphologies and the crystal structures of the Cu@GO. Before scanning the element distribution, the powder was sonicated in ethanol for around half hour. Then, the suspension was carefully drop-casted on copper grids, which were supported by holey carbon films. The test was obtained after the sample drying with EDX scanning mode on JEOL. In order to get the oxidation state of catalyst and to know more surface physical properties, X-ray photoelectron spectroscopy (XPS) and X-ray Auger electron spectroscopy (XAES) were reached. All measurements were achieved under an ultra high vacuum via an ESCALAB 259Xi spectrometer with Al Kα radiation (1486.6 eV) and a multichannel detector. Then the binding energy with C 1s at 284.6 eV with accuracy of ± 0.2 eV was calibrated. Computational methods and models [28–30]. During the molecular simulations, periodic boundary conditions were implemented along the Cu basal plane, and the velocity Verlet algorithm with a time step of 1 fs was used. All the MD simulations were conducted in the NVT ensemble. The Langevin thermostat were employed to keep the constant temperature of the substrate. The interatomic interaction of C atoms could be described by airebo pair style, the interaction between Cu atoms could be reflected by the EAM potential, and the Cu–C interaction could be reflected by the Lennard-Jones potential (r = 3.0825 Å, e = 0.02578 eV). Fig. 1 illustrated the MD simulation model, where Cu1 represented the first layer and Cu2 represented the second layer.The first-principles [25] were employed to perform all spin-polarization density functional theory (DFT) calculations [36] within the generalized gradient approximation (GGA) using the Perdew-Burke-Ernzerhof (PBE) formula. The projected augmented wave (PAW) potentials [19,20] were chosen to describe the ionic cores and used a plane wave basis set with a kinetic energy cutoff of 400 eV to consider the valence electrons. The Gaussian smearing method allowed partial occupancies of the Kohn−Sham orbitals with a width of 0.05 eV. When the energy change was smaller than 10-6 eV, the electronic energy was considered self-consistent. When the energy change was smaller than 0.05 eV Å−1, a geometry optimization was considered convergent. The vacuum spacing perpendicular to the plane of the structure is 15 Å. The Brillouin zone integration is performed on the structure using 2 × 2 × 1 Monkhorst-Pack k-point sampling. Finally, the adsorption energies (E ads) were calculated as E ads = E ad/sub - E ad - E sub, wherein E ad/sub, E ad, and E sub are the total energies of the optimized adsorbate/substrate system, the adsorbate in the structure, and the clean substrate, respectively. The free energy was calculated using the equation: (1) G = E + Z P E − T S where G, E, ZPE, and TS are the free energy, total energy from DFT calculations, zero point energy, and entropic contributions, respectively. In the calculation, U correction had been set as 3.68 for Cu atoms and the bottom layers had been fixed in this system.Figure S1 showed the pyrolysis process of the dried Cu@GO catalyst in N2 atmosphere. Four weight losses from room temperature to 400 °C in TGA curve can be seen. The first loss of 4.6% is deemed as the loss of the absorbed guest molecules within the porous structure of the dried Cu@GO catalyst [31]. The second and third losses of 7.5% and 1.9% corresponded to the loss of bound water under the temperature lower than 250 °C. The last weight loss of 38.7% around 300 °C could be considered to be the decomposition of the dried precursor of Cu@GO catalyst Cu3(BTC)2 [32,33].The specific surface areas and pore size distribution of Cu@GO and Cu-catalyst were studied by N2 adsorption-desorption, and the details were illustrated in Fig. S2 and Table 1 . The S BET of Cu@GO and Cu-catalyst were 96.9 and 1.6 m2 g-1 and the average pore sizes of these two catalysts were 3.0 and 20.4 nm with the pore volume of 0.07 and 0.01 cm3 g-1, respectively. In comparison with Cu-catalyst, Cu@GO catalyst possessed a larger specific surface area, which was expected to provide more accessible active sites [34].Accordingly, the pore size of Cu@GO catalyst was quite lower than that of Cu-catalyst, implying the smaller Cu particles formed in Cu@GO catalyst promoted the accessible active sites. As shown in Fig. S2a, according to IUPAC classification, N2 isothermal physisorption curve of Cu@GO catalyst can be classified as type IV with H1-type hysteresis loops, indicating the wedge hole of the pore [2,35]. While the physisorption curve of Cu-catalyst was type III without hysteresis loop, suggesting the accumulation of cooper species [18]. Fig. S2b showed that the pore size distribution of Cu@GO catalyst was relatively narrow and sharp with a range of 3–10 nm, promoting the formation of relatively smaller Cu particles as well as implying the pore size and pore structure of the as-prepared Cu@GO catalyst were highly random [24]. Meanwhile, ICP was used to determine the Cu contents of the attained catalysts collected in Table 1, and the results indicated that the Cu contents in Cu@GO and Cu-catalyst were 76.2% and 93.4%, respectively.As shown in Fig. 2 a, the XRD spectrum of the dried precursor of Cu@GO catalyst could match well with Cu3(BTC)2 crystal that the diffraction peaks 2θ of 9.9o, 10.9o, 13.4o, 21.3o, 26.1o, and 28.6o could be assigned to (220), (400), (440), (600), (731), and (751) crystal planes of Cu3(BTC)2. The result proved that the Cu MOF of Cu3(BTC)2 could be synthesized via the ultrasonic method [27,37-40]. As for the reduced Cu@GO catalyst, the (111), (200), and (220) crystal planes of Cu2O at the positions of 36.5o, 42.6o, and 61.7o were observed and the (111), (200), and (220) crystal planes of Cu at the positions of 2θ of 43.5o, 50.6o, and 74.2o were also discovered in the attained catalyst [33,41], and there was no characteristic peaks of CuO residual were observed, thereby it can be speculated that all of the Cu2+ were reduced during the calcination and reduction process. Accordingly, the foundation of graphite oxide in Fig. 2a verified the formation of carbon film which promoted the stability of Cu@GO catalyst to a large extent [42]. For the sake of exploring the differences between the contrastive Cu-catalyst and Cu@GO catalyst, XRD analysis of Cu-catalyst was also performed, as illustrated in Fig. 2b. The main diffraction peaks at 2θ values of 35.5o, 37.7o, 38.8o, 48.8o, 53.4o, 61.6o, 66.2o, 68.1o, and 72.3o could be assigned to (110), (111), (200), (202), (020), (022), (311), (220), and (311) planes of CuO, and there was no any other phase could be detected [43,44], implying the dried precursor of Cu-catalyst was the purity CuO after drying overnight. For the reduced Cu-catalyst, the observed peaks can be divided into two parts, 2θ values of 36.5o, 42.6o, and 61.7o were assigned to (111), (200), and (220) planes of Cu2O and 2θ values of 43.5o, 50.6o, and 74.2o were assigned to (111), (200), and (220) planes of Cu, respectively. Whereas, the planes of CuO (110), (111), and (200) corresponded to 2θ values of 32.4o, 35.5o, and 38.8o still existed in the reduced Cu-catalyst [45], which could be ascribed to the incomplete reduction or oxidized by oxygen in the atmosphere when it was exposed to the air atmosphere. The as-prepared Cu-catalyst possessed similar crystal planes with Cu@GO catalyst. It could be speculated that the exit of residual carbon could obviously prevent the oxidization of Cu@GO catalyst in the air atmosphere.In order to analyze the functional groups of the dried precursor of Cu@GO and the reduced Cu@GO catalysts, FT-IR was also employed, and the results were shown in Fig. 3 . For the precursor of Cu@GO catalyst, it was found that there was an obvious adsorption peak of 474 cm-1, which could be attributed to the Cu–O stretching vibration of pure CuO with a monoclinic structure [46,47]. Furthermore, the observation of 739 cm-1 could be attributed to the out-of-plane bending vibration of C–H group [48]. The characteristic band of Cu3(BTC)2 located at 1377 cm-1 standing for the stretching vibration of aromatic ring. Another characteristic band at 1104 cm-1 can be identified as the stretching bond of C–C on the aromatic ring as well [29,35]. The strong vibrational bands at 1627 cm-1 and the strong broad and bands between 3180 and 3510 cm-1 could be pointed out as –COOH of H3BTC which was used as a linker in Cu-BTC [49]. Through the existence of these six distinct absorption peaks, it was not hard to speculate that the successful formation of Cu3(BTC)2 which was the dried precursor of Cu@GO catalyst. For the curve of the reduced Cu@GO catalyst, the peak at 739 cm-1 could be assigned to the C–H group out-of-plane bending vibration. The intensity declined remarkably, implying that the main C–H structure was destroyed during the pyrolysis process in the N2 atmosphere while there was still some carbon residual of C–H structure. The bands of 1055 cm-1, 1107 cm-1, and 1242 cm-1, which could be assigned to the characteristic C–C stretching vibration, CO–Cu stretching, and epoxy groups C–O–C stretching vibration of the carboxyl group [44], indicated the existence of carboxyl groups, the result of which further proved the structure of graphite oxide. Therefore, it could be speculated that the graphite oxide was formed over the copper particles. Furthermore, the strong absorption bands of 630 cm-1 could be assigned to the inorganic network Cu–O of Cu2O phase and this bond shifted from 474 cm-1 of the dried precursor Cu3(BTC)2 during the reduced process, implying that Cu+ was formed [47].Raman test was performed to study the shape of carbon in Cu@GO catalyst. It was shown in Fig. 4 that the G-band (∼1598 cm-1) existed in as-prepared catalyst, which was ascribed to the stretching vibration of the sp2-hybridized carbon atoms in graphite oxide. The existence of sp2 hybridization was favorable to the interaction between Cu species and graphite oxide film [50,51]. There was a weak D-band (∼1327 cm-1), suggesting that there was less disordered density of carbon atoms at the plane edges in the reduced Cu@GO catalyst. The presence of D-band and G-band verified that carbon and copper exist simultaneously in Cu@GO catalysts, which was in accordance with the results of XRD and FT-IR.The H2-TPR profiles of Cu@GO and Cu-catalysts were depicted in Fig. 5 . It could be found that both reduction peaks from Cu@GO and Cu-catalyst are symmetrical. Thereby the onset reduction temperature of Cu@GO catalyst was 475 K, ascribed to the well dispersed CuO particles [52]. Whereas, the onset reduction temperature of Cu-catalyst was 518 K, presumably ascribed to the larger CuO x particle size with deteriorated dispersion. In this situation, it could be concluded that after calcination, Cu@GO catalyst was reduced to both Cu+ and Cu0 species selectively according to its reduction temperature. While it also could be deduced that the reduction process of the precursor calcinated Cu-catalysts was incomplete, which further verified the results from FT-IR and XRD. Fig. 6 a-b is the SEM images of the dried precursors of Cu@GO. It is seen that the precursors of Cu@GO and Cu3(BTC)2 were formed into octahedral structures which were consistent with the previous observation for cubic MOF crystal [40]. The whole morphology of Cu@GO catalyst was illustrated in Fig. 6c-d, and it was clearly seen that the metallic Cu particles was embedded in the porous carbon matrix uniformly with the range of Cu particles sizes from 20 to 130 nm. It could be also found that the abundant and disordered nanoporous was formed in the attained Cu@GO catalyst and was connected via a network of channels which was consistent with the result of pore size distribution from N2 physisorption. Meanwhile, high-resolution was further performed to study the microstructure of Cu@GO catalyst. The structure of Fig. 6e revealed that there was a film of carbon which sealed the Cu nanoparticles in the graphite oxide layer. The lattice spacing of Cu@GO catalyst shown in Fig. 6f was also measured afforded by HRTEM with 0.21 nm of (111) crystal plane of metallic Cu and 0.27 nm of (111) crystal plane of Cu2O [53,54], respectively, coinciding with the identified crystal planes from XRD.As shown in Fig. 6g, selected-area electron diffraction (SAED) was adopted to probe the diffraction planes of the attained catalyst. There were a series of rings in SAED pattern, which could be assigned to the characteristic diffraction planes of (111), (200), and (220) of metallic Cu and the Cu2O planes of (111), (200), and (220), which was consistent with the results from XRD [40,55,56]. As for the comparison of Cu-catalyst, the Cu particles distributed irregularly, as shown in Fig. 6h-i. Fig. 7 showed that the elements of C, Cu, and O were uniformly dispersed in Cu@GO catalyst. It is seen from Fig. 7b that there were a series of vacancies in C-mapping, and the residual carbon from Cu3(BTC)2 existed by connecting network channels. Fig. 7c illustrated the uniform distribution of Cu is in the form of plenty of in Cu nanoparticles which match well with the vacancy in C-mapping, indicating Cu nanoparticles were encapsulated in carbon film equably which coincided with the results from N2 physisorption, XRD, FT-IR, and Raman. Moreover, the graphite oxide film could encapsulate the copper particles. Furthermore, Fig. 7d of O-mapping revealed the uniform co-existence of C, Cu, and O indicating the energy bonds between these three elements. In summary of the TEM image, FT-IR, and EDX mapping, a conclusion could be drawn that the ultrasonic precipitation was an excellent pathway for producing the Cu-based catalyst covered by carbon which promoted the catalytic activity and stability simultaneously.The high angle annular dark-field scanning/transmission electron microscope was executed to verify the core-shell structure of Cu@GO catalyst. The cross-sectional compositional line profile was carried out and illustrated in Fig. 8 . Fig. 8b exhibited that Cu and C co-existed in each part, while the Cu content went up from the margins to centers of the particles. Meanwhile, the contents of carbon around the margins of the particles were much higher than the content of Cu, and the contents of carbon were lower in the center of catalyst particles. Therefore, it could be speculated that the nanoparticles Cu cores and carbon shells were successfully prepared via the as-mentioned method which were accordant with the results from TEM, XRD, and EDX.In Fig. 9 a, the XPS survey spectra showed three different elements, C, O, and Cu in Cu@GO catalyst, where the element of C was derived from the residual carbon during the pyrolysis process. Three peaks in C 1s high-resolution spectrum of Cu@GO catalyst, centering at 284.6, 285.4, and 286.5 eV which can be assigned to C−C/CC, C−O, and CO in Fig. 9b, respectively [57,58]. These carbon peaks were the characteristic peaks of graphite carbon derived graphite oxide corresponding to the results of XRD and FT-IR, and further verified the formation of carbon film over Cu core. The binding energy at 933.8 eV from Cu 2p high-resolution spectrum in Fig. 9c was ascribed to Cu in Cu0 and Cu + state [59,60], hinting that both Cu2O and Cu phases were detected in Cu@GO and Cu-catalyst. The intensive broad peak of 943.0 eV, which was attributed to Cu2+ in Cu-catalyst [61], revealed that the residual CuO was oxidized again when it was exposed to the air atmosphere, which was in accordance with the XRD result. Whereas, the Cu2+ appeared in Cu-catalyst was considered to limit the catalytic performance to some extent, because the atom efficiency of Cu for the formation of active Cu species inevitably decreased. Therefore, it could be speculated that the existence of carbon inhibited the oxygenation of catalyst, facilitated the distribution of Cu nanoparticles. XAES was further employed to differentiate Cu0 and Cu + species in catalysts, and the asymmetric peaks were divided into two overlapping Cu LMM Auger kinetic energy around 913.0 and 917.0 eV correspond with Cu+ and Cu0 shown in Fig. 9d [62], respectively. There was an interesting phenomenon that the broad peak of Cu catalyst could be deconvoluted to three different peaks thereby the Cu2+ was observed at kinetic energy of 918.2 eV [63], suggesting the incomplete reduction of Cu-catalyst which was also confirmed by XRD and XPS results. As calculated in Table 2 , the ratio of Cu+/(Cu+ ​+ ​Cu0) was 0.50 for Cu@GO catalyst which was appropriate, while the ratio of Cu+/(Cu+ ​+ Cu0) in Cu-catalyst was 0.61 as a favorable catalyst. Whereas, the Cu2+ appeared in Cu-catalyst limited the catalytic performance. In comparison, the ratio of Cu+ in Cu@GO catalyst attained a balance level which was also elevated by the residual graphite oxide [64]. The results certified the co-existence of Cu+ and Cu0 in the as-prepared catalysts, and the synergistic effect on the surfaces of catalysts.The performance of Cu@GO and Cu-catalyst in the hydrogenation of EC was operated under the temperature of 473 K, time of 4 h and the hydrogen pressure of 5 MPa. It is seen in Table 3 that, the conversion of the EC attained 71.2% with 69.2% methanol and 98.2% EG selectivity, showing excellent catalytic activity in EC hydrogenation by applying Cu@GO catalyst. As for Cu-catalyst, EC conversion was only 44.9% with methanol and EG selectivity were 21.1% and 80.7%, respectively, the results of which was extremely lower than Cu@GO catalyst although the Cu content of Cu-catalysts (93.4%) is higher than Cu content of Cu@GO catalyst (76.2%). Therefore, it is speculated that high distribution of Cu species in Cu@GO catalyst assured the higher catalytic activity thereby the disordered carbon nanoporous facilitated the distribution of Cu nanoparticles as well as inhibited the oxidizing process of low valence Cu species. These results implied that the structure of Cu@GO catalyst with Cu core and graphitic oxide shell derived from Cu3(BTC)2 enhanced the catalytic activity effectively, synchronously increased the selectivity of co-product methanol and EG.The effect of reaction time and temperature on hydrogenation of EC was studied to find the optimized reaction conditions. As illustrated in Fig. 10 a, the reaction temperatures ranging from 453 to 493 K were tested. EC conversion increased with the reaction temperatures all the time, while the selectivity of EG declined slightly from 99.9% at 453 to 94.6% at 493 K. The selectivity of methanol increased with temperature and reached to the highest of 69.0% at the temperature of 473 K, then kept unchanged with further increase the temperature to 493 K. Moreover, the high TOF of 1526 mgEC gcat −1 h−1 could be attained in the hydrogenation of EC at 493 K.The result of the temperature effect indicated that the temperature is of great importance in the catalytic activity thereby higher temperature is benefited to the transformation of EC, but the co-product methanol and EG will be deteriorated if the reaction temperature was too high. The effect of reaction time was also studied, which was illustrated in Fig. 10b. Conversion of EC increased gradually with the reaction time and attained complete conversion at 493 K for 6 h with 67.1% methanol and 91.7% EG selectivity. Noteworthy to say that, EG selectivity decreased from 94.6% to 91.7% with reaction time prolonged to 6 h, implying the reaction carried out at high temperature for a long time accelerated the deterioration of EG. The attained results verified that Cu@GO catalysts exhibited superior catalytic activity for EC hydrogenation.The stability of catalysts plays a significant role in its application in chemical industry, and the present unstable Cu-based catalyst or the unclear stabilization mechanism is desired to be developed. The reusability of Cu@GO catalyst was shown in Fig. 11 . Compared with the first run, EC conversion increased to 75.0% with selectivity of methanol and EG increased to 83.3% and 99.9% in the second run, respectively, assigned to the excitation of Cu@GO catalytic sites in the first run. EC conversion declined to 62.8% in the third run and maintained at a relative balance value around 65.0% in the subsequent cycles. The methanol selectivity also decreased in the subsequent cycles and reached to 67.3% after six runs, showing excellent stability. Furthermore, EG selectivity kept at 99.9% without any change after the second run. In summary, 64.3% EC conversion, 67.3% methanol, and 99.9% EG selectivity could be reached after 6 runs. The reason could be speculated that the graphitic oxide inhibited the aggregation of Cu, thus enhanced the stability of Cu-based catalysts, simultaneously promoted the distribution of Cu species to improve the catalytic activity.In order to further study the stabilization mechanism of graphite oxide film of Cu-based catalyst, characterizations of XRD, TEM, and XPS of the reused Cu@GO catalyst were employed. As depicted in Fig. 12 a, the same planes of metallic Cu and Cu2O could be observed in the reused Cu@GO catalyst, and it could be speculated that the formation of graphite oxide film inhibited the microstructure variation of Cu-based catalyst effectively. The TEM image shown in Fig. 12b reflected that there was hardly any changes of the nanoporous channel by connecting network, thereby the pore size of the reused catalyst also maintained the same level compared with the fresh Cu@GO catalyst. Herein, the formation of graphite oxide film inhibited the volume diffusion to restrain the aggregation of Cu particles. Meanwhile, XAES was performed for analyzing the distribution of Cu species, and results showed that Cu+ accounts for 51% of (Cu+ ​+ ​Cu0), the value of which was essentially preserved in comparison with the fresh Cu@GO catalyst. This phenomenon implied the Cu species of the as-prepared catalyst was not obviously influenced under the high temperature reaction. In combination of the characterization result of the reused and fresh Cu@GO catalyst, it can be deduced that the existence of graphite oxide film is beneficial to the catalyst stability. Furthermore, it can be concluded the higher balance of Cu species should be contributed to the excellent catalytic activity.Molecular simulation calculations were carried out to study the effect of C on the stability of catalyst, and its radial distribution function was shown in Fig. S3. It is shown that the maximum radial distribution probability of Cu1 and Cu2 in the first and second layers of the two models is 2.5 Å. However, in the Cu–C model, the distribution range of Cu1 is narrower and the maximum probability value is higher, indicating that the relaxation range of Cu1 is smaller than that of Cu2, Cu atoms on the surface are more stable, which decrease the aggregation of Cu particles effectively. The molecular simulation results indicated that graphite oxide can promote the stability of Cu-based catalyst, which is consistent with the experimental results.The adsorption and dissociation behaviors of EC and H2 on Cu with different valence states were simulated. The adsorption of EC and H2 on Cu/C and Cu2O/C was simulated to study the effect of graphite oxide membrane on the adsorption of EC and H2. The optimized geometric structure of DFT is shown in Fig. S4. It is seen that the initial distance between cyclic oxygen and Cu+ site of EC molecule is 1.908 Å, and the initial distance between carbonyl oxygen and Cu+ site is 2.454 Å. The cyclic oxygen in EC molecule is closer to Cu+ site than its carbonyl oxygen, the adsorption energy of EC molecules on Cu0 and Cu+ are -0.49 and -0.64 eV, respectively. These results indicate that EC molecule is stably adsorbed to Cu0 and Cu+ sites, and the interaction between carbonyl oxygen and Cu+ sites is dominant. Fig. S4c and d showed the adsorption of H2 on Cu0 and Cu+. The adsorption energy of H2 on Cu0 and Cu+ is -0.35 and -0.40 eV, respectively. This indicated that the presence of graphite oxide film is conducive to the adsorption of H2 molecules on Cu0 and Cu+ sites. Fig. 13 showed the dissociation process of H2 on Cu0 and Cu+, respectively. As shown in Fig. 13a, the initial distance between hydrogen molecule and Cu0 site is 1.985 Å, and the initial distance between the two H atoms and Cu0 site is 1.228 Å. After the dissociation, the distance between the two H atoms and Cu0 site becomes 1.991 Å and 1.834 Å, the distance between the two H atoms becomes 2.689 Å. Similarly, as shown in Fig. 13b, the initial distance between hydrogen molecule and Cu+ site is 1.987 Å, and the initial distance between the two H atoms is 1.267 Å. After the dissociation, the distance between the two H atoms and Cu+ site becomes 1.800 Å and 1.766 Å. As shown in Fig. 14 , the dissociation energies of H2 on Cu0 and Cu+ are 31.08 and 23.16 kcal·mol-1, respectively. The results showed that H2 adsorbed on the catalyst is stable and easily dissociated through the synergistic action of Cu0 and Cu+ due to the presence of graphite oxide. Fig. 15 showed the differential charge density of H2 and EC adsorbed by Cu/C and Cu2O/C structures, where yellow is the area of electron accumulation and green is the area of electron dissipation. The results showed that Cu0 and Cu+ exhibited special electronic states when the catalyst adsorbs H2 and EC. At the same time, the C layer also has electron transfer, which further explained the role of graphite oxide.In this work, Cu@GO catalyst with carbon film derived from MOF was prepared by ultrasonic precipitation method, which was then used in the hydrogenation of EC derived from CO2 to produce methanol and EG simultaneously. The Cu particles were encapsulated in graphite oxide film with random nanoporous channel connected via network. The results of the catalytic performance proved that the residual carbon of Cu@GO catalyst promoted the catalytic activity. Meanwhile, the ratio and the synergistic effect between the copper species of Cu+ and Cu0 contribute to the superior catalytic activity. A high TOF of 1526 mgEC gcat -1 h-1 was attained in the hydrogenation of EC at 493 K. Moreover, the reusability study of Cu@GO catalyst showed excellent stability of it. DFT calculation results indicated that the graphite oxide film inhibited the aggregation of Cu particles, therefore promoted the stability of Cu-based catalyst effectively. In conclusion, a facile method was provided for preparing uniformly distributed and stable Cu@GO catalyst via ultrasonic precipitation method derived from MOF, which showed excellent catalytic performance towards the hydrogenation of CO2-derived EC to synthesize methanol, as well as EG.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.We are grateful to the National Natural Science Foundation of China (21576272), “Transformational Technologies for Clean Energy and Demonstration”, Strategic Priority Research Program of the Chinese Academy of Sciences (Grant No. XDA 21030600) and Science and Technology Service Network Initiative, Chinese Academy of Sciences (KFJ-STS-QYZD-138), the National Key Research and Development Program of China (2019YFC1906701) for the financial supports.The following is the Supplementary data to this article: Multimedia component 1 Multimedia component 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.gce.2021.12.004.

The synthesis of sustainable methanol and ethylene glycol (EG) via hydrogenation of ethylene carbonate (EC) has caught researchers’ growing interests on account of the indirect chemical utilization of CO2. Core-shell Cu@GO catalysts with random nanoporous network of graphite oxide (GO) were synthesized via a simple method of ultrasonic precipitation. Cu@GO catalysts were analyzed systematically by N2 physisorption, TGA measurement, XRD, FT-IR, Raman, TEM, SEM, and XPS (XAES). In particular, the mentioned method was confirmed to be effective to fabricate the high dispersity core-shell Cu@GO catalysts through promoting the specific surface area. The as-prepared Cu@GO catalyst was then successfully applied in the hydrogenation of CO2-derived EC to produce methanol and EG. A high TOF of 1526 mgEC gcat -1 h-1 could be attained in EC hydrogenation at the reaction temperature of 493 K. Accordingly, the correlation of catalytic structure and performance disclosed that the synergistic effect between Cu+ and Cu0 was responsible for achieving high activity of the catalyst. In addition, the reusability of Cu@GO catalyst suggested that graphite oxide shell structure could decrease the aggregation of Cu particles, thus enhance the stability of Cu-based catalysts. DFT calculation results suggested that the involvement of carbon film on Cu was favorable for the stabilization of the active sites. This study is helpful for developing new and stable catalytic system for indirect chemical utilization of CO2 to synthesize commodity methanol and EG.

Accumulation of degradation-resistant pollutants such as nitroaromatics and azo dyes in water bodies poses a threat to the aquatic ecosystem as well as human health. The high water solubility of certain nitroaromatics like 4-nitrophenol (4-NP) and azo dyes can make their removal quite challenging. This triggered the development of methods based on efficient adsorbents (Yagub et al., 2014; Gupta et al., 2013; Dias and Petit, 2015; Parida et al., 2021), photocatalytic degradation (Dias and Petit, 2015), bio-degradation (Marvin-Sikkema and de Bont, 1994; Ju and Parales, 2010), and catalytic conversion (Fu et al., 2019; Zeng et al., 2013) as viable purification strategy to maintain good water quality. Among these options, catalytic conversion is the preferred method as it offers a possibility to convert the pollutants to valuable products and less harmful counterparts with high efficiency. For example, upon hydrogenation, carcinogenic 4-NP can be converted to 4-aminophenol (4-AP), which is an intermediate for corrosion inhibitors (Thenmozhi et al., 2014; Guenbour et al., 2000), pharmaceutical molecules, and dyes (Buschmann, 2007). Transforming nitro pollutants to useful amino compounds is a green and commercially beneficial approach to get rid of pollutants.Among various methods, noble metal nanoparticles (NMPs) catalyzed hydrogenation continues to draw considerable attention owing to their high activity and oxidative stability (Fu et al., 2019; Zheng and Zhang, 2007; Qin et al., 2019; Lu et al., 2006; Soğukömeroğulları et al., 2019; Kästner and Thünemann, 2016). With the rise in demand for active catalysts, the use of noble metals like Au, Pt, and Pd-based catalysts have grown recently (Fu et al., 2019; Soğukömeroğulları et al., 2019; Ansar and Kitchens, 2016; Nguyen et al., 2019; Sun et al., 2014; Johnson et al., 2013; Goepel et al., 2014; Fu et al., 2019). However, the high cost of these metals is a major drawback and thus triggered the development of supports for enhanced recoverability. Supports containing magnetic particle is one such approach for easy recovery and reuse of expensive NMP catalysts (Zeng et al., 2013; Yang et al., 2020; Xu et al., 2020). The development of bimetallic nanoparticle catalysts is another approach to reduce the cost along with improvement in the activity compared to their monolithic counterparts (Qin et al., 2019; Fu et al., 2018). Despite these attempts, the overall price of such catalysts remains higher than catalysts prepared from metals like Ni, Cu, and Ag.In this context, MNPs of moderately active and low-cost metals (Cu and Ag) are still attractive choices as catalysts (Li et al., 2015; Zhou et al., 2020; Das et al., 2019; Dong et al., 2014; Qian et al., 2020; Bhaduri and Polubesova, 2020; Sudhakar and Soni, 2018; Budi et al., 2021). To improve the activity of silver nanoparticles (AgNPs), supports have been designed to boost the catalytic activity via increasing the available surface area for catalysis. Using this principle, nanosheets (Li et al., 2015; Qian et al., 2020; Mao et al., 2018), conductive polymers (Das et al., 2019), and fibrous silica (Dong et al., 2014) were employed for the successful enhancement of the catalytic activity of AgNPs. On the other hand, porous supports prepared from carbon and organic polymers are also known to have a positive effect on the catalytic activity of NMPs (Zhou et al., 2020; Bhaduri and Polubesova, 2020; Gong et al., 2019; Xia et al., 2016; Budi et al., 2020). In the case of these supports, confined space reaction and enrichment of micro-environment abound NMPs by adsorption of substrate molecules enhance the activity (Qin et al., 2019; Gong et al., 2019; Cárdenas-Lizana et al., 2013). Moreover, fast electron transfer from support like carbon black to NMPs is also known to improve catalytic efficiency (Qin et al., 2019). Although these supports are known to enhance the catalytic activity of NMP, a portion of the NMP surface is shielded by the support, thus, resulting in their underutilization.To counter the underutilization of NMPs, relatively mobile cross-linked polymeric networks or hydrogels are investigated as supports (Lu et al., 2006; Li et al., 2010, 2011; Wang et al., 2010; Irene, 2018; Begum et al., 2019). These supports allow easy access to the NMP surface along with substrate and product exchange. Such polymeric supports also facilitate tuning the catalytic activity via an external stimulus such as temperature, pH, and salt concentration (Lu et al., 2006; Li et al., 2010; Li et al., 2011; Wang et al., 2010; Irene, 2018). However, complicated preparation methods of such stimuli-responsive catalysts make them unattractive (Li et al., 2010; Li et al., 2011). Lack of complete control over activity also hinders the wide acceptability of such responsive catalysts (Lu et al., 2006; Kästner and Thünemann, 2016; Li et al., 2011; Irene, 2018). Considering the potentials of responsive supports, a simplified method needs to be developed to improve the catalytic activity, controllability, and selectivity of embedded NMPs.Herein, a simple water-in-oil emulsion route is reported for the preparation of AgNP embedded responsive hydrogel microsphere catalyst. Such catalyst was prepared by the emulsification of an aqueous solution of Trivinylphosphine oxide (TVPO), Piperazine, and AgNO3 followed by simultaneous Michael addition crosslinking of TVPO-Piperazine and simultaneous in-situ formation of AgNPs within the emulsified micro-droplets. The composite microspheres were characterized by X-Ray diffraction (XRD), Scanning electron microscopy (SEM), and X-ray photoelectron spectroscopy (XPS). Dynamic light scattering (DLS) analysis confirmed the pH-responsive behavior of the composite hydrogel-microsphere. pH-responsive swelling-deswelling of the composite microsphere was utilized to control the access to AgNPs and modulate their catalytic activity ( Fig. 1). This feature was investigated using 4-NP, Congo red (CR), and Methylene blue (MB) as model pollutants. Switching ON-OFF of the catalytic activity was demonstrated by changing the pH of the reaction medium. Substrate selectivity of the novel catalyst was also investigated using a mixture of 4-NP and MB. Finally, the reusability of the catalysts was demonstrated to highlight their practical application potential.AgNO3 (≥ 99.0%), Sodium borohydride (NaBH4), MB, CR, 4-NP, Phosphoryl trichloride, vinyl magnesium bromide (1 M in THF), dry THF, piperazine, and Span 80 were purchased from Sigma-Aldrich and used as received. Trivinylphosphine oxide (TVPO) was synthesized by a reported procedure and confirmed by NMR analysis (Nazir et al., 2020b).Hydrogel was prepared by Michael addition of TVPO (64.0 mg, 0.50 mmol) and piperazine (64.6 mg, 0.75 mmol) in 2.5 mL of water (40 °C for 16 h) (Nazir et al., 2020b). pH-responsive swelling of the transparent hydrogel was determined by measuring the swelling ratio (SR) by the procedure reported in our previous publication (Nazir et al., 2020b) and details can also be found in Sec. S1.1.2, 5, and 10.5 mol% of AgNO3 was added to 2.5 mL aqueous solution of TVPO (64.0 mg) and piperazine (64.6 mg) kept in an ice bath. The solutions were then transferred to cuvettes, sealed, and transferred to an oven (at 40 °C). All operations were carried out in dark and separate cuvettes were used for each duration. UV–vis spectra of solutions were recorded at intervals.For the preparation of the responsive catalyst, the process was developed to prevent the AgNP formation on the surface of microspheres. Therefore, AgNO3 was not premixed with the Michael adducts, rather it was added after the initiation of gelation. As shown in Fig. 2a, 0.5 g of Span 80 was mixed with 20 mL of cyclohexane using Ultraturax (19,000 rpm, 5 min). Then, a freshly prepared 5 mL aqueous solution of TVPO (128 mg, 1 mmol) and piperazine (129.2 mg, 1.5 mmol) was added dropwise to this mixture under stirring (19,000 rpm) to obtain a milky emulsion. Then, the centrifuge tube was covered with aluminum foil and transferred to a water bath under stirring (40 °C, 500 rpm). After 15 min, 0.1 mL aqueous solution of AgNO3 (48 mg/mL) was added to the emulsion and stirred for 16 h to obtain a light brown emulsion (Fig. S2). Then, the emulsion was dialyzed (24 h) using ethanol as dialysate and 15 KD RC dialysis tubes. The dialysate was replaced 3 times followed by dialysis in deionized water (24 h) to replace ethanol with water. This sample is named E1 (Fig. 2a) and the emulsion prepared without the addition of AgNO3 is named E0.Emulsion of TVPO (128 mg, 1 mmol) and piperazine (129.2 mg, 1.5 mmol) was prepared in a pressure reactor by the procedure as described in Section 2.3.1, and the pressure reactor was transferred to a water bath at 40 °C under stirring (15 min). 0.1 mL aqueous solution of AgNO3 (48 mg/mL) was added to the emulsion and stirred for 2 h (Fig. 2b). Then, the reactor was pressurized with H2 (~2.5 bar) and the stirring was continued for 14 h at 40 °C. The emulsion was then purified by dialysis as reported in Method 1 and the sample was named E2.UV–vis spectra for in-situ AgNP formation and catalytic reaction were recorded using a Varian Cary 50 UV–Vis Spectrophotometer. Catalytic reduction of 4-NP, MB, and CR were analyzed by recording the UV–vis spectra of the reaction solution as a function of time and the residual concentration was determined using their corresponding UV–vis calibration curves.NMR analysis was carried out Bruker AV-III 400 NMR spectrometer (Bruker Biospin AG, Fällanden, Switzerland). The 1H, 13C, and 31P NMR spectra were recorded using Bruker standard pulse on a 5 mm CryoProbe™ equipped with z-gradient employing 90° pulse lengths of 11.4 µs (1H), 10.0 µs (13C), and 16.0 µs (31P).ICP-OES analysis was used to determine the silver content (Ag-content) of samples. ICP-OES 5110 (Agilent, Basel, Switzerland) apparatus was used for these experiments. Samples preparation for ICP-OES consisted of mixing 10 mg of the sample with 3 mL HNO3, followed by the digestion at 250 °C for 30 min using microwave heating.XRD analysis was carried out in a Stoe IPDS-II instrument, operating at a voltage of 50 kV and a current of 40 mA with Mo Kalpha radiation (λ = 0.71073 Å) at an angular range (2θ) of 5–50°. The instrumental contribution was taken into consideration by measuring the diffraction pattern of LaB6 as the reference material and used within Topas software (Coelho, 2018).DLS analysis was carried out in a Malvern Zetasizer Nano ZS90 at 25 °C to determine the particle size of hydrogel-microspheres. Before analysis, the pH of the dispersion was adjusted to the desired value and kept for 1 h at 25 °C to achieve an equilibrium swelling. The same samples were used to measure the ζ-potential using folded capillary Zeta Cell DTS1070.SEM analysis was carried out in a Hitachi S-4800 SEM equipment operating in scanning and transmission mode (30 kV). For SEM images, samples were prepared by putting a drop of emulsion on a silicon wafer and evaporating the water at room temperature for 16 h followed by 7 nm Au/Pd coating. For transmission images, the sample was prepared by putting a drop of emulsion on a lacey carbon grid and evaporating the water over 16 h at room temperature.XPS analysis was carried out on a Physical Electronics (PHI) Quantum 2000 X-ray photoelectron spectrometer equipped with a monochromatized AlKα source (at 15 kV, 28.8 W), and a hemispherical electron energy analyzer fitted with a channel plate and a position-sensitive detector. The sample was analyzed with an electron take-off angle of 45° and spectra were recorded with constant pass energy mode (46.95 eV and energy resolution of 0.95 eV). Spectra were processed with PHI MultiPak.14.4 mg of 4-NP (0.1 mmol) and 35.4 mg of NaBH4 (0.94 mmol) were dissolved in 12.0 mL water (40 °C) maintained at different pH. The reduction was initiated by adding 1.5 mL of E2 (at the same pH) to the freshly prepared 4-NP-NaBH4 solution under stirring (500 rpm). The reduction of 4-NP was monitored by measuring the UV–vis intensity of 4-NP centered around 400 nm after required dilution. Separate samples were prepared for each duration. TOF of the reaction was calculated as the moles of 4-NP reduced by a mole of Ag-atom in an hour.For MB hydrogenation, 60 μL of MB solution (10.0 g/L) and NaBH4 (35.4 mg, 0.94 mmol) were added to 12.7 mL of water at desired pH. Then, 0.75 mL of E2 was added to the solution and the reaction was monitored by UV–vis spectroscopy after required dilution.For CR reduction, 60 μL of CR solution (20.0 g/L) and NaBH4 (35.4 mg, 0.936 mmol) were added to 12.75 mL of water at desired pH followed by the addition of E2 (0.75 mL) under stirring. Then the catalytic reduction of CR was monitored by UV–vis spectroscopy. Both MB and CR reduction were carried out at 40 °C and separate samples were prepared for each duration.The hydrogels were synthesized via previously reported Michael addition crosslinking of TVPO and piperazine (at 40 °C) in presence of 2, 5, and 10.5 mol% AgNO3 (Nazir et al., 2020a, 2020b). Colorless solutions turned dark red with time, indicating the formation of AgNPs ( Fig. 3 and S1). The appearance of a UV–vis band of nanosilver at λmax 380 nm just after 5 min confirmed the formation of Ag0 nuclei at the early stage (Fig. S1d). The redshift of the band with time indicated the growth of nuclei to AgNPs (Agnihotri et al., 2014; Zhao et al., 2013). The presence of AgNO3 led to faster gelation (Fig. S1a-c) and the solution containing 10.5 mol% of AgNO3 formed the gel within 45 min compared to 16 h for pristine hydrogel (at 40 °C). Fast gelation in the case of precursor solution containing AgNO3 can be attributed to physical reasons. In-situ nanoparticle formation (Mishra et al., 2014; Hoogesteijn von Reitzenstein et al., 2016) and crosslinking of precursors lead to a rapid rise in viscosity compared to pristine precursor solution. A combination of these factors reduces the flow behavior and the solutions behave like a hydrogel.As explained earlier (Fig. 1), the swelling-deswelling behavior of the composite hydrogel is important to control the catalytic activity. As a measure of responsiveness, the swelling ratio (SR) of AgNP-hydrogel composites was determined (Eq.S1, Sec. S1.1). From Fig. 3b it can be seen that a slight decrease in deswelling was observed in composites prepared with 1.0 and 2.0 mol% of AgNO3. However, composite prepared with 10.5 mol% AgNO3 displayed only limited deswelling, which can be attributed to the formation of the rigid matrix due to the presence of a large number of AgNPs. Finally, gel-phase NMR analysis of composite hydrogel (Fig. S2) confirmed the absence of any side reaction in presence of AgNO3. Details of NMR analysis can be found in Sec. S2.Direct synthesis of AgNP-hydrogel composite from the precursors (TVPO and piperazine, silver salts) has simplified the catalyst preparation. The use of bulk composite as a catalyst can lead to poor catalytic activity due to diffusion limitations. To overcome such drawbacks and facilitate easy access to catalytic sites, the AgNP-hydrogel composite was prepared in the form of microspheres by an emulsion process (Fig. 2). Considering moderate Ag-content and minimum loss in swelling behavior, hydrogel-microspheres were prepared using 2.0 mol% of AgNO3.The absence of AgNPs on the surface of the microsphere is important to achieve complete control over the catalytic activity of AgNPs by swelling-deswelling. Therefore, the method was developed to prevent the formation of AgNPs on the surface of hydrogel-microspheres. Particularly, premixing of AgNO3 with the Michael addition precursors was avoided, rather AgNO3 was added to the emulsion after initiation of gelation within droplets to facilitate the formation of AgNPs a few nanometers below the surface. Hydrogel-microspheres without AgNPs (E0) and with AgNPs (E1) were prepared by this method (Section 2.3.1). The light brown color of E1 indicates the presence of AgNPs within microspheres (Fig. S3). The solid content of the purified E1 in water was found to be 0.9% with an Ag-content of ~0.35 wt% on dried E1 (Table S1), which is significantly lower than the calculated Ag-content of 1.1%. This can be attributed to the fast gelation of Michael precursors within the droplets. This leads to the formation of tertiary amines before the complete reduction of AgNO3. The reduction potential of tertiary amines is known to be lower than the secondary amines (Piao et al., 2011). As a result, unreduced AgNO3 is removed from E1 during the subsequent purification step.To enhance the Ag-content in the hydrogel-microspheres, H2 was introduced in the reactor (~2.5 bar) 2 h after the addition of AgNO3 (E2, Fig. 2b), and the emulsion was stirred for a further 14 h. H2-assisted method (E2) resulted in a much darker emulsion than E1 (Fig. S3), indicating higher Ag-content in E2. UV–vis spectra of purified E0, E1, and E2 were recorded after equal dilution in water ( Fig. 4a). In the case of E0, UV-absorption was observed only below 320 nm, which is consistent with our previous report (Nazir et al., 2020b). This makes easy detection of embedded AgNPs within E1 and E2 hydrogel microspheres. The absorption band of AgNPs was visible at ~415 nm (Parida et al., 2020; Sirohi et al., 2019), owing to the transparency of the hydrogel matrix within this wavelength range. The higher intensity of the band at 415 nm for E2 indicates a higher concentration of AgNPs within microspheres.The formation of AgNPs within microspheres was also confirmed by powder XRD of dried E1 and E2 (Fig. 4b). Diffused diffraction pattern of hydrogel (E0) was present in the XRD patterns of both E1 and E2. Diffraction peaks at 2θ of 17.7°, 20.4°, 29.0° and 33.5° in case of E1 and E2 can be assigned to (111), (200), (220), and (310) reflections of a face-centered cubic Ag0 crystals (Parida et al., 2020). The size of AgNPs determined using 111-plane was found to be ~8 nm for both E1 and E2. Then, Ag-content in dried E2 was found to be 0.7 ± 0.04 wt%, which is close to the calculated value (i.e. 1.1%, Table S1). Therefore, only E2 was selected for further characterization and catalytic study. From the solid content and ICP-OES analysis (Table S1), the Ag-content of E2 was calculated to be 0.059 mg/mL.The pH responsiveness of E2 was determined by DLS analysis at different pH values. At pH 4, the particle size was 980 nm and ζ-potential of + 34 mV indicates its good dispersion stability at this pH ( Fig. 5a, Table 1). A decrease in the particle size (807 nm) and ζ-potential (+13 mV) was observed with an increase in the pH to 7. This led to poor dispersion stability of E2 and precipitation after ~2 h (Fig. S4). To compare the swelling behavior of E2 with the bulk composite hydrogel, the ratio of particle volume at pH 4 and pH 7 (VpH4/VpH7) was calculated using the average particle size obtained in DLS experiments. VpH4/VpH7 of E2 was found to be 1.8, which is in agreement with the ratio of SR of the bulk hydrogel at pH 4 and pH 7 (SRpH4/SRpH7 =1.9).Increasing the pH to 10 resulted in a decrease of the ζ-potential to − 6 mV (Table 1), indicating the unstable nature of the dispersion (Fig. S4), and two populations of particles were observed during DLS analysis (Fig. 5a). Although the decrease in particle size of E2 is expected at pH 10, the larger particle size observed at this pH is due to the aggregation of smaller particles. Two particle populations make it difficult to compare the change in volume at pH 10 (Fig. 5a). Further increasing the ≥ pH 12, an extensive aggregation was observed with unreliable DLS results.SEM analysis of E2 was carried out both in scanning and transmission mode to determine the location of AgNPs in microspheres. As expected, no sign of AgNPs was observed during SEM analysis of E0 (Fig. 5b). SEM images of E2 showed the presence of AgNPs a few nanometers below the surface of the microsphere (Fig. 5c). Transmission images also confirmed the presence of AgNPs (13 ± 3 nm, Fig. 5d, and e) few nanometers within the microsphere. The size of AgNPs determined by transmission images was found to agree with XRD analysis (~8 nm). It is worth mentioning here that, absence of AgNPs on the surface of microspheres prevents any catalytic activity at its deswollen state, which offers excellent control over the activity just by swelling and deswelling of the E2.XPS analysis was carried out to determine the oxidation state of the silver in E2 (Sirohi et al., 2019; Ruíz-Baltazar et al., 2018; Cheng et al., 2015). The high-resolution scan in Fig. 5f shows the presence of two peaks between 365 and 377 eV due to spin-orbital splitting to Ag 3d5/2 and Ag 3d3/2 core levels. Both the peaks can be resolved into two components, with major peaks at 368.2 (3d5/2) and 374.2 eV (3d3/2) assigned to Ag0. The small peaks at 369.6 and 375.3 eV were assigned to oxidized silver (Ag+). Based on the peak area, the Ag0 component was found to be ~95%. Analyzing the N1s spectra, an overall shift towards lower binding energy was observed in E2 as compared to E0 (Fig. 5g). This can be attributed to an increased number of secondary amines in E2 than E0, which is in agreement with the NMR analysis. Analysis of N1s spectra confirms the absence of any AgNPs-hydrogel chemical interactions, and AgNPs are physically stabilized within the hydrogel.To study the possible reduction of 4-NP in the absence, aqueous solutions of 4-NP and NaBH4 at different pH were prepared (Section 2.5.1) and the solutions were monitored by UV–vis spectroscopy (Fig. S6a). No change in the intensity of 4-NP absorption peak (at 400 nm) was observed over time. This indicates a lack of 4-NP reduction in absence of E2 at all pH ranges (Table S2). Adding the required quantity of E2 to the reaction solution, the yellow color started to disappear ( Fig. 6a) along with a decrease in the intensity of the 4-NP UV-absorption peak and the appearance of a new band at ~300 nm corresponding to 4-AP (Fig. 6b). The conversion and rate of reaction were determined from the intensity of the UV-absorption peak at 400 nm and given in Fig. S6b and Fig. 6c. C 0 and C are the initial and residual concentrations of 4-NP at a given time, respectively.Linear correlation between −ln(C/C 0 ) vs. time indicates the pseudo-first-order reduction of 4-NP at all pH values (Fig. 6c), with a visible effect of pH on the rate of hydrogenation of 4-NP (Fig. 6c). The highest k value at pH 4 (0.11 min−1) followed by a decrease with an increase in pH (Fig. 6c) indicates the effect of deswelling of hydrogel matrix surrounding the AgNPs. Increasing the pH ≥ 12, a very low k value (0.0006 min–1) was observed, which signifies the lack of 4-NP hydrogenation (Fig. 6c and S6b). These observations, highlight the ability to control the activity of AgNPs by controlling the swelling of the microspheres. TOF of the 4-NP hydrogenation (at pH 4) was determined at different reaction duration (Fig. S6c) and was found to decrease with time, which is in agreement with earlier reports (Kozuch and Martin, 2012). TOF at full conversion was found to be ~170 h−1 and ~100 h−1 for pH 4 and 7 respectively ( Table 2). TOFs achieved by E2 at full conversion are higher than the recently reported state of art Au (Qin et al., 2019; Fu et al., 2018) and Ag (Zhou et al., 2020; Mao et al., 2018) based catalysts.Based on the swelling behavior of composite microsphere and catalytic activity at different pH, a mechanism to explain the tunable behavior is proposed in Fig. 7. As observed during DLS analysis, E2 reached a swollen state around pH 4, which coincides with the pH of the highest catalytic activity. This signifies the easy accessibility of AgNPs for the hydrogenation of 4-NP. Positive ζ-potential of E2 at this pH (Table 1) also favors the adsorption of 4-NP and BH4 - on the catalyst. As a result, the local concentration of 4-NP and BH4 - around the micro-environment of AgNPs remains high and facilitates high catalytic activity. A similar strategy has been used to enhance the catalytic activity of Palladium/MOF catalysts for styrene hydrogenation (Huang et al., 2016). Additionally, pores of the hydrogel can also act like nano-reactors to enhance the activity via the well-known confined space reaction (Fig. 7) (Gong et al., 2019; Cárdenas-Lizana et al., 2013). A combination of these factors led to fast hydrogenation of 4-NP. Increasing the pH of the reaction medium decreased the access to AgNPs due to the deswelling of the E2 (Fig. 7a). The decrease in ζ-potential (Table 1) with pH also reduces the local concentration of 4-NP. As a result, the catalytic reduction of 4-NP slowed down. Complete deswelling along with negative ζ-potential of E2 above pH ≥ 12 block the access to AgNPs and, prevents any catalytic reduction (Figs. 6c, 7a). E2 was also used for the reduction of pollutants of anionic (Congo red) and cationic (methylene blue) nature. The reduction pathway of carcinogenic azo dye like CR by AgNPs in the presence of NaBH4 and AgNPs is well established and shown in Fig. S7a (Rajesh et al., 2014; Nasrollahzadeh et al., 2020). The residual CR in water is quantifiable by recording the UV–vis intensity of CR solution at different times (Fig S7b). Hydrogenation of CR under acidic medium (pH 4) was fast (k = 1.8 min–1) with a TOF of ~124 h–1 ( Fig. 8a, Table 2) highlighting the excellent activity of E2. The activity decreased with an increase in the pH and activity was turned OFF at pH 13 (Fig. 8a, S7c). Such behavior resembles the reduction of 4-NP and can be explained by Fig. 7a. In our previous study, we have shown that under an acidic medium this hydrogel favors adsorption of anionic dyes (Nazir et al., 2020b). Changing the pH to alkaline results in the repulsion of dye molecules from the hydrogel matrix (Nazir et al., 2020b). This behavior favors the adsorption of CR and faster degradation under an acidic medium. Repulsion between CR and hydrogel microsphere along with deswelling of E2 prevents any degradation above pH 12.During MB reduction, the highest catalytic activity was observed at pH 4 (Fig. 8b, S8) with a k value of 0.52 min–1, which is lower than the k-value observed during the reduction of 4-NP and CR. It is due to the decreased adsorption of cationic MB by the positively charged E2 (Fig. 7b). As a result, the TOF of 27 h–1 was observed during MB reduction at pH 4 (Table 2). On the contrary to 4-NP and CR hydrogenation, it was not possible to completely stop the hydrogenation of MB at pH 13. This can be explained as the adsorption and slow diffusion of MB molecules to reach AgNPs for catalysis owing to the negative surface charge of E2 at pH 13 (Figs. 7b and 8b).Differential catalytic reduction of cationic and anionic molecules by E2 at pH ≥ 12 prompted the investigation of the substrate selective activity of the novel catalyst. A mixture of MB to 4-NP (3 × 10–4 mmol) maintained at pH 13 was subjected to hydrogenation. The green color of the solution slowly turned yellow ( Fig. 9c), suggesting the selective reduction of only MB. UV–vis spectra of the solution also confirmed the hydrogenation of only MB (Fig. 9a, b). Interestingly, the concentration of 4-NP remained unchanged even after 3 h. This can be attributed to the selective and competitive adsorption of cationic MB molecules on E2, due to the negative surface charge at this pH.To demonstrate the practical applicability, the catalyst E2 was subjected to reuse cycles using 4-NP as a substrate. E2 maintained its activity even after the fifth cycle (Fig. 9d) and activity could be turned OFF at the third cycle by raising the pH to 12. E2 recovered its activity in the fourth cycle (at pH 4.5), highlighting the reversibility of its activity. No loss in catalytic activity after the fifth reuse cycle indicates easy recovery of E2 and low loss of silver. Ag-content of 0.65 ± 0.06% determined by ICP-OES analysis after the fifth cycle confirmed a very low Ag loss (Table S1) and minimizes the risk of secondary pollution due to silver leaching from E2. Transmission images indicate hydrogel matrix successfully prevents agglomeration of AgNPs (Fig. S9), although some degree of agglomeration of hydrogel microspheres can be observed in Fig. S9. Agglomeration of microspheres can take place during sample preparation on a TEM grid or during catalyst recovery. From the catalyst reuse study, it is clear that such agglomeration is temporary and has no effect on its catalytic activity.TOF of the novel catalyst (E2) calculated after 4-NP reduction (at pH 4) is superior to most of the recently reported AgNP-catalysts ( Table 3, Entry 2–6). In the case of entry 7, excellent catalytic activity was reported for AgNP based catalysts, which is attributed to their ultrafine particle size and porous support. The catalyst prepared from an alloy of gold and silver also displayed lower TOF than E2 (entry 8). AuNP based catalyst reported in entries 9 and 10 displayed significantly higher activity than E2, because of porous support that provides easy access to AuNPs and inherently high activity of AuNPs. Interestingly, E2 requires a significantly lower quantity of NaBH4 to achieve such high TOF compared to highly active catalysts listed in Table 3. Additionally, ease of controlling the activity and catalytic selectivity are distinct advantages of E2 compared to the reported catalysts.In summary, this work demonstrates a simple method to prepare AgNP based pH-responsive catalysts with excellent catalytic reduction of pollutants like 4-NP and dyes (CR and MB). The pH-responsive hydrogel support offers the possibility to control the access to AgNPs and the micro-environment around them, thereby tuning the catalytic activity. At its swollen state, AgNP-hydrogel-microsphere (E2) displayed an excellent catalytic reduction with TOF of 170 h–1 and 124 h–1 for 4-NP and CR respectively. Increasing the pH of the reaction media resulted in a decrease in activity and was turned OFF once the pH was above 12. The composite catalyst also displayed a selective reduction of only MB at pH 13 from a mixture of MB and 4-NP, owing to the selective adsorption of MB (cationic pollutant) due to negative ζ-potential at this pH. Successful recyclability of this novel catalyst makes it suitable for various practical applications. Additionally, the substrate selectivity of this catalytic system can be utilized for further development of selective catalysts using other metals. Dambarudhar Parida: Conceptualization; Methodology; Investigation; Data curation; Formal analysis; Supervision; Validation; Visualization; Writing - original draft; Writing - review & editing. Eva Moreau: Investigation; Data curation; Formal analysis. Rashid Nazir: Investigation. Khalifah A. Salmeia: Formal analysis; Writing - review & editing. Ruggero Frison: Investigation. Ruohan Zhao: Investigation. Sandro Lehner: Investigation. Milijana Jovic: Investigation. Sabyasachi Gaan: Methodology; Project administration; Resources; Supervision; Writing - review & editing.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.The authors thank Dr. Daniel Rentsch for the NMR analysis. We appreciate the support of Mr. Ton Markaj for ICP-OES analysis and the support of Dr. Roland Hauert for XPS analysis. The NMR hardware was partially granted by the Swiss National Science Foundation (SNSF, grant no. 206021_150638/1). The manuscript was written through the contribution of all and all authors have approved the final version of the manuscript.Supplementary data associated with this article can be found in the online version at doi:10.1016/j.jhazmat.2021.126237. Supplementary material. .

A simple method is reported for the preparation of silver nanoparticle (AgNP) embedded pH-responsive hydrogel microparticle catalyst via Michael addition gelation and in-situ silver nitrate (AgNO3) reduction. The AgNP-hydrogel microsphere exhibited an efficient reduction of pollutants like 4-Nitrophenol (4-NP) and Congo red (CR) under acidic medium with turn over frequency (TOF) of ~170 h–1 and ~124 h–1 respectively. Interestingly, the activity of the catalysts was turned-OFF under a basic medium (≥ pH 12) due to the deswelling pH-responsive matrix surrounding the AgNPs. On the contrary, turning-OFF the hydrogenation of a cationic pollutant like methylene blue (MB) using high pH (≥ 12) was not possible, due to ionic interaction of MB molecules with the negatively charged catalyst at this pH. This feature was used to demonstrate selective hydrogenation of only MB from a mixture of 4-NP and MB. Finally, five recycling steps confirmed the reusability and practical application potential of the catalyst.

Hydrocracking of long-chain paraffins into transportation fuels constitutes one of the most versatile processes in the petrochemical industry. It involves the cracking and isomerization of hydrocarbons and is primarily used to obtain high-quality middle distillates. Hydroisomerization refers to processes in which the branching occurs with very limited cracking. It is used to improve the cold-flow properties of diesel fuel and to obtain high-octane gasoline blending components and lube oils with good cold-flow properties. These hydroconversion processes follow a bifunctional mechanism in which a dehydrogenation/hydrogenation (metal or metal sulphide sites) function is combined with acid-catalyzed isomerization and cracking functions [1].The hydrocarbon product distribution depends amongst other on the balance and (spatial) intimacy of these two functions, giving rise to “ideal” vs. “non-ideal” hydrocracking. Shape selectivity refers to the influence of the pore size in which the cracking reactions take place [2–4]. In the classical interpretation of the hydrocracking mechanism [5,6], n-alkane is dehydrogenated on a metal site to the corresponding n-alkene, which desorbs and diffuses to Brønsted acid sites where it further reacts via carbenium ion chemistry. This includes skeletal rearrangements and β-scission (carbon‑carbon bond cleavage) reactions. The rate of the ß-scission step depends on the degree of isomerization previously attained, i.e., alkylcarbenium intermediates with a larger number of branches will crack faster. The product of the primary β-scission reaction can undergo additional transformations, including a secondary ß-scission event if diffusion of the intermediate to hydrogenation sites is too slow. Thus, desorption, diffusion and the relative location of the hydrogenation to the acid function play a role in the distribution of the cracked products. Typical hydrogenation functions include metals such as Pt, Pd and Ni, because they provide sufficient dehydrogenation activity, although with a tendency to paraffin hydrogenolysis in the case of Ni and Pt [7,8]. The acidic component is usually provided by a zeolite or amorphous silica-alumina. While these solids can also serve as a support for the hydrogenation function, composite catalysts may also contain other supports such as γ-alumina on which the metal phase can be dispersed. Strongly acidic zeolites are favoured for hydrocracking purposes, while supports with a limited acidity are more selective towards isomerization. Zeolites contain acid sites located in micropores with dimensions in the order of the size of hydrocarbons to be converted, inducing specific shape selectivity. Amorphous silica-alumina typically has larger mesopores, which are arranged in a random manner compared to the micropore topology of zeolites. Ordered mesoporous silicas such as MCM-41, which are acidic by introducing aluminum in the amorphous silica network, have also been used for hydrocracking purposes to establish the effect of a uniform pore size and ordered pore arrangement on cracking reactions [9]. The industrial relevance of such ordered mesoporous materials is, however, limited because of the high cost associated with the organic templates used in their preparation as well as their low hydrothermal stability. While MCM-41 presents a tubular pore system, MCM-48 exhibits a three-dimensional system of mesopores. SBA-15 is another ordered mesoporous silica with a similar hexagonal structure as MCM-41 but with larger pores, thicker silica walls and, therefore, higher hydrothermal stability [10,11].In the present work, we compared the performance of different ordered mesoporous silicas (SBA-15, MCM-41, MCM-48) and amorphous silica-alumina (ASA), which contains disordered mesopores, as acidic supports for the hydroconversion of n-hexadecane (n-C16) in a trickle-bed microflow reactor. The aim was to understand the influence of the size and order of the (meso)pores on the product distribution. Acidity was introduced into the ordered mesoporous silicas by aluminum either in the synthesis step or by post-synthesis alumination of the calcined silica materials. The acidic, textural and morphological properties of the porous materials were characterized by elemental analysis, X-ray diffraction (XRD), N2 porosimetry, transmission electron microscopy (TEM), solid state NMR spectroscopy and IR spectroscopy. The Pd metal phase was characterized by H2 chemisorption and IR spectroscopy of adsorbed CO.A range of SBA-15 samples with a target Si/Al ratio of 20 were synthesized under acidic conditions, employing HCl solutions with pH 1.0, 1.5, 1.6, 1.7, 2.0, and 2.5. In a typical procedure based on a literature recipe [12], a solution A was prepared by dissolving 2 g Pluronic P123 in 70 ml in the acid solution followed by stirring at 313 K for 6 h. A second solution B was obtained by dissolving 0.22 g aluminum triisopropoxide and 3.2 ml tetramethylorthosilicate (TMOS) in 5 ml of the acid solution, followed by stirring at room temperature for 2 h. Solution B was then added to solution A, followed by further stirring at 313 K for 20 h. The resulting suspension was transferred to a Teflon-lined stainless-steel autoclave, which was then sealed and heated at 373 K for 48 h. The resulting materials are denoted according to the pH of the starting solution as P1.0, P1.5, P1.6, P1.7, P2.0 and P2.5. An siliceous MCM-41 sample was prepared according to literature [13], by dissolving 3.8 g tetramethyl ammonium hydroxide solution (TMAOH, 25% wt in water) and 4.6 g cetyltrimethyl ammonium bromide (CTAB) in 34.1 g water. After stirring at 308 K for 1 h, 3.0 g fumed silica was added and the gel was further stirred at room temperature for 20 h. The gel was transferred to a Teflon-lined stainless-steel autoclave and heated at 423 K for 48 h. A siliceous MCM-48 sample prepared according to Ref. [14] was synthesized by dissolving 2.14 g fumed silica in 30 g a 10 wt% cetyltrimethylammonium hydroxide solution in water (CTAOH). After stirring at room temperature for 2 h, the gel was transferred to a Teflon-lined stainless-steel autoclave and heated at 408 K for 24 h. After the hydrothermal synthesis step, the solids were filtrated and washed with deionized water. All these samples were dried overnight at 373 K and the template was removed in a next step by calcination at 773 K for 10 h (SBA-15 samples) and 823 K for 6 h (MCM-41 and MCM-48 samples).The siliceous MCM-41 and MCM-48 samples were aluminated by a dry alumination method [15,16]. For this purpose, 2.0 g calcined MCM-41 was dispersed in 50 ml of dry n-hexane. Solutions containing 0.17 g (0.11 g) aluminum isopropoxide in 150 ml of n-hexane were prepared to obtain materials with Si/Al ratios of 40 (60) (denoted hereafter as M41–40 and M41–60). The MCM-41 suspension was added to the aluminum-containing solution and the mixture was stirred at room temperature for 24 h. For MCM-48, 1 g silica was directly added to a solution containing 0.06 g of aluminum isopropoxide in 50 ml of n-hexane and the resulting dispersion was further stirred at room temperature for 24 h to obtain a Si/Al ratio of 60 (sample M48–60). The solids were recovered by filtration and washed with n-hexane. The samples were dried overnight at 373 K and subsequently calcined at 823 K for 4 h. A commercial ASA (amorphous silica alumina 75/25 w/w from supplier JGC) was used as received.The elemental composition of the solids was determined by ICP-OES (Spectro CirosCCD ICP optical emission spectrometer). X-ray diffraction (XRD) patterns were recorded on a Bruker D2 Endeavor diffractometer using Cu Kα radiation with a scanning speed of 0.02 o/s in the 2θ range of 0–10 o. N2 adsorption and desorption isotherms were measured at 77 K on a Micromeritics TriStar II 3020 instrument. For this purpose, 100 mg of the sample were transferred in a glass sample tube, followed by drying at 393 K in a N2 flow overnight.IR spectra were recorded in the range of 4000–400 cm−1 using a Bruker Vertex V70v Fourier-transform infrared spectrometer. CO IR spectroscopy was used to probe the acid and metal sites. To evaluate the acidity, the sample was cooled to ~90 K and CO was introduced into the cell via a sample loop connected to a six-port valve. After each CO dosage, a spectrum was recorded. For metal site determination, a similar procedure was followed with the difference that the samples were reduced before IR characterization carried out at 303 K. Reduction was performed in pure hydrogen by heating to 673 K at a rate of 3 K/min, followed by a dwelling time of 1 h. The sample was evacuated until a residual pressure of 1 × 10−5 mbar was reached. The sample was cooled to 303 K and CO was introduced into the cell.For IR spectroscopy of adsorbed pyridine, the probe molecule was introduced from an ampoule at its vapor pressure at room temperature. After exposure of the dehydrated sample to pyridine for 10 min, the cell was evacuated and a spectrum was recorded. Further spectra were recorded after outgassing for 1 h at 423 K, 573 K and 773 K. The acidic properties were also characterized by H/D exchange of the hydroxyl groups with deuterated benzene (C6D6) followed by IR spectroscopy. In a typical experiment based on literature [17], C6D6 was dosed to the dehydrated sample compartment from a manifold until a pressure of 10 mbar is reached, which corresponds to a total amount of 0.33 mmol (sample density 7.5 mg/cm2). The sample was exposed to C6D6 for 10 s, followed by evacuation for 1 h. The final pressure was lower than 10−6 mbar. Then, a spectrum of the partially exchanged sample was recorded. This sequence was automatically repeated with exposure times of 30 s, 5 min, 10 min, 20 min, and 30 min at 303 K, 30 min at 323 K, 30 min at 373 K, 30 min at 423 K and 30 min at 523 K. 27Al nuclear magnetic resonance (NMR) spectra were recorded on a 11.7 T Bruker DMX500 NMR spectrometer operating at 132 MHz. Transmission electron microscopy (TEM) images were taken using a FEI Tecnai 20 at an acceleration voltage of 200 kV. The samples were suspended in ethanol and dispersed over a holey Cu grid coated with carbon film.H2 uptake measurements were used to determine the metal surface area in reduced catalysts. Typically, 50 mg sample was loaded in a quartz reactor. Prior to dosing, samples were reduced in flowing H2 (1 h, 673 K, 3 K/min), evacuated at 723 K for 1 h to remove chemisorbed hydrogen and cooled to 353 K under vacuum. Chemisorption analysis was then carried out at 353 K.The solid acids were loaded with 1 wt% Pd by wet impregnation with an aqueous Pd(NH3)4(NO3)2 solution. The resulting samples were calcined at 723 K in flowing air for 4 h. In order to perform n-C16 hydroconversion activity measurements, the catalyst was dried in the reactor at 1 bar and 473 K for 1 h in a He flow and subsequently reduced at 60 bar in a H2 flow. During reduction, the temperature was increased at a rate of 3 K/min to 673 K followed by an isothermal period of 1 h. Then, the temperature of the catalyst bed was lowered to 473 K and the packed bed was wetted by maintaining a liquid flow rate of 1 ml/min for 10 min. The reactor was operated at a H2/n-C16 molar ratio of 20 and a weight hourly space velocity (WHSV) of 10 gn-C16 gcat −1 h−1. The reaction temperature was increased stepwise and the reaction was equilibrated for 3 h before product sampling. The reactor effluent was analyzed by gas chromatograph equipped with an RTX-1 column and a flame ionization detector. The identification of isomers and cracked fractions was done in accordance to the elution sequence reported in the literature [18]. Due to the large number of products observed, monobranced (ex. 3-methylpentadecane or 4-methylpentadecane) and multibranched (ex. 2,13-dimethyltetradecane or trimethyltridecanes) C16 isomers were lumped as ‘C16 isomers’, while the fractions including normal paraffins from methane to n-pentadecane and their corresponding isomers were lumped as ‘cracked products’.The composition of the calcined materials as determined by ICP analysis is presented in Table 1 . The Si/Al ratios of the SBA-15 samples are lower with increasing pH of the synthesis solution. Small-angle XRD patterns (Fig. 1 , left) show that SBA-15 samples with a low Al content obtained at a pH below 2 exhibit three well-resolved peaks related to 100, 110 and 200 planes, characteristic of the p6mm hexagonal symmetry of SBA-15 [11]. From the a 0 value of about 12 nm and the pore size distribution derived from porosimetry, a wall thickness between 2.4 nm and 4 nm was determined for these samples [19]. The absence of low-angle features in the silica materials prepared at pH 2 and higher show that the samples with a higher Al content do not contain ordered mesopores. The isoelectric point of silica is around 2. This means that, when SBA-15 is prepared at a pH of 2 or higher, there are not enough positively charged protonated hydroxyl groups that can interact with the poly(ethylene oxide) groups of the Pluronic P123 mesoporogen. As a consequence, mainly disordered silica is obtained as discussed before [20,21]. In general, a pH lower than 1.5 is needed to obtain a sufficiently charged silica surface to assemble a well-ordered silica-polymer mesophase using Pluronic P123 [22].The MCM-41 sample with a low Al content (M41–60) shows the relevant diffraction peaks related to 100, 110, 200 and 210 planes of the p6mm hexagonal symmetry [13,23]. At higher Al content (sample M41–40), the intensity of these peaks is lower, indicating a lower order of mesopores. This phenomenon, which has been mentioned in the literature [24,25], is likely related to a decrease in the long-range order at higher Al loading, although the local hexagonal order of the material is largely maintained. The XRD pattern of M48–60 contains the typical diffraction peaks associated with 211, 220 and 420 planes, which is due to the Ia3d cubic structure [16]. As expected, the walls of the MCM-41 (1.3 nm) and MCM-48 (1.2 nm) samples are thinner than those of SBA-15. The physisorption isotherms of these samples were of type IV, typical of mesoporous materials (Fig. 2 ). The presence of sharp pore filling/emptying steps within a narrow p/p 0 range for SBA-15 and MCM-41 materials is indicative of uniform cylindrical pores, presenting an H1 hysteresis loop [12,23]. The MCM-41 samples prepared using CTAB as structure-directing agent normally have pores in the range of 2–5 nm with relatively closed hysteresis loops in comparison to the ones observed in SBA-15 materials in which capillary condensation is more pronounced due to the larger pore size [26,27].The NL-DFT method was used to calculate the pore size distribution from the adsorption branch of the isotherms (Fig. S1 of supporting information). Increasing pore sizes were obtained for SBA-15 samples containing more Al. For example, the Si-SBA-15 sample (P1.0) has an average pore size of 7.9 nm. Addition of more Al to the gel to arrive 0.25 wt% (P1.5) caused the pores to swell to a size of 8.8 nm. A higher Al content (i.e., comparing 0.48 wt% and 0.80 wt%, samples P1.6 and P1.7) led to pores larger than 9 nm. This behavior has been observed before [19] and can be explained by the swelling properties of the micellar arrays by isopropanol obtained by hydrolysis of the Al precursor. A comparison of the textural properties of M41–40 and M41–60 (Table 1, Fig. S1) shows that alumination by the dry-grafting method causes the pores to shrink, decreasing the pore volume and surface area [28]. The MCM-48 sample shows a sharp pore filling step in the p/p 0 range from 0.20 to 0.35, suggesting a good mesostructural order with uniform pore channels and a relatively narrow pore size distribution [29]. A weak hysteresis loop between p/p 0 values of 0.40 and 0.80 points to capillary condensation in secondary mesopores arising from interparticle voids [30,31]. ASA, on the other hand, exhibits an isotherm with a broad pore filling profile and a hysteresis loop in the p/p 0 range from 0.43 and 1.00, which points to pores with broad size distribution as might be expected for this type of material. The pore size distribution ranges between 2 nm and 60 nm. The much wider pores result in lower surface area and pore volume in comparison to the ordered mesoporous samples. Fig. 3 shows representative TEM images of the mesoporous materials. The images of samples P1.5, P1.6 and P1.7 confirm a morphology consisting of regular arrays of long cylindrical channels with a uniform hexagonal arrangement along the 100 plane [12,19,32]. The mesopores are not always running in a straight way through the matrix, but they can be curved to some extent. It is also evident that sample P1.7 (Fig. 4 -f) exhibits regions where the characteristic features of SBA-15 are not present, indicating a lower order of the mesopores. In spite of the absence of low-angle features in the XRD patterns, samples P2.0 and P2.5 still preserve areas with uniform pores, which are consistent with the shapes of the adsorption isotherms presented in Fig. 2 and the uniform pore size distribution observed in Fig. S1. Nevertheless, the ordered hexagonal pore system is not present to the same extent as in the other SBA-15 materials prepared at lower pH. The M41–60 sample (Fig. 3-i) displays mesopores arranged in a hexagonal honeycomb-like structure, separated by thin amorphous silica pore walls [9,33].The image of M41–40 shown in Fig. 3-j reveals that an irregular pore arrangement of cylindrical pores can still be observed, making apparent that the incorporation of aluminum into the framework affects the long-range order of the mesopores without affecting the mesoporous nature of the material [34]. MCM-48 (Fig. 3 k-l) has been reported to present uninterrupted channels along the 100 and 111 planes, but due to changes in the curvature of the material, no channels are observed in other directions such as the 110 plane [35].The coordination environment of aluminum in the samples was investigated by 27Al MAS NMR spectroscopy (Fig. 4). The spectra are characterized by a strong peak at 55 ppm due to tetrahedrally coordinated aluminum (AlIV), indicating that a significant fraction of Al atoms has been grafted as tetrahedral species on the silica of the mesoporous materials. Part of these Al species may also be in the silica network due to the calcination procedure. An additional weak feature at 0 ppm can be associated to Al species in octahedral coordination (AlVI) [33]. The Al speciation for the investigated materials is given in Table 1. The AlIV fraction of all the samples lies between 0.54 and 0.66. As expected, no Al was incorporated in the SBA-15 samples synthesized at pH 1 [36].IR spectroscopy of adsorbed CO at 90 K was used to determine the density and strength of Brønsted acid sites (BAS). The carbonyl stretching region of IR spectra for samples M48–60 (left) and P1.7 (right) is displayed in Fig. 5 . The band at 2138 cm−1 belongs to physisorbed CO and the one at 2156 cm−1 to CO coordinating to silanol groups. Brønsted acid sites (BAS) are represented by the band at 2174 cm−1. Previous work has shown that this feature is a composite of bands due to a small amount of strong BAS at 2178 cm−1 and a larger amount of relatively weak BAS at 2172 cm−1 [37]. The weak feature observed at 2190 cm−1 is due to weak Lewis acid sites (LAS).Deconvolution of the peak related to BAS according to a procedure described before [38] provides an estimate of the population of these two types of BAS [39,40]. Spectra of sample P1.7 at different CO coverages are given in the supporting information. Table 2 shows that the concentration of strong BAS increases with the Al content for the SBA-15 samples. The concentrations are however very low, explaining why CO-perturbed bridging OH groups at 3300 cm−1 cannot observed as a distinct feature for these samples [41]. IR spectra of adsorbed pyridine followed by evacuation at 423 K, 573 K and 773 K for the SBA-15 samples are presented in Fig. 6 . The bands at 1545 cm−1 and 1455 cm−1 are assigned to pyridine adsorbed on, respectively, BAS and LAS, while the band at 1490 cm−1 can be associated with both types of adsorbed pyridine [41,42]. The acid site concentrations derived from pyridine IR are listed in Table 2. In general, the SBA-15 samples contain small amounts of BAS in accordance with the results obtained with CO IR spectroscopy. The absence of a feature related to BAS in these spectra recorded after evacuation at 773 K shows that only minor amounts of strong BAS are present in these samples.To determine the presence of small amounts of strong BAS in a semi-quantitative manner, H/D exchange of OH groups with deuterated benzene was followed by IR spectroscopy. The OD region for samples M48–60 (left), P1.7 (center) and the OH region for P1.7 (right) are shown in Fig. 7 . All the samples present a clear peak at 2683 cm−1, corresponding to selective H/D exchange of bridging hydroxyl groups, even at short exposure times (10–30 s) for materials with high aluminum content.These samples also display an additional feature at 2632 cm−1, related with the exchange of bridging OH groups exhibiting additional electrostatic interactions to adjacent oxygens [43] or bridging OH interacting instead with hydrogen from residual water [17], which become predominant at increasing exposure temperatures for all the samples. The H/D exchange of weak silanol groups was also observed as a band developing at 2752 cm−1. The analogous bands of bridging OH groups located in the OH region at 3638 cm−1 and 3570 cm−1 [44] cannot be observed in the IR spectra because of the broad OH band at 3746 cm−1 (Fig. 6 top, right). The concentration of strong BAS was determined from the spectra after H/D exchange at 323 K for 30 min (Table 2), using a molar extinction coefficient of 2 × 106 cm/mol [17].For n-C16 hydrocracking activity, the samples were loaded with ~1.0 wt% Pd (Table 2). The atomic Pd/H+ ratios were between 1.6 and 82.2, indicating that there should be enough metal function to ensure an appropriate supply of olefins to the acid sites and to establish acid-catalyzed isomerization and cracking reactions as the rate-limiting steps [45,46]. The dispersion of the Pd metal phase was determined by H2 chemisorption (i.e., the ratio between irreversibly adsorbed hydrogen (Hirr) and the total amount of Pd [47]). We assumed a Hirr/Pds = 1 stoichiometry for the Pd nanoparticles [48]. The Pd dispersion is below 0.5 for all samples and Pd/ASA exhibits the lowest Pd dispersion. The Pd particle size distribution was determined in more detail by TEM for the SBA-15 and MCM-41/MCM-48 samples (Fig. S2). Relatively uniformly sized Pd particles were observed for these samples with average sizes ranging from 3.0 ± 0.7 nm for sample P1.7 to 11.3 ± 3.4 nm for P1.6. Based on the average pore size determined by Ar porosimetry, we estimate the fraction of Pd particles located inside the mesopores. For SBA-15, the majority of Pd particles have sizes smaller than the pore size (Table S1), except for P1.5 and P1.6 for which only, respectively, 15% and 30% of the Pd particles are smaller than the pore size. On the contrary, the Pd particles in the M41–40, M41–60 and M48–60 sample are mostly larger than the mesopores, indicating that most of the Pd particles are located on the external surface of the M41S silicas.We also studied by IR spectroscopy the adsorption of CO on metallic Pd at 303 K. Before CO dosing, the samples were reduced at 673 K and evacuated to a pressure lower than 10−5 mbar. An example spectrum of CO adsorbed on sample P1.7 (Fig. S4) shows a broad low-frequency signal between 1700 and 2000 cm−1 and a high-frequency signal between 2000 cm−1 and 2150 cm−1, generally assigned to CO bridged and linear chemisorbed to Pd, respectively [49]. Deconvolution of the IR bands was carried out following literature [50]. The peak at 1895 cm−1 has been attributed to μ2-bridge-bonded CO on the (100) plane of Pd [51]. The 1930 cm−1 peak is due to μ2-bridge-bonded CO on (111) planes and the ones at 1961 cm−1 and 1976 cm−1 relates to bridge-bonded CO on Pd particle edges and steps [52,53]. The signal located at 2074 cm−1 is assigned to linear CO bound to (111)/(111) and (111)/(100_ edges sites and the one at 2091 cm−1 is due to CO residing on corner Pd atoms [50]. Bridge-to-linear adsorbed CO molar ratios (B/L) calculated after saturation with CO are presented in Table 2 using molar extinction coefficients of 4.1 × 106 cm/mol (bridge species) and 0.36 × 106 cm/mol from literature [54]. B/L values are in line with the dispersion trend following from H2 chemisorption, i.e., samples with high dispersion present a lower amount of CO bound on bridge sites.The hydroconversion of n-C16 was carried out at a weight hourly velocity of 10 gn-C16 gcat −1 h−1, a total pressure of 60 bar and a H2/n-C16 ratio of 20. The n-C16 conversion for the investigated samples is shown as a function of the temperature in Fig. 8 . For all sample families with the exception of M41–40, the activity correlates with the Al content.The apparent activation energies fall in the range of 169–220 kJ/mol, which is higher than typical values for catalysts containing amorphous silica-alumina as the acidic component [55]. Both skeletal isomers and cracked alkanes were obtained as reaction products (Fig. S5). Overall, the highest isomers yield was obtained for the ASA-based catalyst. For the SBA-15 catalysts, the isomers yields were very similar as a function of conversion with the exception of P1.7, which presented a much lower isomers yields (more cracking). The other samples presented lower isomers yields in the order M41–40 ≈ M41–60 > M48–60. Fig. 9 (Fig. S5) shows the molar ratio between cracked (C) and total (T) products extrapolated to zero n-C16 conversion [56,57], denoted by the parameter (C/T)x→0, as a function of Al content. In ideal hydrocracking and at low conversion, monobranched isomers should be dominant products without formation of cracking products, i.e., (C/T)x→0 should approach zero. There is a strong relation between the initial formation of cracked products and the structure of the materials. (C/T)x→0 decreases with increasing Al content for the SBA-15 samples. This is counterintuitive, because hydrocracking is expected to be ideal for low acidity materials. The observed trend cannot be directly related to the Pd dispersion (cf. Table 2). Instead, we attribute the decreasing cracking tendency with increasing Al content to the loss of order of mesopores in these samples. The lowest (C/T)x→0 values are obtained for the disordered samples prepared at a pH of 2 or higher. A similar trend with respect to pore ordering can be observed for the MCM-41/MCM-48 catalysts. Especially, the catalyst presenting long cylindrical pores (M41–60) exhibits higher (C/T)x→0 than M48–60 with interconnected channels. This behavior, which is likely enhanced by the low Pd dispersion of the latter sample, relates to the longer residence of olefinic intermediate in one-dimensional ordered pore systems, resulting in enhanced cracking. Partial disorder of the mesopores will decrease the diffusion length inside these pores. We further investigated whether improving the metal function would affect the product distribution. For this purpose, we added 0.5 wt% Pt to the P1.7 sample by a follow-up impregnation procedure using an aqueous H2PtCl6*6H2O solution and calcination at 723 K in flowing air for 4 h. This catalyst is denoted as P1.7(+Pt) in Fig. 10 . For this sample, the (C/T)x→0 ratio decreased to the same values obtained for the disordered materials prepared at higher pH, indicating that in the original P1.7 sample the metal function was not strong enough relative to the diffusion lengths. Fig. 10 shows the n-C16 conversion as a function of temperature for the P1.7 and P1.7(+Pt) samples. It can be observed that the addition of Pt results in a significantly higher n-C16 conversion. The apparent activation energy was lowered from 220 kJ/mol to 154 kJ/mol, approaching the values reported in the literature [55]. Together with the much higher isomers yields observed for P1.7(+Pt), we conclude that the performance of P1.7 was limited by the metal function. We also estimated the average number of acid-catalyzed steps involved in n-C16 hydroconversion extrapolated to zero conversion (nas, x→0, Table S1), considering the selectivity of monobranched/multibranched isomers and cracked products. It takes one acid-catalyzed step to form monobranched isomers, 2 and 3 for dibranched and tribranched isomers, respectively, and typically 4 for cracking. Therefore, nas, x→0 would approach unity for an ideal hydrocracking catalyst. Conversely, a catalyst operating outside this regime, i.e. with a poor balance of hydrogenation and cracking sites or presenting severe shape selectivity effects will present higher nas, x→0 values, indicating consecutive isomerization and cracking of the more reactive multibranched isomers already at low conversion [57]. Values of nas, x→0 around 1.4 and 1.6 seem to indicate that the samples do not suffer from shape selectivity or a poor metal/acid balance. Nevertheless, the values are slightly higher for the samples with ordered mesopores.We also observed that the distribution of monobranched isomers at low and 50% n-C16 conversion is very similar for all samples (Fig. S6). At both conversion levels, the product distribution is dominated by 7-methylpentadecane (7-meC15) and 8-methylpentadecane (8-meC15) with lower amounts of other isomers. At a n-C16 conversion of 50%, nonetheless, the amounts of 7-meC15 and 8-meC15 are less pronounced and the concentration of other isomers increases.The product distribution of cracked hydrocarbons is plotted in Fig. 11 . Samples supported on supports with a well-ordered mesopore structure present higher cracking selectivity than samples with a disordered mesopore structure (i.e., ASA, M41–40, P2.0 and P2.5). In general, all Pd-loaded catalysts exhibit an asymmetric distribution of cracked products shifted to lighter products with C4-C6 being the dominant products. Cracking is more substantial for samples with ordered mesopores in comparison with samples having disordered pore systems. Adding Pt to P1.7 also leads to a lower degree of cracking and a cracked product distribution corresponding more to the one expected for ideal hydrocracking. These differences suggest that some degree of secondary cracking takes place, which associated with an increased residence time of the olefinic intermediates as discussed previously [58]. It is evident from the results that, in the case the metal hydrogenation function is not strong enough, the product distribution is influenced not only by the order of the pores, but also by the concentration of acid sites. The occurrence of secondary cracking suggests that the diffusion of hydrocarbons plays an important role in the reaction mechanism, especially inside the long cylindrical pores of SBA-15 and MCM-41 based catalysts [59].It appears also that the addition of Pt as a stronger hydrogenation function can also impact the product distribution. The observation that Pt addition also increases the overall n-C16 conversion implies that the dehydrogenation function of P1.7 is not strong enough. This may be because Pt presents a stronger hydrogenation activity or Pt is more readily dispersed than Pd. While this may be due to the low Pd dispersion in the P1.7 sample, this result shows that a too low hydrogenation function leads to overcracking in 1-dimensional mesopores. Wang et al. earlier showed the sensitivity of the product distribution on the location, size and shape of Pt nanoparticles [60].In this work we obtained insight into the role of size and order of the pores of (ordered) mesoporous silica (SBA-15, M41S and ASA) on the bifunctional hydrocracking of n-hexadecane. A series of Al-modified ordered mesoporous SBA-15, MCM-41 and MCM-48 and disordered ASA were employed as the acid component for the bifunctional catalysts, which were also loaded with Pd as the hydrogenation component. SBA-15 materials were prepared at different pH, which was found to simultaneously influence the Si/Al ratio and order of mesopores. Al was introduced in MCM-41 and MCM-48 by a post-synthesis grafting method. All materials including ASA exhibited low acidity compared to zeolites. A general trend is that increasing Al incorporation (alumination for M41S, higher pH synthesis for SBA-15) led to a loss of ordering of the mesopores. The Pd metal phase was Pd sites homogeneously dispersed as ~10 nm particles. In n-C16 hydroconversion, it was observed that secondary cracking is more pronounced for catalysts containing long one-dimensional cylindrical pores (SBA-15 and MCM-41) than for catalysts containing a three-dimensional ordered or disordered mesoporous texture, which can be attributed to the difference in residence time of intermediates in the mesopores. From the observation that secondary cracking increased for lower Pd dispersion, it is inferred that the distance of acid sites in the mesopores and the metal phase mainly located outside these pores also plays a role. This was further corroborated by adding Pt to a mesoporous Pd/SBA-15 sample, which improved the overall catalytic activity and decreased secondary cracking reactions. Ideal hydrocracking operation is approached for ASA, MCM-48, and SBA-15 prepared at a high pH with a disordered mesopore system.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This research was financially supported by Shell Global Solutions International B.V. The authors thank A.M. Elemans-Mehring for ICP-OES analysis and the Soft Matter Cryo-TEM Research Unit of Eindhoven University of Technology for access to TEM facilities. The authors thank Arno van Hoof, Tobias Kimpel and Jiadong Zhu for TEM analyses. Supporting information Image 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.fuproc.2022.107259.

The catalytic performance in n-hexadecane hydrocracking of a set of bifunctional catalysts composed of acidic mesoporous silicas, i.e., SBA-15, MCM-41, MCM-48, and amorphous silica-alumina (ASA), and Pd as acid and (de)hydrogenation components, respectively, was investigated. The selectivity to cracked products and the occurrence of secondary cracking depended on the pore topology, acidity, and Pd dispersion. The Si/Al ratio and the mesopore order of SBA-15 were modified by changing the pH in the synthesis step. Al was introduced in the M41S materials by post-synthesis grafting. All materials including ASA exhibited low acidity compared to crystalline zeolites. Increasing Al content led to a decrease of the order of mesopores. Secondary cracking of n-hexadecane was more pronounced for catalysts containing long one-dimensional cylindrical pores (SBA-15 and MCM-41) in comparison with catalysts containing three-dimensional ordered (MCM-48) or disordered (ASA) mesopores. The selectivity difference is attributed to differences in residence time of intermediates in the mesopores. The distance between acid sites located in mesopores and Pd nanoparticles primarily located outside these pores also influences the product distribution. Ideal hydrocracking operation is approached for ASA, MCM-48, and SBA-15 prepared at a high pH containing disordered mesopores.

Data will be made available on request.Supramolecular chemistry paves a bright avenue for constructing numerous functional materials with advanced applications in adsorption/separation, catalysis, sensing, biomedical technologies etc. [1,2]. For porous materials, such as zeolites [3], mesoporous silicas [4] and mesoporous organosilica nanoparticles [1,5], supramolecular host-guest systems show a particular potential.Relatively novel crystalline hybrid materials [6], metal-organic frameworks (MOFs) or porous coordination polymers, which are composed of organic molecules (linkers) and metal ions, can be regarded themselves as supramolecular systems. Under selected conditions, their self-assembly from organic and inorganic building blocks, which are polytopic ligands (linkers) and metal ions, leads to a unique MOF structure in a predictable way [7]. Furthermore, MOF modular building principle assists in constructing a large diversity of topologies, cavity sizes, and functionalities. One of most exciting MOF features is the infinity of ligands that can be used to generate them [8,9].Alongside their inorganic counterparts, zeolites, MOF matrices show a potential in the development of functional (nano)materials including supramolecular systems. To date, the significant research efforts have been directed towards design and development of host-guest structures using MOF host matrices due to their exciting properties, such as extremely high porosity, large pore volume, open porous structure and versatile chemical composition. In this way, a number of functional molecules and species can be encapsulated in the MOF host matrix, which is regarded as a porous scaffold for their accommodation [10]. These molecules can impart some novel functions in MOF materials or enhance/optimize existing intrinsic properties [11]. This approach holds great potential for catalytic applications of MOFs. Following this strategy, new catalytic properties can be introduced in the MOF porous host.The MOF-based host-guest materials show efficiency in number of catalytic applications from fine chemical synthesis to photocatalysis [12,13]. Nowadays, there are three main established approaches for the immobilization of functional molecules or nanoscale objects in MOFs, known as ‘‘ship in a bottle’’ approach [14–16], ‘‘bottle around a ship’’ approach [15,17,18], and one-step synthesis approach ( Scheme 1) [14–18]. In all cases, guest molecules are bound by weak host-guest interactions with MOF walls. The ‘‘ship in a bottle’’ strategy involves the encapsulation of guest molecules in the cavities of MOFs, followed by further treatment leading to the desired functional structure. The second “bottle around a ship” strategy involves introducing functional guest molecules in the reaction mixture containing reagents for MOF synthesis [17,19]. Supramolecular host-guest composites based on MOF host matrices have been successfully prepared with different functional inorganic [20–22] (Table S1, Supporting Information (SI)) and organic guests ( Table 1) [23–30]. Most of these composites were effective catalysts for photocatalytic reactions, hydrogenation of α,β-unsaturated carbonyl compounds, synthesis of 1,5-benzodiazepines from 1,2-phenylenediamine and ketones etc.Propylene carbonate (PC) as one of the cyclic carbonates is an important chemical, which is widely applied as a solvent, in plasticizers, lithium battery electrolytes, monomer in the preparation of polycarbonates, and the intermediate in the manufacture of fine chemicals [31]. Nowadays, the cycloaddition reaction of propylene oxide (PO) with CO2 provides an industrial route to PC using CO2 in place of phosgene with both economic and environmental benefits ( Scheme 2).This reaction can be catalyzed by homogeneous catalysts, but the problem of recyclability is inherent to homogeneous catalysts. Heterogeneous catalytic systems also were used to catalyze this reaction. Heterogeneous catalysts include tetraalkylammonium salts of transition-metal-substituted polyoxometalates, such as [(n-C7H15)4N]6[α-SiW11O39Co] and [(n-C7H15)4N]6[α-SiW11O39Mn] [32], zeolites, such as SSZ-13 [33], Ga-TS-1 [34], M/H-ZSM-5 (M - Zn2+, Fe3+, Co2+, Ni2+) [35] etc. Metal-organic frameworks (MOFs) also were demonstrated to be effective heterogeneous catalysts that can capture and convert CO2 to cyclic carbonates under mild conditions (Table S2 , SI) [36–40]. The accessible and unsaturated metal cations in the framework of MOFs are important for the high activity because of their function as Lewis acid sites (LAS) to activate the epoxide.Metalloporphyrins as ligands in MOF structures can serve as the catalytic sites for the reaction with CO2. Thus, Gao and co-workers explored a metalloporphyrin-based MOF, denoted as MMPF-9, for heterogeneous catalysis in the synthesis of PC from PO and CO2. MMPF-9 exhibited excellent catalytic performance under mild conditions (1 atm of CO2 at room temperature) [41]. The yield of PC was 87.4 % after 48 h of the reaction. Note that the activity of MMPF-9 was significantly higher than the catalytic activity of HKUST-1 (49 %). Functionalization of the linker also allows improving catalytic properties of MOFs. Thus, Ma and co-workers [42] prepared two MOFs functionalized by quaternary ammonium or phosphorus bromide ionic liquid, MIL-101-N(n-Bu)3Br and MIL-101-P(n-Bu)3Br, by post-synthetic modification of the parent MIL-101-NO2. Such approach allows the process to be carried out without a co-catalyst, despite the fact that a higher temperature and CO2 pressure are required.In our investigation we would like to draw attention to the synthesis of a novel supramolecular composite (MIL/K-OH) based on calix[4]arene with hydroxyl groups in the arene “bowl” (K-OH, Fig. 1) as a functional guest molecule and NH2-MIL-101(Al) as a porous host and investigation of its catalytic potential in synthesis of PC from PO and CO2.This study is focused on three main issues: (a) evaluation of the effect of the synthesis method, i.e., MW-assisted technique and the solvothermal procedure, in respect of structural, morphological, and textural properties of the produced MIL/K-OH composite, (b) the understanding of the role of the functional calix[4]arene molecule on the catalytic behavior of this host-guest system in the synthesis of PC, and (c) assessment of its potential for further applications in acid-base catalysis.Propylene oxide (> 98 %, Acros Organics), tetra-n-butylammonium bromide (TBABr) (Sigma-Aldrich), AlCl30.6 H2O (Aldrich), 2-aminobenzene-1,4-dicarboxylic acid (ABDC, 99 + %, Acros Organics), 1,3,5-benzenetricarboxylic acid (H3BTC) and 1,3,5-trimethyl-benzenetricarboxylate (Me3-BTC, 98 %, Aldrich), ortho-phosphoric acid (H3PO4, 85 wt %, Merck), sodium hydroxide (NaOH, 4 M), nitric acid (HNO3, 60 wt %), N,N-dimethylformamide (DMF, Aldrich) were used without any further purification. NH 2 -MIL-101(Al) sample was prepared by MW-assisted synthesis under atmospheric pressure according to [43]. 0.51 g of AlCl30.6 H2O, 0.56 g of 2-aminobenzene-1,4-dicarboxylic acid, and 40 ml of DMF were transferred into a glass ampoule and heated at atmospheric pressure in a chamber of a MW oven “Vigor” (200 W, 20 min, 130 °С). The formed solid was separated on a centrifuge and washed with DMF (3 ×10 ml) and acetone (3 ×10 ml). Then the crystalline product was treated with boiling methanol (20 ml) under stirring (24 h), isolated on a centrifuge, and evacuated at 130 °С for 7 h. Calix[4]arene K-OH (25,26,27,28-tetrahydroxycalix[4]arene) was synthesized according to [44]. Anhydrous aluminum chloride (21.0 g, 158 mmol) was added gradually to a mixture of tert-butylcalix [4]arene (20 g, 0.03 mmol), phenol (13.56 g, 0.144 mmol) and toluene (125 ml). The resulted mixture was stirred (RT, 1 h), then was poured in water (25 ml) acidified with conc. HСl (5 ml). The organic layer was separated, and the solvent was removed by distillation. Methanol was added to the residue. The obtained K-OH material was recrystallized from a chloroform-methanol mixture. The product yield was 7.40 g (56 %). 1HMR (CHCl3, 300 MGz) 3.57 (s, 4 H, CH2), 4.27 (s, 4 H, CH2), 6.75 (t, 4 H, ArH), 7.07 (d, 8 H, ArH), 10.22 (s, 4 H, ArOH). MIL/K-OH(MW) composite. A solution of AlCl30.6 H2O (1.02 g, 4.22 mmol), 2-aminobenzene-1,4-dicarboxylic acid (1.12 g, 6.18 mmol) and K-OH (0.873 g, 2.04 mmol) in DMF (70 ml) was transferred into a glass ampoule and heated at atmospheric pressure in a chamber of a MW oven “Vigor” (200 W, 20 min, 130 °С). The crude product was rinsed with DMF (3 ×10 ml) and acetone (3 ×10 ml), then dried at 130 °C for 6 h under a vacuum. The chemical composition of the MIL/Ks-OH(MW) sample is shown in Table 2. MIL/K-OH(Solv) composite. A solution of AlCl30.6 H2O (1.02 g, 4.22 mmol), 2-aminobenzene-1,4-dicarboxylic acid (1.12 g, 6.18 mmol) and K-OH (0.873 g, 2.04 mmol) in DMF (70 ml) was transferred into a Teflon autoclave and heated in a thermostated oven (130 °С, 72 h). The crude product was rinsed with DMF (3 ×10 ml) and acetone (3 ×10 ml), then dried at 130 °C for 6 h under a vacuum. The chemical composition of the MIL/K-OH(Solv) sample is shown in Table 2.C, H, N, and S analyses of the synthesized materials were performed using a Euro EA Elemental Analyzer.The porous structures of samples were determined from the adsorption isotherm of N2 at − 196 °C using an “ASAP 2020 Plus” (“Micromeritics”) instrument. The specific surface area (SBET) was calculated from the adsorption data over the relative pressure range between 0.05 and 0.20.X-ray powder diffraction (XRD) data were collected in a reflection mode using a Panalytical EMPYREAN instrument with a linear X′celerator detector and non-monochromated Cu Kα radiation (λ = 1.5418 Å), measurement parameters: tube voltage/current 40 kV / 35 mA, divergence slits of 1/16 and 1/8°, 2θ range 2–30°, speed 0.1° min−1. High-resolution synchrotron measurements (λ = 0.45085 Å) were carried out at room temperature at beam line ID22 of the European Synchrotron Radiation Facility (ESRF, Grenoble, France).DRIFT spectra were recorded using a Shimadzu FTIR-8300S spectrometer with a DRS-8000 diffuse reflectance cell in the range between 400 and 6000 cm–1 with a resolution of 4 cm−1. All spectra are presented in F(R) Kubelka-Munk scale: F ( R ) = ( 1 − R ) 2 2 R , where R is the reflection coefficient.The basicity of ZIFs was determined by DRIFT spectroscopy with CDCl3 as a probe molecule using the technique reported in Ref. [45] (SI).Electron microscopy investigation (TEM) was performed using a JEM-2200FS (JEOL, Tokyo, Japan) electron microscope operating at 200 kV with a lattice resolution of 0.1 nm.Before the reaction, all catalysts were activated at 150 °C for 2 h in air in order to remove adsorbed water. The standard procedure was as follows: the cycloaddition of CO2 to propylene oxide was carried out in a 30 ml stainless-steel autoclave. Typically, 24 mmol of PO, 50 mg (0.85 mmol based on PO) of a catalyst, 50 mg (0.85 mmol based on PO) of TBABr were loaded into the autoclave. The autoclave was purged 3 times with CO2 to remove air and was charged with CO2 up to 0.8 MPa. Then the autoclave was heated to 80 ◦C (the heating time was 15 min). The mixture was kept under stirring for 5 h. After the reaction completion, the autoclave was cooled, and excess CO2 was released. The product was diluted with toluene and analyzed with a gas chromatograph (Agilent Technologies 7820 A) equipped with a capillary column (Agilent HP-5).The chemical composition of the NH2-MIL-101(Al) sample and MIL/K-OH composites is shown in Table 2. It can be seen from these data that experimental and theoretical weight amounts of C, H, N elements in the NH2-MIL-101(Al) are close. The composite synthesis method dictates remarkably the K-OH content in the NH2-MIL-101(Al) matrix. So, using MW-assisted synthesis, rather negligible loading (3.1 wt %) of K-OH in the porous host is achieved. This phenomenon is explained by a very short reaction time (∼ 30 min) under conditions of fast MW-synthesis, which is insufficient for the ideal accommodation of K-OH molecules in the NH2-MIL-101(Al) pores. On the contrary, the solvothermal procedure realized at a prolonged reaction time (72 h) results in a larger K-OH loading (16.9 wt %) in the NH2-MIL-101(Al) host matrix.IR spectra of NH2-MIL-101(Al) and MIL/K-OH samples are shown in Fig. 2 . There are bands at 552 and 1398 cm−1 from the benzene ring in the spectrum of NH2-MIL-101(Al), as well as the bands at 897, 1581 and 1675 cm−1 from carboxylic groups, at 1261, 1338, 1625 and in the region of 3100–3500 cm−1 from -NH2 groups [46–49]. Two bands at 3301 and 3383 cm−1 can be assigned to antisymmetric (νasym(NH2)) and symmetric (νsym(NH2)) stretching vibrations of the -NH2 groups, respectively. This may indicate the existence of two types of amine groups in the framework [48,49]. The main types of characteristic bands in the FTIR spectrum of NH2-MIL-101(Al) are shown in Table 3.All these bands are also observed in the spectrum of MIL/K-OH composites. Unfortunately, the bands of K-OH overlap with those of the NH2-MIL-101(Al) material; however, some differences can be observed. First of all, the bands assigned to the C-H bending vibrations of the benzene ring and stretching vibrations of Al-O shift from 552 to 544 cm−1 and from 470 to 459 cm−1, respectively. The red shift is observed for the band at 3501 cm−1 assigned to the antisymmetric stretching vibrations of –NH2 groups (ν sym(NH2)). These changes can be caused by (a) the interaction between the –NH2 groups of the NH2-MIL-101(Al) framework and –OH groups of K-OH, and (b) interaction between –OH groups of K-OH and Lewis acid sites formed by Al3+ ions. Interaction between functional groups of K-OH and NH2-MIL-101(Al) is confirmed by the disappearance of the band at 3145 cm−1 attributed to the -OH stretching hydrogen bonds, i.e., cooperative ••OH •• OH •• OH chains of the framework [50] in the spectrum of K-OH (Fig. 2).Secondly, the characteristic band at 897 cm−1 corresponding to the vibrations of the substituted aromatic ring (C-H bending and CC-H stretching) is not observed in spectra of MIL/K-OH samples. This disappearance can arise from π-π stacking interaction between the aromatic rings of K-OH and the aromatic structure of NH2-MIL-101(Al).Another feature in the spectra of the samples is the dependence of the position of the band from the –NH2 groups (νsym(NH2)) on the method of the composite synthesis. In the spectrum of MIL/K-OH(Solv), the shift is larger (Δ = 26 cm−1) than that in the spectrum of the MIL/K-OH(MW) material (Δ = 16 cm−1). This phenomenon can be provoked by the difference in the strength of the interaction between K-OH molecules and the framework of NH2-MIL-101(Al).as a C-H acid probe molecule (DRIFT-CDCl3) also point to the interaction between –OH groups of K-OH and basic sites of NH2-MIL-101(Al). The spectrum of CDCl3 adsorbed on the NH2-MIL-101(Al) material is shown in Fig. 3 . The interaction of CDCl3 with basic sites of samples leads to the appearance of two bands at 2252 and 2212 cm−1 that can be assigned to basic sites formed by –NH2 and Al-OH groups, respectively. This suggestion follows from our early investigations of MIL-100(Al) and NH2-UiO-66(Zr) (Table S3, SI) [51]. The band at 2212 cm−1 is not observed in spectra of MIL/K-OH composites, which can indicate that basic sites formed by -NH2 groups are absent due to the interaction with –OH groups of K-OH.XRD patterns of NH2-MIL-101(Al) and MIL/K-OH composites are shown in Fig. 4. The main diffraction peaks of samples are similar to those of the previously reported patterns of the NH2-MIL-101 family, which has a cubic unit cell [52–54]. The values of the unit cell parameter of NH2-MIL-101(Al) is 89.71 Å, which is lower in comparison with MIL/K-OH(MW) and MIL/K-OH(Solv) (Table S4, SI). The refined unit cell parameter is equal to a = 87.860(2) Å according to synchrotron examinations (Figs. S1-S2 , SI).The coherent scattering region, DXRD was calculated using the Debye-Scherrer's formula (Eq. 1) [55,56]: (1) D XRD = 0.9 ⋅ λ β ⋅ cos θ where DXRD is the coherent scattering region (nm), β is the full-width at half maximum of the peak (radian), Θ is the Bragg angle of a diffraction peak (grad), and λ is the X-ray wavelength of CuKα (0.1542 nm). DXRD values for MIL/K-OH(MW) and MIL/K-OH(Solv) (23.6 ± 1.0 nm) are also larger than those for NH2-MIL-101(Al) (Table S4, SI). Moreover, there are noticeable differences in the ratio of the heights of the low-angle peaks. Thus, the increase in the ratios of integral intensities of I(333)/I(442) and I(662)/I(664) reflections is observed in XRD patterns of MIL/K-OH composites (Table S4, SI). There is also a noticeable difference in the heights of the (222) reflection in XRD patterns of MIL/K-OH(MW) and MIL/K-OH(Solv) materials (Fig. 4). The XRD pattern of the MIL/K-OH(MW) composite differs less from the XRD pattern of the NH2-MIL-101(Al) sample, unlike the XRD pattern of the MIL/K-OH(Solv) material. The lower intensity of the (222) reflection and the higher I(333)/I(442) and I(662)/I(664) ratios in the XRD pattern of MIL/K-OH(Solv) as compared to the XRD pattern of the pristine NH2-MIL-101(Al) material can indicate a partial pore blockage by calix[4]arene guest molecules. Note that K-OH reflexes (Fig. S1-S3, SI) are almost not observed in the synchrotron X-ray powder diffraction patterns of the MIL/K-OH(MW) and MIL/K-OH(Solv) composites. Probably, this phenomenon could be explained both by the location of the K-OH molecules inside the MIL porous host and a low K-OH content in the MIL/K-OH(MW) composite.TEM images of the MIL/K-OH are presented in Fig. 5. The MIL/K-OH(MW) sample is composed of nanocrystals with a highly-ordered mesoporous structure. The outside surface of MIL/K-OH(MW) is relatively dense with roughness and has nano-sized finger-like channels fitted for the K-OH encapsulation. The morphology of the MIL/K-OH(Solv) nanomaterial is similar to that of MIL/K-OH(MW) nanocrystals (Fig. 5). At the same time, the size of K-OH particles is larger in MIL/K-OH(Solv) in comparison with MIL-K-OH(MW). In the TEM image of MIL/K-OH(Solv) (Fig. 5), the particles with the interplanar spacing of the crystallites around 0.24 nm can be revealed, which corresponds to the lattice spacing of K-OH [57]. The differences in the guest K-OH location in the host structure affects the textural properties of the MIL/K-OH composites.were studied by N2 low-temperature adsorption. The corresponding results are shown in Table 4. The introduction of K-OH into the NH2-MIL-101(Al) matrix leads to the decreasing specific surface area and total pore volume, and this reduction is dependent on the preparation method of the composite. The effect of the MW-assisted procedure on the specific surface area and total pore volume expressed as a decrease of these parameter values as compared with the NH2-MIL-101(Al) reference sample is lower in comparison with the solvothermal method.[a] VΣ was estimated from the adsorption value at p/p o = 0.99; [b] V μ = V Σ − V meso ; [c] Cumulative mesopore volume calculated from the desorption branch of the isotherm by the BJH method and the standard thickness of the adsorption film.Mesoporosity of MIL/K-OH(MW) is higher (Vmeso/VΣ = 0.39) than that of MIL/K-OH(Solv) (Vmeso/VΣ = 0.26) due to the different filling of the framework with guest K-OH molecules. Actually, according to the elemental analysis data (Table 2), the MIL/K-OH(Solv) material contains much more K-OH molecules than its MIL/K-OH(MW) counterpart.In order to evaluate the application prospective of the synthesized MIL/K-OH composites, they have been tested for their catalytic activity in the chemical fixation of CO2 gas through the conversion of CO2 and propylene oxide (PO) to propylene carbonate (PO) (Scheme 2 ). Catalytic properties of MIL/K-OH composites were investigated at 80 °C under 0.8 MPa of CO2. The main results are shown in Table 5. According to the test results and GC-MS analysis, PC was the major product with about 99 % selectivity in all cases. A blank experiment indicated that the conversion of PO was < < 1 % after 5 h. In the presence of K-OH, the conversion of PO for 5 h was less than 3 % (Table 5, run 2). The addition of tetra-n-butylammonium bromide (TBABr) as a co-catalyst leads to the increasing conversion of PO to 35 % (Table 5, run 3). A similar effect of TBABr was observed for NH2-MIL-101(Al) and NH2-UiO-66(Zr) (Table 5, runs 7–8 and 12–13). Therefore, a binary system was used for the investigation of catalytic properties of MIL/K-OH composites.Conversions of PO in the presence of MIL/K-OH(Solv) and MIL/K-OH(MW) were 74 % and 82 %, respectively (Table 5, runs 4–5). These results show that the activities of the MIL/K-OH composites depend on their preparation method. The MW assisted procedure favors the higher activity of the composite than the traditional solvothermal method. The difference in catalytic activities of MIL/K-OH(Solv) and MIL/K-OH(MW) can be explained by several reasons, which mainly are based on the difference in textural properties of samples and particle size of K-OH. First of all, the higher activity of MIL/K-OH(MW) in comparison with MIL/K-OH(Solv) can be a result of the higher specific surface area and mesoporosity (Vmeso/VΣ) (Table 4) that can affect the accessibility of active sites for reagents. Another reason can be related with the difference in the particle size of K-OH that can affect both the number and reactivity of active sites [58].The reaction mechanism of cycloaddition of CO2 to epoxide to produce cyclic carbonates is ascribed to the action of the “Lewis acid site and basic site” pair (LAS-BS) [37–40]. Physicochemical investigations and catalytic data point that the structure of active sites in the NH2-MIL-101(Al) material and MIL/K-OH composites are different ( Scheme 3). Thus, LAS and BS are formed by Al3+ ion and -NH2 group of the linker in the NH2-MIL-101(Al) sample. The coordination of oxygen atom of the propylene oxide ring to LAS favors the electron density redistribution and, therefore, activation of propylene oxide, while interaction of CO2 with BS leads to its activation. In MOF/TBABr binary systems, the main role of the co-catalyst is to increase the rate of ring-opening of propylene oxide. Bromine ion attacks the least hindered carbon atom of propylene oxide with formation of reactive oxygen anion. Subsequently, CO2 reacts with the oxygen anion of the opened propylene oxide to form an intermediate, and finally, bromine ion is eliminated by a ring-closing step to produce propylene carbonate from the intermediate while regenerating the catalyst. In contrast to NH2-MIL-101(Al), -NH2 groups in the MIL/K-OH composite are blocked by K-OH due to the interaction of -OH groups of K-OH with the amine groups (Scheme 3). It can be suggested that the other three -OH groups can interact with TBABr to form an ion pair. In this case, the bromine ion is more mobile and can accelerate the ring-opening step due to the nucleophilic attack to C1 atom of epoxide to produce an intermediate. Probably, for this reason, the activity of the MIL/K-OH(MW) composite is higher than that of the pristine NH2-MIL-101(Al) material.[a] 24 mmol of propylene oxide; [b] Catalyst was used in second cycle; [c] [Zn(EIM)2], EIM – 2-ethylimidazole; [d] [Zn(cbIM)2], cbIM - 5- chlorobenzimidazole; [e] - [Zn(abIM)2], abIM - 4-Azabenzimidazole.Stability of MIL/K-OH(Solv) was investigated in a cyclic test at 24 mmol of PO, catalyst/TBABr of 0.85/0.85 mmol/mmol, 0.8 MPa of CO2, 80 °C for 5 h. After the first cycle, the catalyst was separated from the reaction mixture by filtration, washed with toluene, dried in air, calcined at 150 °C in air for 3 h and used in the next cycle. These tests revealed a slight activity decrease from 74 % (first cycle) to 67 % (second cycle) (Table 5 ). Comparison of the IR spectra of the MIL/K-OH(Solv) before and after catalysis clearly shows that no changes are observed in the IR spectra of the sample after the first and second catalytic cycles when compared to the pristine material ( Fig. 6). The decreasing activity is due to the blocking of active sites with reaction products as it follows from the appearance of new bands in the region of 850–1200 cm−1.The comparison of the efficiency of the most active sample, i.e. MIL/K-OH(MW), with the activity of binary systems based on metal-organic frameworks with NH2-groups in the structure of the linker, such as NH2-MIL-101(Al), NH2-MIL-53(Al) and NH2-UiO-66(Zr), is shown in Table 5 (runs 5, 7, 10 and 12). Results indicate that activities of all NH2-containing MOFs are lower in comparison with that of MIL/K-OH(MW). The activity of MIL/K-OH(MW) also was compared with activities of binary systems based on zeolitic imidazolate frameworks (ZIFs) reported in the literature [58–60]. Data shown in Table 5 and Table S2 (SI) indicate that MIL/K-OH(MW)/TBABr is definitely a promising catalytic system for the cycloaddition of CO2 to PO. Moreover, this binary system allows producing PC under relatively mild reaction conditions. The conversion of PO was 77 % at 50 °C for 24 h.For the first time, novel host-guest composite materials based on calix[4]arene with hydroxyl groups in the arene “bowl” (K-OH) and the aluminum(III) 2-aminoterephthalate metal-organic framework (NH2-MIL-101(Al), MIL) have been synthesized according to one-step in situ or “bottle-around-the-ship” approaches. The effect of the synthesis method, i.e., MW-assisted technique (MW) and the solvothermal procedure (Solv), on the structural, morphological, textural and catalytic properties of MIL/K-OH composites was demonstrated.The possibility of the interaction via the strong hydrogen bond between protons of –OH groups of K-OH and –NH2 groups of the NH2-MIL-101(Al) framework has been proven by spectroscopic investigations. According to IR spectroscopy, the interaction between the functional groups of K-OH and NH2-MIL-101(Al) was stronger in MIL/K-OH(Solv). According to SEM study, both MIL/K-OH(MW) and MIL/K-OH(Solv) composites have a highly-ordered mesoporous structure with a partial pore blockage by K-OH guest molecules. However, the particle size of K-OH is larger in MIL/K-OH(Solv) in comparison with MIL-K-OH(MW) as confirmed by TEM data. The effect of the MW-assisted procedure on the specific surface area and total pore volume is lower in comparison with the solvothermal method due to the different filling of the framework with guest K-OH molecules.The catalytic performance of the novel MIL/K-OH(MW) and MIL/K-OH(Solv) composite materials was investigated in solvent-free coupling of CO2 and propylene oxide (PO) to produce propylene carbonate (PC) at 0.8 MPa of CO2 and 80 °C and compared with that of the individual components. K-OH guest molecules modify the catalytic properties of the NH2-MIL-101(Al) host, e.g., the activity of the MIL/K-OH(MW) composite is higher than that of the pristine NH2-MIL-101(Al) material. In turn, the use of the MIL/K-OH(MW) catalyst results in the increased PO conversion in comparison with the MIL/K-OH(Solv) material, which is related to the difference in the textural properties of both composites. The binary system MIL/K-OH(MW)/[n-Bu4N]Br was demonstrated to be a promising catalytic system for the cycloaddition of CO2 to PO. In the presence of this binary system, the conversion of PO was 77 % with 99 % selectivity towards PC at 1.2 MPa of CO2, 50 °C for 24 h. Leonid M. Kustov: Conceptualization, Writing – review & editing. Vera I. Isaeva: Supervision, Writing – review & editing. Maria N. Timofeeva: Supervision, Writing – review & editing. Ivan A. Lukoyanov: Conceptualization, Formal analysis, Investigation, Methodology, Visualization, Writing – original draft. Valentina N. Panchenko: Conceptualization, Formal analysis, Investigation, Writing – original draft. Evgeny Y. Gerasimov: Formal analysis, Investigation, Writing – original draft. Vladimir V. Chernyshev – Formal analysis, Investigation. Lev M. Glukhov: Formal analysis, Investigation.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This work was supported by the Ministry of Science and Higher Education of the Russian Federation within the governmental order for Boreskov Institute of Catalysis (АААА-А21-121011390055-8) and project no. 075-15-2021-591 for N.D. Zelinsky Institute of Organic Chemistry. The authors acknowledge the Novosibirsk State University Shared Equipment Center “Applied Physics” and “VTAN”, and ESRF for providing access to the ID22 station (experiment MA-4527).Supplementary data associated with this article can be found in the online version at doi:10.1016/j.jcou.2022.102262. Supplementary material. .

In this study, novel composite materials based on calix[4]arene with hydroxyl groups in the arene “bowl” (K-OH) and the aluminum(III) 2-aminoterephthalate metal organic framework (NH2-MIL-101(Al), MIL) have been synthesized according to in situ or “bottle around ship” strategy. The effect of the synthesis method, i.e., MW-assisted technique (MW) and the solvothermal procedure (Solv), on the structural, morphological, textural and catalytic properties of MIL/K-OH composites was demonstrated. Catalytic performance of the MIL/K-OH(MW) and MIL/K-OH(Solv) composites was studied in solvent-free coupling of CO2 and propylene oxide (PO) to produce propylene carbonate (PC) at 0.8 MPa of CO2 and 80 °C. The activity of the MIL/K-OH(Solv) catalyst was higher in comparison with MIL/K-OH(MW) material as a result of the differences in the textural properties. The binary system MIL/K-OH(MW)/[n-Bu4N]Br was demonstrated to be a promising catalyst for cycloaddition of CO2 to PO. In its presence, the conversion of PO was 77 % with 99 % selectivity towards PC at 1.2 MPa of CO2, 50 °C for 24 h.

Enantioselective metal-catalysis is one of the most important synthetic approaches for preparing enantioenriched compounds, which occupy a central position in areas ranging from medicinal chemistry to chiral materials. The performance of chiral metal-catalysts depends, mainly, on the correct choice of the chiral ligand [1]. Among the large amounts and variety of ligands built up, only some have a broad applicability. A large reaction, substrate and/or reagent scope are worthwhile to minimize the time devoted to ligand finding and preparation. The most efficient, called “privileged chiral ligands”, derive from a few core structures [2]. Surprisingly, most of them possess C2 symmetry (e.g. BINOL, BINAP, TADDOL …). The reason for initially selecting bidentate ligands with C2-symmetry was to reduce the number of catalyst-substrate arrangements and transition states, helping mechanistic studies and therefore facilitating the understanding of the correlation between ligand architecture and catalytic results crucial to find the optimal catalyst. However, the intermediates that take place in the catalytic cycle may not be symmetric and in these cases the desymmetrization of the ligand has proven to allow a better enantiocontrol in certain reactions. An easy and suitable way to desymmetrize a ligand is to introduce in the ligand design two different donor atoms. In the last decades, heterodonor ligands have therefore increased their use in catalysis, by being able to facilitate the stereocontrol thanks to the different electronic and steric properties of two distinct coordination groups [3]. Moreover, the presence of the two different donor groups facilitates the discovery of highly effective ligands for a given reaction since it is easy to independently tune both donor groups. Among heterodonor ligands, those containing both P- and N-donor groups have a predominant position, with phosphorus-oxazoline ligands the most commonly investigated due to their ready accessibility and modular construction (see Section 2). During the last decade the oxazoline group has been substituted for other more robust N-donor groups (e.g. imines, amines, oxazoles …) that in some cases have allowed further catalytic improvement (see Section 3). Over the last years, there has been a new renaissance in the use of chiral P,O- and P,S- ligands in asymmetric catalysis, which have led to very interesting new successful applications (see Section 4). We here offer a critical review on the design and application of the most successful bidentate heterodonor P-oxazoline, P-other N, P-O and P-S ligand families in asymmetric processes. In addition to the comprehensive contents, for each group of ligands this review also offers a new perspective of presenting such ligands. While other reviews present the ligands by time or asymmetric reaction, for the first time we grouped them by their relationship between their structure and catalytic performance, which helps to associate the structural characteristics of a set of catalysts with their catalytic capacity and will facilitate further research on the field. Finally, the representative applications of these heterodonor family of ligands in total synthesis are collected at the end of this review (Section 6).Most of P-oxazoline ligands are prepared from easily obtainable chiral amino alcohols in short and efficient synthetic sequences [3a-e]. The beginning of P-oxazoline ligands can be found in 1993, with the preparation of the phosphine-oxazoline PHOX ligands 1 (Fig. 1 ) [4] that have been applied with huge success to a large variety of asymmetric transformations (e.g. allylic substitution and decarboxylative allylation reactions, hydrogenation, inter- and intramolecular Heck reactions, conjugate additions to enones, Diels-Alder and aza-Diels-Alder reactions, etc.) [2]. Despite being reported >25 years ago, PHOX ligands maintain their success in new enantioselective transformations, emphasizing their category as privileged ligand [5].Since then, many new P-oxazoline ligands have been synthesized by varying either the ligand skeleton or the steric/electronic properties of the phosphine group or by swapping the phosphine moiety by other P-donor groups (e.g. phosphoroamidite, phosphinite) [3a-e]. These modifications allowed to improve enantioselectivities in some specific reactions. However, a few of them have been used with success to different asymmetric reactions and have shown a wide substrate and/or reagent scope. Next, we collect the families of P-oxazoline ligands that have successfully showed a large range of reaction and/or substrate scope and the relationship between their structural design and their catalytic performance. We will center on recent works and a quick overview of previous reports will also be incorporated.Some successful modifications of the PHOX ligands are electronics by introducing withdrawing groups in the phenyl backbone ring or/and in the phosphine moiety. In this respect, it can be highlighted the recent synthesis of ligands L1 developed by Stoltz’s group (Scheme 1 ). The synthesis of these ligands relies in the Cu-catalyzed Ullmann-type coupling which allows the modular synthesis of ligands L1 even in demanding steric and electronic cases, avoiding the discrete synthesis of anionic reagents (Scheme 1) [6a].Stoltz’s group found that L1 (X = CF3 and R = 4-CF3-C6H4) was highly favorable in the Pd-catalyzed decarboxylative allylation of cyclic allyl carbonates, providing higher catalytic performance than PHOX ligands (Scheme 2 a) [6]. This methodology has facilitated the transformation of substituted cyclic allyl enol carbonates into a range of natural products (e. g. (+)-hamigeran, (−)-cephalotaxine and elatol; see Section 6) [7]. The same group has recently expanded this methodology to acyclic enol carbonates, achieving also excellent enantioselectivities (Scheme 2b) [8]. Interestingly, the same catalytic system was also able to achieve excellent enantiocontrol in the decarboxylative allylation of bench-stable β-keto allyl esters. This latter finding has facilitated the synthesis of many quaternary cyclobutanones and cyclopentanones as well as heterocyclic compounds (e.g. piperazines, diazepanones, etc., Scheme 2c) [9]. This methodology has again enabled the total synthesis of several natural products (e.g. (+)-sibirinine, (−)-goniomitine, (+)-limaspermidine, nigelladine A, etc; see Section 6) [10].The Stolz’s group also demonstrated the benefits of ligand L1 in the Pd-catalyzed enolate alkylation cascade procedure that allows the installation of vicinal quaternary and tertiary stereocentres at the α-carbon of a ketone [11]. This multiple bond-forming procedure takes place by conjugate addition of the chiral Pd/L1-enolate, generated in situ from β-keto allyl esters, to malononitriles (Scheme 3 a). More recently, You’s group disclosed the usefulness of Pd/L1 catalyst in the highly diastereo- and enantioselective synthesis of tetrahydrofurobenzofurans and tetrahydrobenzothienofurans through a Pd-catalyzed dearomative [3 + 2] cycloaddition of nitrobenzofurans (Scheme 3b) [12].Guiry’s group also disclosed the usefulness of ligand L1 (X = CF3 and R = 4-CF3-C6H4) in the preparation of sterically demanding tertiary α-aryl ketones, such as isoflavones and α-aryl-1-indanones, via Pd-catalyzed decarboxylative protonation of α-aryl-β-keto allyl esters [13]. Interestingly, they also showed that the nature of the proton source has an impact on enantioselectivity, which clearly indicates that it is involved in the enantioselective determining step. By switching the proton source form Meldrum’s acid to formic acid both enantiomers of the α-aryl-1-indanones can be accessed (Scheme 4 ).Another notable modification of the PHOX ligands was to add a biaryl phosphite functionality (a–e) instead of the phosphine moiety, which increases the π-acceptor character of the ligand [14]. Ligands L2a–e (R = tBu, iPr, Et, Ph; Scheme 5 ) have the advantage to be air-stable solids, which are easily prepared by attaching several amino alcohols and phosphorochloridites to the 2-hydroxyphenyl cyanide scaffold (Scheme 5) [15].Ligands L2 were initially designed to increase the substrate range of PHOX ligands in Pd-catalyzed allylic substitutions [14,16]. In Pd-catalyzed allylic substitution reactions, ligands able to induce high enantioselectivities in a large range of substrates' type and nucleophiles are exceptional [1c,17]. Improving Pd-PHOX catalysts, which gave excellent enantioselectivities with rac-(E)-1,3-diarylallyl substrates but low-to-moderate ee’s for cyclic and 1,3-dialkylallyl substrates, respectively [2a,3], Pd/L2 catalysts provided an excellent catalytic performance in the allylic substitution for all of them [14]. Thus, higher activities (TOFs > 2400 mol substrate × (mol Pd × h)−1) than with PHOX ligands were achieved due to the π-acceptor character of the phosphite moiety. In addition, the high enantioselectivities (up to 99% ee) were reached not only for disubstituted substrates, but also for tri- and monosubstituted ones (Scheme 6 ) [15]. The highest enantioselectivities for the benchmark substrate, rac-1,3-diphenylallyl acetate, were reached with the simple tropoisomeric ligand L2b independent of the oxazoline substituent (R = Ph, Et, iPr or tBu). Pd/L2b was also tolerant with the variation of the type of nucleophile (Scheme 6). In this respect, excellent enantioselectivities were reached with butenyl-, pentenyl-, propargyl and allyl-substituted malonates, fluorobis(phenylsulfonyl)methane (a fluoromethide synthon) and non-aromatic alcohols (ee's up to >99%, Scheme 6), whose resulting substitution products have proved valuable in the preparation of more complex chiral products [18]. Pd/L2b was also effectively applied to symmetrical 1,3-diaryl- and 1,3-dialkylallyl acetates with different electronic and steric demands with a large variety of C-nucleophiles (Scheme 6). For cyclic and monosubstituted substrates, the highest ee’s (up to >99% and 92%, respectively; Scheme 6) were achieved using ligands L2b and L2e. It should be noted the high regioselectivities in favor to the branched chiral product attained in the allylic alkylation of monosubstituted substrates, since most of the Pd-catalyst led to the preferential formation of the achiral linear isomer [16,19]. The phosphite moiety of the ligand is key to understand the high regioselectivities achieved towards the branched isomer. Thus, the phosphite group favors the nucleophilic attack at the most substituted allylic terminal carbon atom thanks to the trans-influence [14]. Excellent results were also reached in alkylation of 1,3,3′-trisubstituted allylic acetates (Scheme 6).The broad substrate scope of the Pd/L2b catalyst system was rationalized by NMR studies and DFT calculations of their Pd-η2-olefin and Pd-η3-allyl intermediates complexes [14b]. These studies indicated that: (a) the tropoisomeric biphenyl phosphite group in ligand L2b adopts an (S)-configuration in the Pd-η3-allyl intermediates with hindered as well as unhindered substrates (Fig. 2 ); and (b) the Pd/L2 catalysts are able to readjust the binding chiral pocket's size to the substrate requirements, which explains the high ee’s achieved in a very diverse set of substrates. This latter feature is crucial to explain the success of Pd/L2 catalytic system in other asymmetric transformations, such as hydrogenation of unfunctionalized olefins [20], intermolecular Heck reactions, with results comparable to Pd/PHOX catalyst, using ligand L2b (R = Ph) [21], and in the hydroboration of olefins [22]. Interestingly, for the latter transformation Ir/L2b (R = iPr) catalyst proved to be of an exceptional effectiveness, attaining higher ee’s (up to 94%) than phosphine-oxazoline PHOX ligands [23]. These results are specially remarkable because achieving high selectivities in the hydroboration of 1,1′-disubstiuted alkenes is difficult, due to face selectivity issues and the difficulties in controlling the regiospecific boration in the terminal β-position [24]. Particularly, Pd/L2b is the only catalytic system able to hydroborate α-tert-butylstyrenes, thus complementing Cu-NHC catalysts, the only other system able to hydroborate α-alkyl styrenes with high ee’s [25].A final benefit of the new phosphite-oxazoline L2 ligands compared to the PHOX ligands is that the most efficient ligand in all of the catalytic asymmetric reactions discussed above are derived from affordable (S)-phenylglycinol or (S)-valinol (R = Ph or iPr) instead of the more costly (S)-tert-leucinol found in PHOX ligands.Other modifications of the PHOX ligands are on the oxazoline group, by attaching the phenyl backbone ring to the stereogenic center next to the oxazoline (ligands L3; Fig. 3 ) [26], or introducing other oxazoline substituents such as, ferrocene, tricyclic and sugar oxazoline groups (e.g. ligands L4 and L5; Fig. 3) [27]. However, in any case the enantioselectivities and the substrate scope improved those attained with the PHOX ligands. Thus, ligands L3 (R1 = Cy or Ph and R2 = tBu) led to high enantioselectivities in the Ir-catalyzed hydrogenation of unfunctionalized olefins but only in the reduction of some methylstilbenes (ee's up to 99% ee) and a range of β-methylcinnamic esters with enantioselectivities up to 99% [26]. Ligand L3 (R1 = Ph and R2 = tBu) has also been used with success in the Pd-catalyzed allylic alkylation of benchmark substrate (ee's of 98%) and in the intermolecular Heck reaction of 2,3-dihydrofuran with the phenyl triflate (94% ee) [28]. Ligands L4 and L5 followed a similar trend than the PHOX ligands in the Pd-catalyzed allylic alkylation. Thus, they provided high enantioselectivities with rac-(E)-1,3-diarylallyl substrates, but low for cyclic ones. Ligand L4 provided also high enantiocontrol in the Pd-catalyzed Heck reaction of 2,3-dihydrofuran using various aryl triflates (ee’s up to 98%) [29].Another modification into the oxazoline ring was to introduce substituents in the 5 and/or 5′ positions (e.g. ligands L6 and L7; Fig. 4 ) [30] providing similar levels of enantioinduction than the usually most effective PHOX ligand, the tBu-PHOX, but with the advantage of being readily accessible as both enantiomers from either the (S) or (R)-valine rather than from expensive tert-leucinol enantiomers.Besides these modifications on the phosphine and oxazoline moieties and in phenyl backbone ring, many changes on the ligand backbone have been studied. One of these modifications includes a methylene spacer linking the phenyl ring of the ligand backbone and the oxazoline ring (ligands L8), forming with the metal a higher seven-membered chelate ring (phosphine-oxazoline ligands L8, R1 = Me, H; R2 = Me, iPr, tBu; Scheme 7 ) [31]. The phosphine moiety in ligands L8 has also been exchanged for a biaryl phosphite group (ligands L9; Scheme 7) [32]. Ligands L8 and L9 were synthesized in few steps from easily available starting material, as illustrated in Scheme 7.Zhou and coworkers used Pd/L8 catalytic systems in the intermolecular Heck reaction (Scheme 8 ) [31]. The intermolecular asymmetric Heck reaction is less developed than the intramolecular version due to regioselectivity issues, which hampers its application for the synthesis of more complex molecules [33]. Pfaltz early demonstrated that the use of tBu-PHOX ligand can overcome the regioselectivity issue, although it requires from 3 to 7 days for full conversion [34]. Ligands L8 (R1 = Me, H and R2 = tBu) provided high regio- and enantioselectivities (up to 95% ee, Scheme 8), with results comparable to PHOX ligands, in the reaction of 2,3-dihydrofuran and various aryl triflates [31]. From these results it should be highlighted that ligands that contained hydrogens in the benzylic position provided the R-enantiomer while ligands with methyl substituents at R1 provided the S-enantiomers (Scheme 8).More recently our group decided to replace the phosphine group in ligands L8 by several π-acceptor biaryl phosphite moieties (ligands L9b, e, f–h; R1 = Me, H; R2 = iPr, tBu, Ph; Scheme 7) [32]. This change increases the activity, because the presence of the phosphite group favors the migratory insertion, which come up to be the rate-determining step. At the same time the substrate scope could be extended to other heterocyclic and carbocyclic olefins and to other triflates including non-aromatic ones (ee's up to 98% and regioselectivities up to 99%). The best results were obtained with the ligand that had biaryl phosphite groups b, e and h, a hydrogen in R1 positions and an iPr oxazoline substituent, avoiding the use of the costly tBu substituent required in the analogous phosphine-oxazoline L8 and PHOX ligands [32].Advantageously, the same family of P-oxazoline ligands L8–L9 also provided an excellent catalytic performance in the reduction of unfunctionalized olefins or olefins with poorly coordinative groups. This was a relevant finding because the reduction of these type of substrates is underdeveloped compared with the asymmetric hydrogenation of alkenes containing coordinative groups [35]. This is because catalysts able to hydrogenate unfunctionalized olefins are very sensitive to changes in the olefin geometry and to changes in the substitution pattern. Thus, for instance, most of the catalysts perform well for trisubstituted E-unfunctionalized alkenes. Only very recently have appeared catalysts able to reduce Z-trisubstituted and 1,1′-disubstituted. The asymmetric reduction of tetrasubstituted unfunctionalized olefins still remains a challenge. This substrate-dependent behavior was already displayed with the pioneering Pfaltz’s design of [Ir(PHOX)(cod)]BArF (cod = 1,5-cyclooctadiene and BArF = 3,5-(F3C)2-C6H3)4B) catalyst precursors [36], which mainly provides high ee’s in the hydrogenation of a small group of E-trisubstituted olefins [37]. The authors first prepared the catalyst precursors [Ir(cod)(L8–L9)]BArF by reaction of the corresponding ligand with [Ir(μ-Cl)(cod)]2 and subsequent Cl/BArF anion exchange to give air stable red–orange solids in high yields (Scheme 9 ). The VT-NMR spectra (from +35 to −85 °C) showed one single isomer in solution [20,38].They found that the use of [Ir(cod)(L8)]BArF (L8; R1 = H and R2 = iPr) allowed to extend the range of E-trisubstituted olefins to include allylic alcohols and α,β-unsaturated ketones and esters (ee’s up to 98%) [38]. Replacing the phosphine moiety in ligands L8 by several π-acceptor biaryl phosphite groups (ligands L9) further extended the array of substrates successfully hydrogenated, including more challenging 1,1′-disubstituted olefins [20]. The highest enantioselectivities were obtained with [Ir(cod)(L9)]BArF containing the ligand with the less expensive Ph or iPr oxazoline substituents and hydrogen atoms in the benzylic position. The phosphite group depended on the substrate to be hydrogenated (a summary of the reduction of 55 olefins with Ir/L9 are shown in Fig. 5 ). Interestingly, environmentally friendly solvent 1,2-propylene carbonate (PC) could be used instead of the commonly used dichloromethane without any deleterious effect on enantioselectivity. High enantioselectivities were therefore attained in the reduction of trisubstituted olefins including the more challenging triarylsubstituted substrates ones and those containing several poorly coordinative groups such as α,β-unsaturated ketones, amide, lactones, lactams, alkenyl boronic esters and enol phosphinates (ee’s up to >99%). Even thought, highly enantioselective hydrogenation catalysts for 1,1′-disubstituted substrates are very scarce [39], it was gratifying to obtain ee’s up to 98% in a large number of tert-butyl-aryl 1,1′-disubstituted alkenes (Fig. 5) which differs in the steric and electronic characteristics of the aryl substituent. Decreasing the bulkiness of the alkyl substituent on these α-alkyl-styrenes results in slightly lower enantioselectivities (ee’s from 83% to 91%), due to a competing isomerization pathway as it was disclosed by means of deuterium labeling experiments. Similar values of enantioselectivities were found in the hydrogenation of 1,1′-disubstituted alkenyl boronic esters and enol phosphinates (Fig. 5).In addition, [Ir(cod)(L9)]BArF were also successfully applied in the reduction of an additional challenging class of substrate; the cyclic β-enamides (Scheme 10 ) [40]. Despite, there is an important number of therapeutic agents (e.g. robalzotan, rotigotine, terutroban and alnespirone) [41] than can be accessed via their hydrogenation, there are only few catalysts able to hydrogenate such substrate class with high ee’s, being the majority based on rhodium and ruthenium [42]. In 2016, Verdaguer’s and Riera’s group demonstrated that Ir-PN catalyst can also be used, exceeding the scope of Ru/Rh-catalyst [43]. Then, our group decided also to study the application of Ir/L9 [40]. Enantioselectivities were high for many cyclic β-enamides derived from, 2-tetralones and 3-chromanones when using Ir/L9f (R1 = H; R2 = iPr) catalyst (Scheme 10). To note, the high enantioselectivity achieved in the reduction of N-(5-methoxy-3,4-dihydronaphthalen-2-yl)acetamide, which provides a crucial intermediate for the synthesis of rotigotine [40]. We also found that both enantiomers of the products can be accessed by exchange iridium to rhodium. Again, the use of PC has not effect on the enantioselectivities.Many of the backbone changes in the PHOX ligands also includes the replacement of phenyl backbone ring of PHOX ligand by other moieties (Fig. 6 ), such as ferro- and ruthenocene groups (e.g., ligands L10–L14) [44], biphenyl or binaphthyl groups (e.g., ligands L15) [45], several heterocyclic backbones (e. g., ligands L16–L19) [46], an alkyl chain (e.g., ligands L20–L28) [47] and bicyclic, sugar, and spiro backbones (e.g., ligands L29–L34). In many of these latter backbones modifications the phosphine group has also been replaced by a phosphinite, a phosphite, an aminophosphine and an stereogenic P groups. Among them those having two carbons linking the two donor functionalities have been employed with great success to several enantioselective reactions.In this respect, we can point up the families of phosphinite/phosphite-oxazoline ligands L22–23 and L24–27 (Fig. 6), where the ortho-phenylene tether of the PHOX has been changed by an alkyl chain. They were initially designed to provide a wider substrate capacity in the asymmetric reduction of unfunctionalized olefins. Starting from different carboxylic acid derivatives, chiral serine or threonine methyl esters, and Grignard reagents a range of hydroxyl-oxazolines were easily attained, which after treatment with the corresponding chlorophosphine or phosphorochloridite gave access to ligands L22–L23 (Scheme 11 ) [48].The phosphinite–oxazolines L22, developed by Pfaltz, (Fig. 6 and Scheme 11) are one of the most successful ligands for the Ir-catalyzed reduction of unfunctionalized olefins [24]. Unlike the PHOX ligands, the phosphorus unit is bonded to the stereogenic center next to the oxazoline.The Ir-catalyst precursors were synthesized using the same process described for previous [Ir(cod)(L8–L9)]BArF complexes, obtaining air-stable orange powders that needed to be purified by column chromatography on silica gel. With [Ir(cod)(L22)]BArF, they reached excellent enantioselectivities for the first time in the reduction of E- and Z-2-aryl-2-butenes (Fig. 7 ) [48a,49]. The author optimized the enantioselectivity for each substrate by systematic modifications of the substituents at the oxazoline ring and ligand skeleton. The best enantioselectivities were achieved with [Ir(cod)(L22)]BArF, containing diphenylphosphinite ligands L22 (R1 = Ph) with a methyl group at R3 and a benzyl at the alkyl chain (R4), although, the correct choice of the oxazoline R2 substituent and the configuration of the carbon of R3 is determined by olefin geometry. Thus, for E-trisubstituted olefins,  ee’s are highest with a 3,5-Me2-Ph or a Ph oxazoline R2 substituent and an S-configuration for R3, while for Z-olefins they are best with a Ph oxazoline substituent and an R-configuration into the ligand. Further optimization of ligand parameters allowed for the first time to reduce some more challenging terminal olefins and 1,1′-disubstituted enamines (ee's up to 99%; Fig. 7) with the ligand that contains a methyl and a benzyl group at R3 and at R4, respectively but a cyclohexenyl at R1 [49b,c]. More recently, Ir-L22 also allowed the hydrogenation of α,β-unsaturated nitriles (ee's up to 98%, Fig. 7) [49d]. These catalysts also perform well in 1,2-propylene carbonate, a green solvent, allowing the catalysts to be recycled several times [50].Then our group synthetized the phosphite-based analogues of ligands L22, which broadened the number of 1,1′-disubstituted substrates to be reduced with success [51]. By using [Ir(cod)(L23j)]BArF and [Ir(cod)(L23b)]BArF (R2 = Ph, R3 = H and R4 = Me) catalysts high ee’s (up to >99%) were therefore attained in the hydrogenation of a range of (het)aryl–alkyl disubstituted olefins, allylic alcohols and allylic silanes (29 compounds; Fig. 8 ), surpassing previous successfully Ir/L9 and Ir/L22 catalysts [39]. It should point up that Ir/L23j (R2 = Ph, R3 = H and R4 = Me) also attained high ee’s for trisubstituted olefins and α,β-unsaturated esters. For Z-trisubstituted olefins and allylic alcohols the highest catalytic performance was attained with Ir/23b (R2 = Ph, R3 = H and R4 = Me) [51]. Advantageously, the use of PC enabled the recycle of the catalysts until five times.Useful, ligands L23 were also used in the allylic substitution of many mono, di- and trisubstituted linear hindered and unhindered linear allylic acetates (up to 99% ee) with C- and N-nucleophiles [48b]. Moreover, reversing the configuration of the alkyl chain or reversing the configuration of the phosphite group led to both enantiomers of the products. The results surpass PHOX ligands and are similar to the best accounted with the previous family of phosphite-oxazoline ligands L2 except for cyclic substrates (ee's up to 83%). To increase the enantioselectivity in the cyclic substrates the oxazoline group was changed by a thiazoline group (see Section 3) [18c]. With this simple modification the enantioselectivities improved significantly (94% ee). Both families of ligands (phosphite-oxazoline/thiazoline) are complementary. The study of the Pd-π allyl intermediates allowed to explain the catalytic performance. Thus, 1,3-diphenyl allyl and 1,3-cyclohexenyl allyl Pd-complexes showed that the substitutents at the backbone ligand chain and at the oxazoline group have to be correctly combined to give the isomer that reacts faster and to avoid complexes with the ligand coordinated monodentated. However, for the unhindered lineal substrates, the enantioselectivity is explained throught a late transition state were the substituent at the alkyl chain favored to reach a specific Pd-olefin complex [48b].Ligands L23 also provided high catalytic performance in the Pd-catalyzed Heck reaction. High regio- and enantioselectivity could be achieved using many substrates and triflate sources, with ligands L23b (R2 = p-CH3-Ph or tBu, R3 = H, R4 = CH3, Fig. 6) [21]. Interestingly, the reaction times were considerably reduced by using microwave-irradiation conditions (from the 24 h with PHOX ligands to 10 min with ligand L23b) and regio- and enantioselectivities were still high (ee's up to 99%; Scheme 12 ).Another modification of ligands L22 was the development of ligands L24 (Fig. 6, R1 = o-Tol, Ph and R2 = tBu, iPr,), with the alkyl chain linked to the C-2 of the oxazoline group as in PHOX ligands [52]. Albeit the substrate range in asymmetric reduction of unfunctionalized olefins is reduced than with Ir/L22, they are complementary. Thus, high enantioselectivities were attained in the hydrogenation of allylic alcohols, alkenes with heteroaromatic substituents and the cyclic substrate 6-methoxy-1-methyl-3,4-dihydronaphthalene. Advantageous, Harmata and Hong have also used Ir/L24 catalyst in the total synthesis of pseudopteroxazole, a natural antitubercular agent (see Section 6). The catalysts was able to hydrogenate the internal double bond, and not the exocyclic CC bond, with high regioselectivity in 90% yield [53].More recently, new families of P-oxazolines (ligands L25–L27; Fig. 6) analogous to Pfaltz ones, still with two carbon atoms between the P- and N-donor functionalities, have been developed. The phosphite/phosphinite-oxazoline ligands L25–L27 were used in the enantioselective Ir-catalyzed hydrogenation and Pd-catalyzed allylic substitutions. Ligands L26 and L27 were prepared in a similar manner than ligand L24 and L25 (Scheme 13 ). Condensation of readily available chiral α-acetoxy acids with a range of chiral aminoalcohols followed by oxazoline formation using diethylaminosulfurtrifluoride (DAST) and subsequent alcohol deprotection yielded a range of hydroxyl-oxazolines. The later were then treated with the corresponding chlorophosphine or phosphorochloridite to give access to ligands L26–L27 (Scheme 13) [40].About the hydrogenation, air stable orange-solids [Ir(cod)(L)]BArF (L = L26 and L27b, f–h, k complexes were the first catalysts able to successfully hydrogenate di-, tri- and tetrasubstituted unfunctionalized olefins (ee's up to 99% in 62 examples, Fig. 9 ) [40]. As early stated, the asymmetric hydrogenation of tetrasubstituted olefins is still a challenge [35f,54]. Thus, there are a very limited number of catalyst able to hydrogenate them and those that do shows poor versatility, except for the recent publication with Ir/L21 catalyst [55] (Fig. 6) [56]. Improving previous reports, high enantioselectivities (up to 98%) were attained in the hydrogenation of several indenes, 1,2-dihydro-napthalene and a broad scope of acyclic tetrasubstituted olefins under mild reaction conditions using Ir-phosphinite-oxazoline L26 catalysts (Fig. 9) [40]. Significantly, it was also found that the phosphinite-oxazoline ligand L26 to be used depend on the olefin to be reduced. In this respect, while the highest enantioselectivity in the reduction of the more bulky cyclic indene substrates is obtained with the less bulky phosphinite group (Ph), for the less bulky indenes and acyclic substrates phosphinite ligands with bulkier substituents are needed (o-tolyl and cylcohexyl groups, respectively) to reach the highest enantioselectivity. Finally, by simple replacing the phosphinite by the right phosphite moiety (ligands L27) the same family of catalysts could also effectively hydrogenated a range of unfunctionalized tri- and disubstituted substrates (Fig. 9). The catalysts could also effectively reduce a variety of olefins with different functional groups from those poorly coordinative (e.g.enones, lactames and vinyl boronates) to highly coordinative ones (e.g. β-enamides) [40].Compared to Pd/L2 catalyst Pd/L27g (R2 = R3 = Ph) also gave higher activities (TOF up to 8000 h−1) and high enantioselectivities in the allylic substitution of a wide number of substrates (ee’s up to >99%, 74 examples in total, Fig. 10 ) [57]. Symmetrically disubstituted linear allylic acetates, containing alkyl or aryl substituents, with a variety of C-nucleophiles, including α-substituted malonates, malononitrile, diketones, 2-cyanoacetates and pyrroles were successfully alkylated with Pd/27 g. High enantioselectivities were reached with: i) both alkyl and aryl amines, ii) benzylic, allylic and iii) silanols. By introducing in ligand L27g a second methyl group at the alkyl chain,  ee’s could be improved up to >99% in the alkylation of cyclic substrates (Fig. 10). In addition, the Pd/L27g catalyst is one of the few catalyst that can deracemize unsymmetrically disubstituted substrates such as 1,1,1-trifluoro-4-phenylbut-3-en-2-yl acetate via dynamic kinetic asymmetric transformation with several malonates (yield’s up to 72% and ee’s up to 80%). Regioselectivities up to 90% and ee’s up to 98% were offered in the alkylation of 1-arylallyl acetates with malonates. Nevertheless, the regioselectivity into the branched product decreased with α-substituted malonates (e.g. it dropped from 83% using dimethyl malonate to 60% using dimethyl 2-methylmalonate). NMR and DFT studies showed an early TS, in which the enantioselectivity is guided by the ratio of the Pd-η3-allyl compounds and the relative electrophilicity of the allylic terminal carbon atoms. It was also found that the population of these intermediates is affected by the ligand parameters. Thus, whereas for cyclic substrates the configuration of the phosphite functionality together with the substituents in the alkyl chain guide the population of exo and endo isomers, for linear substrates its ratio is also affected by the oxazoline group [57].Most heterodonor P-oxazoline ligands developed have the chirality in the stereogenic carbon centers located on the oxazoline ring and/or in the carbon backbone (Fig. 6). We have also showed ligands that combines a chiral oxazoline or/and chiral carbon backbone with a phosphite with axial chirality (Fig. 6). However, few P-oxazoline ligands with a P-sterogenic center have been published, mainly due by the complexity of preparing bulky P-stereogenic phosphines in optically pure form (e.g. ligands L18 [46b], L19 [46c], L20 [58] and L28 [59]; Fig. 6). Verdaguer and Riera’s have recently reported a simpler protocol for the synthesis of P-stereogenic aminophosphine-oxazoline ligands L28 (MaxPHOX ; R2 = Ph, iPr, tBu; R3 = iPr; Scheme 14 ). The synthesis of L28 relies in the fact that upon activation enantioenriched tert-butylphenyl phosphinous acid borane undergoes stereospecific nucleophilic substitution with a range of amino-oxazoline compounds (Scheme 14) [59b]. Reaction of L28 with [Ir(μ-Cl)(cod)]2 and NaBArF using the above mentioned standard protocol led to [Ir(cod)(L28)]BArF catalyst precursors [59b]. Actually, ligands L28 only differ from ligands L26 and L27, in the replacement of phosphinite/phosphite groups by an aminophosphine group, so ligands L28 still have two carbon atoms linking the two donor groups.Usefully, these catalysts were able to efficiently hydrogenate a variety of tetrasubstituted olefins: indenes and 1,2-dihydro-napthalene derivatives (ee’s up to 96%) and also acyclic tetrasubstituted olefins (ee’s up to 99%; Fig. 11 ) [59a]. These excellent results were also attained in the reduction of tetrasubstituted vinyl fluorides (dr’s >99% and ee’s up to 98%) [59a]. Catalysts Ir/L28 (R1 = (S)-iPr and R2 = (R)-tBu or (R)-iPr), which have the oxazoline substituent and the bulky group at the P-center cis to each other, also reached excellent results (>99% ee) for cyclic β-enamides (10 examples; Fig. 11) using only 3 bar of hydrogen pressure [43]. The process could also be performed with greener solvents such as ethyl acetate and methanol. Ir/L28 (R1 = (R)-iPr and R2 = (S)-iPr) catalyst also provided comparable results than the best Ir–P,N systems reported in the hydrogenation of challenging N-aryl imines (Fig. 11) [60]. The reaction was performed with a balloon of H2 at −20 °C, achieving up to 96% ee. The authors found that the configuration at the P*-center had almost no effect on the enantioselectivity of the catalyst. Useful, the authors were able to isolate the active specie, complex 2, which was efficiently used to the direct hydrogenation of N-methyl ketimines (Fig. 11) [61]. Both, N-methyl imines and N-alkyl imines were hydrogenated with ee’s up to 94% using only 1 mol % of catalyst and 3 bar of H2, at 0 °C. The effective hydrogenation of this class of substrates had not yet been achieved, maybe due to the higher basicity of N-methyl amines than the N-aryl amines, which may lead to catalyst deactivation [62]. Ir/L28 (R1 = (R)-iPr and R2 = (R)-Ph) catalyst was also used with effectiveness in the enantioselective isomerization of N-allyl amides to enamides (Fig. 11), which allowed to shorter the reported synthetic route for obtaining the antibiotic R-sarkomycin methyl ester (See Section 6) [63].Andersson group was one of the few pioneering researchers in designing suitable ligands for the challenging Ir-catalyzed enantioselective reduction of unfunctionalized olefins. They synthesized an aminophosphine-oxazoline family of ligands L29 (R1 = Cy, o-Tol, Ph; R2 = tBu, H, Ph; R3 = Ph, H, Fig. 6), with a rigid bicyclic backbone, to overcome the limited substrate scope in this process [64]. Ligands L29 have also two carbons atoms linking the two donor functionalities. Ligands L29 were prepared in a multigram scale from (1S,3R,4R)-2-((benzyloxy)carbonyl)-2-azabicyclo[2.2.1]heptane-3-carboxylic acid or (1S,3R,4R)-2-(((4-nitrobenzyl)oxy)carbonyl)-2-azabicyclo[2.2.1]heptane-3-carboxylic acid, which are accessible via stereoselective aza-Diels Alder reaction [65] followed by Cbz- or p-NO2-Cbz-protection of the free amine. The oxazoline moiety was introduced via amide coupling with the desired 1,2-aminoalcohol followed by addition of mesyl chloride and base. Amine deprotection by hydrogenolysis using Pd/C followed by reaction with the appropriate chlorophosphine completes the synthesis of ligands L29 (Scheme 15 ). The Ir-catalyst precursors [Ir(cod)(L29)]BArF were then prepared as previously described (see Scheme 9). For first time, Ir/L29 catalysts furnished high enantioselectivities in the reduction of enol phosphinates [64b,c], vinyl silanes [64d], vinyl boronates [64f], fluorinated olefins [64e], α,β-unsaturated lactones [64g], α,β-unsaturated acryclic esters [64g] and γ,γ-di- and β,γ-disubstituted allylic alcohols [64h] (Fig. 12 ). The related aminophosphite-oxazoline/thiazole ligands extended the number of substrates that could be reduced with selectivities similar to the best published (see Section 3) [66]. They successfully reduced E- and Z-tri- and disubstituted olefins (ee’s up to 99%), including those with poorly coordinative groups (e.g. alkenylboronic esters, enones, vinylsilanes …) [66].Another notable family of P-oxazoline ligands with two carbon atoms linking the two donor functionalities, are the pyranoside ligands L30 (R = Me, iPr, tBu, Ph, Bn; Fig. 6). As with ligands L22–L23 the P is bonded to the stereogenic center next to the oxazoline nitrogen atom but differs in the presence of a more rigid sugar backbone. Ligands L30 were efficiently synthesized from d-glucosamine, an inexpensive natural feedstock (Scheme 16 ). d-glucosamine was first treated with the desired acid chloride or anhydride to form the corresponding amide [67]. Then, the hydroxyls groups at C-4 and C-5 position were protected with a benzaldehyde, to provide more rigidity to the backbone, and the rest of alcohols were acetylated. Formation of the oxazoline group was then achieved in the presence of anhydrous SnCl4. Deacetylation followed by treatment with the corresponding phosphorochloridite led to ligands L30. Ligand L30 provided high catalytic performance in the hydrogenation of unfunctionalized olefins, allylic substitutions and intermolecular Heck reactions [68]. In fact, [Ir(cod)(L30)]BArF were the first fruitful use of catalyst precursors with biaryl-based phosphite ligand in the hydrogenation of unfunctionalized alkenes [68a]. These Ir catalyst precursors were prepared straightforward as a single isomer following the previously described methodology (Scheme 9) as orange air-stable solids [68a,b]. The use of [Ir(cod)(L30b)]BArF and [Ir(cod)(L30l)]BArF (R = Ph) complexes containing bulky groups in the biaryl phosphite functionality led to high enantioselectivities (ee’s between 91% and >99%) in trisubstituted olefins (until 25 examples, Fig. 13 ), including triarylsubstituted substrates, α,β-unsaturated ketones and esters and vinylboronates among other type of olefins. High enantioselectivities were also reached in a range of terminal olefins (19 examples, Fig. 13) including heteroaromatic ones (ee’s up to 99%). Note that lower ee’s were achieved when using the phosphinite-oxazoline analogues.A computational study showed that the reaction proceed through an IrIII/IrV catalytic cycle where the migratory insertion of the hydride is the step that control the selectivity [68b]. From the calculated TSs structures, a quadrant model describing the ligand-substrate interactions was developed. The occupance in this quadrant model suited perfectly for olefins containing E-geometry (Fig. 14 ). Calculations also shown that by varying the biaryl-phosphite substituents it is possible to modulate the occupance of the semihindered quadrant allowing the coordination of Z-olefins as well.Ligands L30l (R = Me) and L30b (R = Ph) were also used in the intermolecular Heck reaction [68e,f]. They provided high activities (full conversions in minutes with microwave irradiation) and enantio- and regioselectivities (up to 99%) for a range of carbo- and heterocyclic olefins and triflate sources. In contrast to PHOX ligands, having non-bulky substituents at the oxazoline has a positive effect on both, selectivities and activities. Finally, good activities and high enantioselectivities (up to 99%) have also been reached in the substitution of tri- di- and monosubstituted linear substrates and cyclic substrates [68c,d]. For hindered linear substrate ligand L30b (R = Ph) provided the best results, while for unhindered linear substrates ligand L30g (R = Me) provided the best results, and ligand L30f (R = iPr) was the best for cyclic substrates. The related phosphinite-oxazoline analogues reached lower enantioselectivities [67]. The elucidation of NMR of the Pd-η3 allyl intermediates allowed the rationalization of the experimental catalytic results. They showed that the substituents at both the oxazoline and the phosphite moieties are crucial for high ee’s by enhancing the amount of the faster Pd-η3 allyl isomer and, at the same time, eluding the presence of species with the ligand coordinated as monodentated. They also corroborated that the nucleophile attacks the allylic terminal carbon which is trans to the phosphite functionality [68].Since the pioneering work of Chan and coworkers, the spiro backbone has been identified as a privileged arrangement for ligand families and catalysts [69,70]. We can highlight four main types of spiro phosphine-oxazoline ligands (Fig. 6): the SpinPHOX [71] (L31) developed by Ding, SIPHOX [72] (L32) reported by Zhou, HMSI-PHOX [73] (L33) developed by Lin and SMIPHOX [74] (L34) by Teng. The SMIPHOX was developed having in mind some distinct features compared with the other three, such as an spiro indane-based P,N ligand with non-C2-symmetric skeleton and higher rigidity and only one chiral center avoiding the complex sterochemistry. Among them, we can highlight the work of Zhou and coworkers with the spiro phosphine-oxazoline (SIPHOX) Ir-catalysts. SIPHOX ligands were prepared from enantiopure 1,1′-spirobiindane-7,7′-diol (SPINOL), which are prepared from 3-methoxybenzaldehyde followed by resolution with N-benzylcinchonidinium chloride [75]. SPINOL was then transformed to the corresponding phosphine-triflate compounds by ditriflation of the diol, monophosphinylation with the desired diarylphosphine oxide in the presence of Pd(OAc)2 followed by reduction with trichlorosilane. Phosphine-triflates were then transformed to the phosphine-acids by Pd-catalyzed carbonylation followed by hydrolysis of the formed esters. Amide formation with the desired 1,2-aminoalcohol in the presence of DCC (N,N′-dicyclohexylcarbodiimide) and HOBT (1-hydroxylbenzotriazole) followed by treatment with mesyl chloride and base led to spirocyclic phosphine-oxazolines L32 (Scheme 17 ). They could efficiently reduce imines and represented the first Ir-catalyst precursors [Ir(cod)(SIPHOX)]BArF (Fig. 15 ) able to reduce under basic reaction conditions a broad range of unsaturated carboxylic acids [72b,c], getting over the constraints of Rh- and Ru-catalysts, which are mainly restricted to acrylic and cinnamic acids (Fig. 15) [76]. The base is required to form the carboxylate anion that coordinates to iridium. Thus, Ir/L32, containing a bulky phosphine-aryl group (R1 = 3,5-tBu2Ph), proved to be highly active (TONs up to 10.000) and enantioselective (ee’s >99%) with a group of α-aryloxy and α-alkyloxy substituted α,β-unsaturated acids (Fig. 15). Nevertheless, Ir/SIPHOX catalysts do not perform well for α,β-unsaturated esters [77]. The same authors demonstrated the potential of Ir/L32 catalyst with the synthesis of a crucial intermediate in the preparation of rupintrivir (a rhinovirus protease inhibitor, see Section 6) [72c]. It should be noted that the key in the high activity of this catalyst can be found in the ligand’s steric constraints which prevents the formation of inactive trimeric species, which are formed for most of the Ir/P-N catalyst [78].Over the years, Zhou’s group have extensively studied the hydrogenation of several different classes of α,β-unsaturated carboxyxlic acids [77,79]. Thus, for example, excellent ee’s have been attained in the asymmetric hydrogenation of α-aryl- and α-oxymethyl-substituted cinnamic acids using Ir-SIPHOX catalysts (Fig. 16 ). This transformation were used in the preparation of (S)-(+)-homoisoflavone, a natural product with antibacterial activity (see Section 6) [80]. Latter, several heterocyclic olefins containing a carboxylic acid group were also successfully hydrogenated (ee's up to 99% ee, Fig. 16). Again Zhou’s group made use of this finding for the synthesis of the GABA uptake inhibitors (R)-tiagabine and (R)-nipecotic acid (see Section 6) [72e]. Interestingly, Ir/SIPHOX is also able to efficiently reduced a large number of tetrasubstituted acrylic acids (ee's up to 99%; Fig. 16), which have subsequently be used in the synthesis of the pyrethroid insecticide Fenvalerate and the antihypertensive drug Mibefradil (see Section 6) [81]. Interestingly, Ir/SIPHOX catalyst maintained its efficient when the carboxylic is moved away from the olefin. Thus, a range of β,γ-unsaturated acids [72d] and terminal γ,δ-unsaturated acids [72f,82] were successfully hydrogenated (ee’s up to 99%; Fig. 16) using Ir/SIPHOX (Ar = Xyl and R = 2-naphthylmethyl) catalyst. Again, the total synthesis of several natural products (e.g. (R)-xanthorrhizol, (R)-aristelegone-A …) were attained (see Section 6). Finally, Ir/SIPHOX (Ar = Xyl and R = H) demonstrated its usefulness in the hydrogenation of terminal olefins containing a benzoic acid group (ee’s up to >99%; Fig. 16), leading compounds with a benzylmethyl stereocenter like those found in the natural sesquiterpene phenols (S)-curcudiol and (S)-curcumene (see Section 6) [82]. Very recently, Zhang’s group have accounted an oxa-spirocyclic version of L32 that has demonstrated their usefulness in the hydrogenation of terminal methylene-tetrahydro-benzo[d]azepin-2-ones [83].The mechanistic study, including DFT calculations, agrees with an Ir(III)/Ir(V) catalytic pathway [84]. More important, using sodium 2-methyl-3-phenylacrylate as a benchmark substrate they were able to isolate migratory insertion Ir(III) intermediate 3, with the carboxylate coordinated to iridium (Fig. 17 ). They also managed to characterize by X-Ray diffraction dimeric hydrido complexes 4 and 5 (Fig. 16), which further support the coordination of the carboxylate to iridium.In this section, a collection of the developed heterodonor P,N-ligands with a N-donor group other than an oxazoline ring is presented. Besides the common tuneable properties of chiral ligands, e.g phosphorus group, backbone and source of chirality, P,N-other ligands allow variation on the hybridization of the N-atom. The tuning on the N-donor group leads to a wide array of P,N-other ligands. Ligands with amino N-donors (N-sp3), imino N-donors (N-sp2), cyclic imino N-donors (N-sp2) and pyridino N-donors (N-sp2) have been synthesized and used in several metal-catalyzed asymmetric reactions [3d,85].The first reports on chiral P,N-ligands came out back in the 70′s with the work of Hayashi and Kumada, where they showed that chiral aminophosphines were promising ligands for asymmetric catalysis [86]. In their early reports they developed the first P,N-ligands bearing planar chirality, the (aminoalkylferrocenyl)phosphines PPFA and MPFA (L35 and L36, respectively, Scheme 18 ). PPFA and MPFA were synthesized by introducing a phosphino group into α-dimethylaminoethylferrocene (6) through stereoselective lithiation [86a]. The overall synthesis starts with the transformation of ferrocene to intermediate 6 in 6 steps, via the formation of formylferrocene (Scheme 18) [87]. Both ligands were initially applied to the Rh-catalyzed hydrosilylation of ketones with high yields (up to 89%) albeit with 49% ee [86a]. PPFA was also used in the Ni-catalyzed asymmetric Grignard cross-coupling with (1-phenylethyl)magnesium bromide and vinyl bromide providing an enantioselectivity of 63% ee [86b]. Later, the same synthetic strategy was used to prepare the more constrained PTFA ligand L37 reported by Weissensteiner, but ligand L37 was prepared from intermediate 8 through formation of the corresponding imine (Scheme 18) [88]. The decrease flexibility on ligand L37 was beneficial for the enantioselectivity in the Ni-catalyzed asymmetric Grignard cross-coupling, affording an ee of 79% [89]. In 2006, Jin’s group synthesized L38 with a cyclic amine as nitrogen donor. L38 was synthesized from formylferrocene via diphenylphosphinoferrocenecarboxaldehyde obtained through Kagan’s method (Scheme 18) [90]. This modification was found to be beneficial for the asymmetric induction, which was illustrated with the excellent enantioselectivities obtained in the Pd-catalyzed allylic substitution with dimethylmalonate (99.6% ee) [91], which were comparable to the obtained with its P,N-oxazoline analogue [92].Ferrocenes with planar chirality attracted considerable attention in asymmetric catalysis, especially PPFA, which has been widely used for other researchers in many asymmetric transformations (Scheme 19 ). For example, Wang accounted the use of PPFA to the Cu-catalyzed addition of diethylzinc to imines, affording high enantioselectivities (up to 97% ee; Scheme 19a) [93]. Later, Sestelo and Sarandeses employed PPFA as the chiral ligand for obtaining 1,1′-binaphthyls by Pd-catalyzed cross coupling reactions of triorganoindium reagents, with ee’s up to 86% (Scheme 19b) [94]. Very recently, Guo has shown the use of PPFA in the Ag(I)-catalyzed tandem [3 + 2] cycloaddition/1,4-addition between aza-o-quinone methides (ao-QMs) and azomethine ylides, yielding imidazolidine derivatives with excellent diastereo- and enantioselectivities (up to 20:1 dr and 98% ee; Scheme 19c) [95]. The strategy has been extended to the addition of arynes, generated in situ from o-silylaryl triflates (Scheme 19d) [96]. Finally, PPFA has also showed utility in the Ir-catalyzed ring-opening of low-activity azabenzonorbornadiene with various aliphatic and aromatic amines, providing the corresponding chiral vicinal 1,2-diamine scaffolds in high yields and enantioselectivities (up to 97% ee; Scheme 19e) [97].Hayashi described the transformation of (S,R)-PPFA (L35; Scheme 18) to P-imine ligands L39 in three steps (Scheme 20 ). To achieve new ligands L39 the dimethylamino group on (S,R)-L35 was replaced by an amino moiety via acetate intermediate 10. Next, condensation of amino intermediate 11 with benzaldehyde furnished P-imine ligands (S,R)-L39 [98]. Ligands with electron-withdrawing groups at the aryl moiety provided higher enantioselectivities than its N-sp3 counterpart PFFA (L35) (up to 90% ee vs 16% ee) in the Rh-catalyzed asymmetric hydrosilylation of acetophenones [98]. Ligands L39 were also screened in the Pd-catalyzed allylic alkylations and again, the best selectivities were offered with a ligand having an electrodeficient aryl group. Using dimethylmalonate as nucleophile, the substitution of diphenylallyl acetate and pivalate and some cyclic substrates furnished promising enantioselectivities (up to 96% ee and 91% ee, respectively) [99]. Since then, other ferrocenylphosphino-imine ligands have been prepared from ferrocenylphosphino-amine compound 11 and the corresponding aldehyde (ligands L40–L41; Scheme 20) [100]. In contrast, ligand L42 was prepared by mixing 11 with imidate 13, which was prepared from the desired 2-methylbenzonitrile (12) in 4 steps (Scheme 20) [101].In general, ligands L40–L42 attained also high enantioselectivities in the benchmark Pd-allylic alkylation [100,101]. In addition, ligands L41 containing quaternary ammonium salts were also tested in the substitution of the benchmark substrate with different carbon nucleophiles and benzyl amine, providing enantioselectivities up to 94% ee (L41, X = I, n = 0) [100b]. This ligand was also employed in allylic etherification reactions with a range of benzyl alcohols, providing enantioselectivities up to 91% ee, but moderate yields (up to 74%) [102]. Ligands L42 (R = 5-Cl) bearing an imidate moiety showed even higher enantioselectivities in the allylic substitution of 1,3-diphenylallyl acetate using different malonate nucleophiles (up to 99% yield and >99% ee) [101]. In addition, the alkylation of cyclic substrates and the unhindered substrate rac-1,3-dimethyl-3-acetoxyprop-1-ene also resulted in good enantioselectivities (ee’s ranging from 75 to 90%). Phosphino-imidate ligands L42 were also applied in the Ir-catalyzed hydrogenation of di-, tri- and tetrasubstituted olefins, providing moderate to good enantioselectivities (45–91% ee) [103].Similarly, α-phosphino-β-imine ligands L43–L45 were prepared from the corresponding aldehyde and amine. Phosphine-hydrazone ligands L43 were obtained through condensation of intermediate 7 (also used for the synthesis of L38; Scheme 18) and the corresponding pyrrolidine 14 (Scheme 21 ). These ligands were screened in the Pd-catalyzed allylic substitution of the benchmark linear substrate with dimethyl malonate (L43, R = H; 96% ee) and benzylamine (L43, R = Me; 96% ee) [104]. Ligands L44, derived from condensation of 1-ferrocenylalkylamine 15 and 2-(diphenylphosphino)benzaldehyde 16 (Scheme 21), provided a better enantioselectivity of 97% ee in the benchmark reaction [105]. Ligands L45, having a phenyl-chromium tricarbonyl motif as planar chirality source (Scheme 21), gave even a higher enantioselectivity (>98% ee) than L43 and L44 when tested in the benchmark allylic substitution reaction [106]. Studies showed that for these ligands the enantioinduction is predominantly controlled by the planar chiral element and increases with the bulkiness of the N-substituents of the imine.Recently, (S,R)-PPFA ligand (Scheme 18) has been modified to provide ferrocene ligands L46 with a benzimidazole moiety (Scheme 22 a). As for ligands L39–L41, their synthesis proceeds through acetate intermediate 10 but in this case, amine 18 was subjected to reaction with benzimidazole 19 to yield the corresponding ligand [107]. Ligand L46 was key to achieve the highly enantioselective Pd-catalyzed [3 + 2] cycloaddition of propargylic esters with β-ketoesters, providing high yields and enantioselectivities (up to 98% ee). The reaction gave access to a range of chiral 2,3-dihydrofurans with an exocyclic double bond that remain unavailable with the known synthetic methods (Scheme 22b).P,N-ligands with axial chirality have found place in a broad range of applications in asymmetric catalysis [85e,g]. The advances of BINAP ligand in Ru-catalyzed asymmetric hydrogenations [108] together with the contemporary success of ferrocene-based P,N-ligands by Hayashi and Kumada [86a-b,d], led to the development of QUINAP (L47), which can be considered as the first highly efficient axially chiral P,N-ligand [109]. The crucial step in the synthesis of the racemic ligand was achieved via Pd-catalyzed cross-coupling of aryl chloride 20 and boronic acid 21 (Scheme 23 a). After cleavage of the methyl ether group, the phosphine group was introduced through conventional chemistry. At this point, it was required a final diastereomeric resolution of the corresponding palladium salts obtained through reaction of rac-22 and palladium complex 23. Diastereomers 24 could then be decomplexed by reaction with a 1,2-bis-(diphenylphosphino)ethane (dppe) to furnish the enantiomerically pure (R)- and (S)-L47 [109,110]. However, this methodology implied two main drawbacks. First, stoichiometric amounts of chiral palladium complex 23 were required, and second, the introduction of the phosphine group had to be done prior to the resolution step. This implied a resolution for every single ligand and thus, limiting the access to ligand diversity. These limitations have prompted to the search of more straightforward synthetic routes for the synthesis of enantiopure QUINAP and related ligands, by many research groups [111]. A novel way for the synthesis of QUINAP came in 2013 by Stoltz et al [111f]. The method consisted in the Pd-catalyzed asymmetric phosphination of aryl triflate 25 via dynamic kinetic resolution using Pd/26 catalytic system (Scheme 23b). Recrystallization further increased the ee from 90% to up to >99.5%. In 2016, Lassaletta reported a new methodology for the synthesis of QUINAP via Pd-catalyzed dynamic kinetic C–P cross-coupling between triflate 25 and a trimethylsilylphosphine, with the use of a Josiphos-type ligand 27 ( Scheme 23c) [111g]. This method gave access to QUINAP with 91.5% ee as well as several other potential P,N-ligands with axial chirality.Over the years it has been demonstrated that QUINAP is among the most outstanding axially chiral P,N-ligands with applications in many enantioselective transformations [85e,g]. The initial work with QUINAP was focused on Rh-catalyzed hydroboration of aryl alkenes and Pd-catalyzed allylic alkylation. Later it has also showed its utility in several other asymmetric transformations. Brown et al. early demonstrated the value of QUINAP in the Rh-catalyzed hydroboration of vinylarenes, which after oxidation led to variety of secondary alcohols with ee’s up to 96% (Scheme 24 a) [112]. This methodology was also used in the synthesis of primary and secondary chiral amines with good to high enantioselectivities (77–98% ee, Scheme 24b) [113]. In this case, the obtained chiral catecholboronate esters were transformed to the desired amines through alkylation with MeMgCl or ZnEt2, followed by conventional electrophilic amination with NH2OSO3H (for primary amines) or R′NHCl (for secondary amines).Morken and co-workers found that QUINAP was an excellent ligand also for the enantioselective Rh-catalyzed diboration of alkenes with dicathecol diboron [114]. It efficiently catalyzed the reaction of alkenes with dicatechol–diborane, to yield the syn-addition products. Subsequent oxidation yielded the corresponding enantiopure diols (Scheme 24c). The system showed a big scope for trans-disubstituted olefins, and unlike the Rh/Quinap-catalyzed hydroboration, the reaction occurred also with purely aliphatic alkenes. For trisubstituted alkenes, the enantioselectivities were also very high although yields were somewhat lower, while monosubstituted and cis-substituted alkenes reacted with lower enantioselection. Finally, Morken’s system also allowed the Rh-catalyzed tandem diboration/Suzuki/oxidation reaction to provide several chiral 1-aryl-2-ols in one-pot and in an operationally simple way (Scheme 24d) [114b].As stated above, the early work with QUINAP showed that it was also useful in the Pd-catalyzed alkylation of rac-1,3-diphenylallyl acetate and dimethyl malonate (98% ee, Scheme 25 a) [115]. Recently, Lassaletta and co-workers accounted the use of Pd/QUINAP catalyst in the dynamic kinetic asymmetric Buchwald-Hartwig amination and alkynylation reactions (Scheme 25b,c) [116]. Thus, a variety of enantiopure amino- and alkynyl-heterobiaryls were attained in high yields and ee 's up to 93% and 98%, respectively. Both processes were used to access to different axially chiral ligands, such as the IAN-type N,N-ligands [116b] Knochel and co-workers were the first in showing the utility of QUINAP in Cu-catalyzed coupling reactions. Initially, the Cu(I)/QUINAP catalyst was applied to a range of enamines with terminal alkynes providing the corresponding propargylamines in up to 90% ee (Scheme 26 a) [117]. Later, it was found that the same system allowed the three-component reaction between aldehydes, amines, and alkynes (A3 coupling). A wide range of propargylic amines could be afforded with good yields and enantioselectivities without the need to preform sensitive enamines (Scheme 26b) [118]. QUINAP was also screened in the Cu-catalyzed β-borylation of α,β-unsaturated esters, but while excellent conversions were achieved, the enantioselectivities didn’t surpass 79% ee (Scheme 26c) [119]. Schreiber applied the Cu/QUINAP system to the alkynylation of different isoquinoline iminium salts providing chiral 1-alkynyl tetrahydroisoquinoline derivatives (ee’s up to 99%; Scheme 26d) [120]. More recently, the Cu/QUINAP has been combined with a photo redox catalytic system for the cross-dehydrogenative-coupling of alkynes to N-aryl tetrahydroquinolines (Scheme 26e) [121]. This strategy allows the direct use of tetrahydroquinolines without preformation of the iminium salt maintaining the high ee’s (up to 96%), albeit with moderate to high yields (up to 90%).The QUINAP ligand was found to be an excellent candidate for the Ag-catalyzed [3 + 2] cycloaddition reaction of tert-butyl acrylate with azomethine ylides. The reaction gives access to pyrrolidines with multiple stereocentres, with an endo:exo ratio of >20:1 and up to 96% ee (Scheme 27 a) [122]. In 2013, Reisman expanded this methodology to the preparation of pyrrolizidines with up to 6 stereogenic centers in one flask with up to 94% ee (Scheme 27b) [123].In 2010, Murakami reported the Ni-catalyzed allene cycloaddition reaction of 1,2,3,4-benzothiatriazine-1,1(2H)-dioxides and allenes using QUINAP (Scheme 28 ) [124]. The reaction showed a broad scope with enantioselectivities up to 97% ee.The success with QUINAP pushed other researcher to explore other related atropisomeric P,N-ligands L48–L52 (Scheme 29 ) in asymmetric catalysis. QUINAZOLINAP (L48) [125] and PyPHOS (L49) [126] were prepared following a similar route than the used originally for QUINAP. These ligands also required to be resolved for which stoichiometric amounts of the chiral Pd complex 23 were needed. In addition, in the case of QUINAZOLINAP, the resolution procedure had to be further modified depending on the steric bulk of the substituent at the 2-position. This was not the case for PINAP ligands (L50–L52), which were designed to overcome these drawbacks. For these ligands, the diastereoisomers can therefore be separated by crystallization or column chromatography (Scheme 29) [127]. Thus, the racemic backbone was first easily prepared by selective oxidative Friedel–Crafts coupling of the dichlorophthalazine with 2-naphthol. Then, heteroaryl chloride 28 could react with (R)-phenylethanol (29) followed by triflation to provide 30, or being first subjected to triflation and then react with the desired chiral amine 31 to provide 32. Finally, using the same methodology as in the synthesis of QUINAZOLINAP, 30 and 32 were phosphinated to furnish PINAP type ligands L50–L52 [128]. Separation of diastereomeric (R,Sax )- and (R,Rax )-mixtures were then easily done by column chromatography [128].The application of ligands L48–L52, showed in some cases better results and even broader applicability than QUINAP. For instance QUINAZOLINAP ligands exhibited even slightly higher enantioselectivities in the Rh-hydroboration of a broad selection of vinylarenes, thanks to the tunability of the substituent at the 2-position of the atropoisomeric backbone [125b,h]. Concretely, the use of 2-methyl QUINAZOLINAP (L48; R = Me) provided excellent enantioselectivities (up to 99.5% ee). Note that the high ee’s are maintained when using tri- and tetrasubstituted vinyl arenes (e.g. indene, stilbene and 1,2-dihydronaphthalene), which usually proceeded with low enantioselectivities. L48 were also applied to Pd-catalyzed allylic alkylations. The enantioselectivity was dependent on the 2-position of the quinazoline backbone, being ligands with a 2-iPr substituent the most enantioselective (up to 94% ee) [125b,f,g]. PyPHOS (L49) ligands also proved to be useful in the Rh-catalyzed hydroboration of vinylarenes.As previously commented, PINAP (L50–L52) ligands have the advantage over QUINAP and related ligands of not requiring chiral Pd salts for their resolution, and therefore they are easily accessed. Moreover, Carreira showed that O-PINAP ligand (L50) gave comparable ee’s than QUINAP in the hydroboration of styrenes (up to 94% ee) and the cycloaddition of azomethine ylides and acrylates (up to 95% ee) [127a]. More important is the excellent performance achieved in Cu-catalysis by the PINAP ligand family. N-PINAP (L51) exhibited even higher enantioselectivities in the Cu-catalyzed A3 coupling for the preparation of propargyl amines (90–99% ee) [127a]. Later, the scope of the reaction has been further extended by Carreira and Ma. For instance, Carreira reported the preparation of propargylic amines bearing the more labile group 4-piperidone, which allowed easy deprotection to afford propargylic primary amines 33 in high yields (up to 92%) and up to 96% ee (Fig. 18 ) [129]. Ma developed a highly enantioselective A3 coupling of pyrrolidine, 2-methylbut-3-yn-2-ol and several aromatic aldehydes, which previously had shown lower enantioselectivities than aliphatic aldehydes (Fig. 18, compounds 34). A range of propargylic amines were obtained in 91−>99% ee [130]. The Cu/PINAP three-component coupling of propargylic alcohols, aldehydes and pyrrolidine was also used for the synthesis of chiral allenols 35 (Fig. 18) [131]. The reaction proceeds through formation of the corresponding propargylic amine and posterior Zn-mediated deamination. More recently, Ma has disclosed the A3 coupling of terminal alkynes, aldehydes and 3-pyrroline or isoindoline and subsequent [1,5]-hydride transfer catalyzed by CuBr to provide the (E)-N-allyl pyrroles with high yields. By using N-PINAP ligands (L51) it was possible to achieve the chiral (E)-N-allyl pyrroles in 97% ee and the four possible diastereoisomers of 37 in >99% ee (Fig. 18) [132]. The synthetic applicability of the A3 coupling using the Cu/N-PINAP system has been recently shown by Oguri, who used this methodology to prepare anti-malarial 6-aza-artemisinins in only four steps [133]. It has been also found that the use of tetrahydroisoquinoline as an amine source in A3 couplings affording tetrahydroisoquinoline-alkaloid derivatives (38, Fig. 18) [134]. The corresponding products were furnished with excellent yields and high enantioselectivities (up to 95% ee). This strategy has been fruitfully applied in the total synthesis of several natural products, such as (+)-crispine A and (+)-dysoxyline with an excellent ee of 98% [134], and various naturally occurring alkaloids (see Section 6) [135].Another important application of Cu/PINAP catalytic systems is the alkyne conjugate addition to Meldrum’s acid derivatives. The initial catalyst screening showed that while phosphine ligands (e.g. Josiphos, BINAP and Monophos) and N,N-ligands gave low ee’s (up to 25%), the first generation of PINAP ligands (L50–L52) gave moderate yields (up to 58%) and enantioselectivities (up to 80% ee). The incorporation of amino alcohols derived from amino acids in the 2-position of the PINAP scaffold, lead to more enantioselective ligands for this transformation (Scheme 30 , ligand L52). In addition, methoxy-substituted ligands (L52) catalyzed the reaction faster. With the optimized ligand, the addition of phenylacetetylene to various Meldrum’s acids in aqueous media proceeded smoothly and with enantioselectivities of 82–97% ee [136].Finally, Gu et al. showed a different application for O-PINAP (L50) ligands. Thus, a range of quinilinoferrocenes with a planar chirality were attained via Pd-catalyzed asymmetric intramolecular CpH bond functionalization/cyclization reaction of 2-halophenyl ferrocenecarboxylic amides [137]. However, only moderate enantioselectivities were achieved (up to 67% ee, Scheme 31 ).Another important family of axially chiral ligands are the nitrogen analogues of Hayashi’s MOP ligands [138], the MAP ligand family L53 (Scheme 32 ) [139]. Kočovský and co-workers synthesized for the first time the MAP ligands from the known biaryl amino alcohol NOBIN. First, alkylation of the amine group takes place, followed by the introduction of the diphenylphosphine group on triflate intermediate 39 trough Pd-catalyzed coupling with R2P(O)H. Successive reduction of intermediate 40 affords the desired MAP ligand. Although designed to act as bidentate P,N-donor ligands, NMR studies of the PdCl2/MAP complex showed a mixture of three species, where the major one was a cyclometallated complex [140].MAP ligand (R = Ph) was applied to various Pd-catalyzed asymmetric transformations, namely asymmetric allylic alkylation, Hartwig-Buchwald aminations and Suzuki cross-couplings (ee's up to 73%) [141]. Next, related L53 with R = Cy allowed for the first time the preparation of enantiopure chiral biaryls with up to 92% enantiomeric excess through asymmetric Suzuki-Miyaura (Fig. 19 [142]). Later, the substrate scope of boronic acids and aryl halides was expanded, including also axially chiral heteroaromatic and biphenyl compounds (Fig. 19) [143].Following the same synthetic strategy than for MAP-ligands, Ding prepared the octahydro analogues H8-MAP (H8-L53, Fig. 20 ) which gave higher enantioselectivities than L53 in the Pd-catalyzed alkylation of (E)-1,3-diphenylallyl acetate (83% ee vs. 73% ee) [144]. The higher enantioselectivity obtained was attributed to the larger bite angle of the H8-L53 ligands [145].Axially chiral P,N-ligands L54–L57 with a rigid amide linker are also derived from NOBIN (Scheme 33 ). To synthesize L54, NOBIN is first transformed to amino-phosphine compound 41 in 3 steps, which then reacts with 2-picolinic acid 42. Ligands L54 were screened in the Cu-catalyzed 1,4-addition of diethylzinc to different linear enones. A ligand with a 2-Me-pyridine moiety (L54, R = Me) provided the highest enantioselectivities (up to 98% ee). A promising enantioselectivity (up to 98 ee) was obtained also for the purely aliphatic enone (E)-5-methylhex-3-en-2-one [146]. More recently, the presence of a phenyl group in the ortho-position of the pyridyl moiety (L54, R = Ph) allowed the conjugate addition of various aldehydes with good yields (78–90%) and good to high enantioselectivities (75–98% ee) [147].Later, Hu and co-workers described the analogues phosphinite ligands L55–L57 (Scheme 33). The synthesis of (S)-L55 is shown is Scheme 33. In this case the pyridine-carboxylate moiety was installed prior to the insertion of the phosphorus group. Next, intermediate 43 was mixed with (S)-Feringa’s phosphoroamidite (S)-44 to afford the desired ligand. (S)-L55 (R = Me) afforded enantioselectivities up to 97% ee in the Cu-catalyzed conjugate addition of diethylzinc to linear enones (uo to 97% ee) [148]. H8-NOBIN-derivatives L56 and L57, which were obtained in a similar manner than L55, showed also an excellent catalytic performance [149]. However, ligands L55 as well as L54 failed for cyclic enones, with ee’s not higher than 53% ee.Ligands (R)-L54 and (S,S)-L55 (R = Me) have been efficiently used in different asymmetric tandem Cu-catalyzed Michael/Mannich reactions, furnishing excellent enantioselectivities (Scheme 34 ). The procedure allowed the synthesis of a broad range of chiral functionalized products with multiple stereocenters, which could be used to prepare valuable compounds, such as pyrrolidines, isoindolinones and azetidines [150].The more electron-rich chiral biphenyl backbone 45 has been also used to prepare atropoisomeric P,N-ligands (Scheme 35 ). Phosphine- and phosphite-pyridyl ligands L58 gave excellent enantioselectivities in the Cu-catalyzed conjugate addition of diethylzinc to linear enones (ee’s up to 96%) [151].Atroposiomeric ligands L59 with a 2-pyridyl moiety to a binepine scaffold have shown excellent results in asymmetric Pd-catalysis. Besides axial chirality, these ligands contain two elements of central chirality, one in the benzylic position and the other at the phosphorus atom (Scheme 36 ). The synthesis of ligands L59 starts from dilithiated binaphthyl compound 46, which is transformed to borane-protected binepine 47 in 4 steps. Then, the benzylpyridine moiety is incorporated in the presence of BuLi, providing 48 as a single diastereoisomers except in the cases where the phosphorus center was a phenyl group [152]. Finally, borane-protected compounds 48 were deprotected by refluxing them in an excess of diethylamine for 4 days, yielding ligands L59 as off-white crystalline with good to excellent yields.. Ligands L59 were first tested in the Pd-catalyzed intramolecular α-arylation of α-branched aldehydes, surpassing the results achieved with QUINAP, PINAP or PHOX. The ligand screening revealed that a ligand with a more electronrich aromatic substituent in the P-group and a non-substituted pyridyl moiety (L59a) showed the best catalytic performance (Scheme 37 a). A selection of aldehydes could be used furnishing the corresponding cyclic products with excellent enantioselectivities (up to >99% ee) and good yields [152] It was also successfully used in the Pd-catalyzed Heck reaction between 2-substituted furans and aryl triflates to yield functionalized 2,5-dihydrofurans with fully saturated C2 stereocenters. In this case the optimal ligand contained a tert-butyl moiety at the phosphorus group (L59b), showing moderate yields but enantioselectivities up to 94% ee for a range of substrates (Scheme 37b) [153]. To gain information about the catalytic species formed during catalysis, the authors performed the complexation of ligands L59 to [PdCl2(CH3CN)2], leading to air-stable complexes [(L59)PdCl2]. NMR and IR spectroscopy proved the bidentate coordination of the ligand, which was also corroborated with the molecular geometry obtained by single-crystal X-ray analysis. It was also found that the ligand adopts a nearly ideal square-planar geometry and that the P-donor has a stronger trans effect than the N-donor atom [152].The first developed generation of axially chiral ligands were based on 6-membered heterocyclic motifs. All of them were built on a binaphthalene or a biphenyl backbones with an element in the ortho-position that hinders the rotation about the biaryl bond [108c-e]. The replacement of one of the naphthalene rings by a 5-membered ring was already attempted by Brown et al., with the synthesis of the indole-based ligand L60 (Scheme 38 ). However, the new ligand turned to be not configurationally stable. In 2003, Aponick designed the StackPhos ligand bearing an imidazole group (L61, R = Ph, Scheme 38), in which the π − π interaction between the electron-rich naphthalene ring and the electron-poor pentafluorophenyl group on the non-coordinating nitrogen is crucial to prevent tropoisomerism [154]. The synthesis of racemic L61 was achieved in 6 straightforward steps starting from 2-hydroxy-1-naphthaldehyde 49, in which the imidazole ring and the diphenylphosphino group were readily introduced (Scheme 38). In contrast to QUINAP and its derivatives, the resolution of L61 was achieved through deracemization instead of resolution. Reaction of racemic ligand with chiral Pd-salt 23 resulted in a single diastereoisomer, which was then treated with dppe to release ligand L61 in high yield and 98% ee. Later, related phosphinoimidazoline ligands L62–L63 (Scheme 38) were independently developed by Guiry (UCD-Phim) [155] and Aponick (StackPhim) [156]. The idea behind them was to circumvent the use of stoichiometric amounts of expensive chiral Pd-amine complex 23 to access the enantiopure ligands. Similarly to Carreira’s PINAP, these ligands bear an element with central chirality that allows the separation of the diastereomeric mixture through recrystallization or column chromatography. Note that StackPhim (L62) and UCD-Phim (L62) are diastereomers.The first application of the StackPhos ligand (L61, R = Ph) was in the Cu-catalyzed A3-coupling between dibenzylamine, trimethylsilylacetylene and a range of aldehydes, including the more challenging aromatic ones. The corresponding propargylamines 50 were yielded in high yields and ee’s up to 97% (Fig. 21 a) [154]. The protocol was extended to the synthesis of amino skipped diynes 51 (up to 96% ee), a class of chiral molecules with minimal differences in two of the substituents rendering them chiral (Fig. 21a) [157]. The Cu/StackPhos system has also allowed the enantioselective copper-catalyzed alkynylation of quinolinium salts and chromanones, delivering the desired products 52 in high yields and enantioselectivities (up to 98% ee; Fig. 21a) [158]. The potential of the reaction was demonstrated in the syntheses of the tetrahydroquinoline alkaloids [158a] (+)-galipinine, (+)-cuspareine, and (−)-angustureinem, as well as (−)-martinellic acid (see Section 6) [159]. More recently, a set of StackPhos ligands bearing different substituents in the imidazole ring have been applied to the alkyne conjugate addition to Meldrum’s acid derivatives. A ligand with a methyl group (L61, R = Me) in combination with Cu(OAc)2 exhibited the best catalytic performance (up to 92% yield and 98% ee) [160]. The transformation gives β-alkynyl Meldrum’s acid building blocks (compounds 53, Fig. 21a), which could be used in the asymmetric synthesis of the vasopressin V2-receptor agonist OPC 51,803 (see Section 6). More recently, Aponick’s group used L61 for the synthesis of chiral δ-lactones via a tandem acetylide addition/alkyne heterofuntionalization process catalyzed by Cu and Ag, respectively (Fig. 21b) [161].The newer UCD-Phim ligands L62, developed by Guiry and coworkers, showed excellent results in the Cu-catalyzed A3-coupling reaction of aliphatic aldehydes, showing in some of the cases greater enantioselectivities than StackPhos ligands (up to 98% ee) [155]. In 2019, the scope was extended to aromatic, alkenylic and alkynylic aldehydes, as well as secondary cyclic amines, achieving up to 99% ee [162]. The StackPhim ligand L63, developed by Aponick, has been used to prepare C2-aminoalkyl five-membered heterocycle motifs (up to 94% ee, Scheme 39 ). The strategy used consists in a convergent alkynylation/cyclization sequence [156].Amino-phosphine ligands with a spiro center (L64–L66, Scheme 40 ) have proved to be highly effective in the hydrogenation of α,β-unsaturated ketones as well as alkenes bearing nitro or carboxylic acid groups when using Ir-catalysts. The first ligands of this class were the SpiroAP ligands (L64) developed by Zhou et al. [163]. The introduction of a CH2-group before the primary amino moiety, and later a CMe2-group, afforded chiral spiro benzylamino-phosphine SpiroBAP [164] and SpiroBAP-R [165] ligands (L65–L66). The synthesis of amino-phosphine spiro ligands L64–L66 starts with the transformation of commercially available SPINOL into diarylphosphine/triflate intermediate 54 (Scheme 40). To obtain SPiroAP ligands (L64), 54 is converted to the dimethyl ester derivative by Pd-catalyzed carbonylation, followed by subsequent basic hydrolysis to provide carboxylic acid derivative 55. In the case of SpiroBAP and SpiroBAP-R ligands (L65–L66), the synthesis proceeds through Pd-catalyzed cyanation of 54. Next, reduction of cyanate intermediate 56 with LiAlH4 or with MeLi afforded SpiroBAP and SpiroBAP-R ligands, respectively.SpiroAP ligands (L64) were screened in the Ir-hydrogenation of exocyclic α,β-unsaturated enones to afford exo-cyclic allylic alcohols and β-arylmethyl cyclic alcohols [163]. The corresponding Ir-complexes were formed in situ during catalysis. The best ligand contained a bulky phosphine group (Ar = 3,5-di-tert-butylphenyl), giving excellent enantioselectivities and yields (Scheme 41 ). The potential of this methodology was demonstrated with the synthesis of a crucial intermediate in the preparation loxoprofen, a nonsteroidal anti-inflammatory drug (see Section 6).Prompted by the excellent results obtained in the hydrogenation of ketones, Zhou and co-workers prepared spiro benzylamino-phosphine SpiroBAP (L65) [164]. In this case, Ir-complexes were synthesized prior to catalysis, following the same procedure used for preparing [Ir(cod)(L8)]BArF. The new complexes were stable to air and and could be stored without degradation for a few months. X-ray diffraction analysis showed that L65 (Ar = Ph) acts as a chelating P,N ligand and creates a rigid chiral pocket around the iridium center. Again, the presence of a bulky aryl phosphine (L65, Ar = 3,5-tBu2-C6H3) exhibited the best catalytic results in the hydrogenation of α-aryl- and α-alkyl acrylic acids (Scheme 42 ). A range of chiral carboxylic acids, including naproxen and related anti-inflammatory drugs, were attained in excellent enantioselectivities (up to 98%) and TOFs (up to 6000 h−1).The presence of a dimethyl group at the benzylic position of the amine group on SpiroBAP-R ligands allowed the enantioconvergent hydrogenation of β-aryl-β-methyl-nitroalkenes (91–98% ee values) and β-alkyl-β-methyl-nitroalkenes (77–95% ee values; Scheme 43 ) [165]. [Ir(cod)(L66)]BArF was able to hydrogenate diastereomeric mixtures of E- and Z-nitroalkenes, thus avoiding tedious isolation of the substrate isomers.Very recently, Jiao and co-workers have published the air-stable ligands L67 bearing a rigid spiro[indane-1,2′-pyrrolidine] backbone (Scheme 44 ) [166]. In contrast to SPINOL-derived ligands L64–L66, the spiral center must be created during the synthesis of ligands L67 (Scheme 44). Thus, the key step to build the spiral center is accomplished through AlCl3-mediated intramolecular Friedel–Crafts-type reaction of 58, which is obtained first from 57 after 5 steps. After three recrystallizations of the diastereomeric mixture of 59a–b, pure diastereoisomer 59a was afforded. Next, compound 60 is obtained via multiple steps. Demethylation of 60 followed by triflation of the phenolic hydroxy group, coupling with diphenylphosphine oxide and two subsequent reduction steps gave ligand L67. Ligands L67 were applied in the Pd-catalyzed asymmetric allylic substitution of benchmark substrate with dimethylmalonate but also with several alcohols and amines as nucleophiles with moderate to high yields and enantioselectivities (60–99% yield and 61–97% ee). The authors were able to obtain a crystal structure of [Pd(II)(η3-1,3-diphenylallyl)(L67, Ar = 3,5-tBu2Ph)]PF6, which gave information about the transition states of the substitution reactions.Many new P,N-ligands bearing central chirality has also been developed. Few years later of the discovery of PPFA ligands (L35), Hayashi and Kumada designed a new library of ligands L68 from natural α-amino acids (Scheme 45 ). Phosphine-amino ligands L68 catalyzed the Ni- Grignard cross-coupling of (1-phenylethyl)magnesium bromide and vinyl bromide, for which the bulkiest tert-Leuphos ligand L68 (R1 = tBu) provided the highest ee value (up to 94% ee) [86c]. The asymmetric induction was thought to result from the hemilability of these ligands. Several analogues of these chiral β-aminophosphine, have been developed throughout the years, because of their stability, low toxicity and ease handling. These ligands can be readily synthesized through nucleophilic phosphide substitution of derivatized amino alcohols 62 containing a leaving group (LG) (Scheme 45). Besides, this scaffold has been also used in the design of other P,N-ligands containing an imine, amide or pyridine as N-donor group. A review about β-aminophosphine derivative has been recently reported [85e]. Another example of β-aminophosphine ligands are compounds L69 derived from L-valine (Scheme 45), which were applied in the palladium catalyzed allylic substitution of (E)-1,3-diphenylallyl acetate with dimethyl malonate. The nitrogen substitution constituted a key factor in the stereochemical outcome of the reaction, with enantioselectivities that ranged from 56% ee (R) to 92% ee (S) [167].Air-stable phosphite-amino ligands L70 (Scheme 46 ) were prepared in two steps also from commercial 1,2-amino alcohols. These ligands were tested in enantioselective Pd-catalyzed allylic substitution reactions. A mechanistic study allowed the optimization of the ligand parameters from a full ligand library, identifying ligands L70a–b as the best. High enantioselectivities were achieved for a linear and cyclic substrate with several C-, N-, and O-nucleophiles (32 examples,  ee values up to 99%) [168]. Ligands L70a–b were easily obtained by methylation of intermediates 63 with formic acid and formaldehyde resulted in dimethylated amino alcohols 64. Subsequent reaction with the desired phosphorochloridite led to ligands L70a–b. Studies on the Pd-π-allyl intermediates provided insights about the effect of the ligand parameters on the origin of the enantioselectivity. It was found that the higher enantioselectivities obtained with ligands containing a hydrogen as the R2 substituent (L70), compared with ligands with a R2 = Me, were mainly due to a higher electronic differentiation between the more electrophilic allylic terminal C atoms, making the major Pd-η3 allyl isomer more reactive [168].β-Aminophosphine ligands derived from starting material other than amino acids have been also found to be efficient in Pd-allylic substitutions. For instance, phosphite-amino ligand L71 with a protected pyrrolidine-3,4-diol moiety has been recently prepared from cheap D-mannose (Scheme 47 ) [169]. N-Boc protected aminoalcohol 65, which was obtained from D-mannose [170], was subjected to Boc deprotection followed by reaction with the appropriate phosphorochloridite. The optimized ligand L71 was employed in the Pd-catalyzed enantioselective allylic substitution of linear and cyclic substrates. Enantioselectivities ranging from 80 to 91% ee were obtained using various C- and N-nucleophiles. In the case of cyclic substrates both enantiomers of the final products could be attained by switching the chirality of the biaryl phosphite group. A study of the Pd–π-allyl intermediates showed that to achieve high enantioselectivities in the substitution of cyclic substrates, the ligand components need to be appropriately chosen to either enhance the difference in the ratio of the Pd–allyl isomers formed or to enhance the reactivity of the nucleophile towards each Pd–allyl isomer. In contrast, the key of success when using linear substrates is to avoid the formation of Pd–allyl complexes with monodentate coordinated ligands. The study also indicated that the sugar backbone is able to control the configuration of the amino group upon coordination [169]. Note that L71 contains an amino group that is part of a cyclic backbone. Most of the P-amino ligands that exhibited remarkable results in asymmetric catalysis contain a non-cyclic amino group [85b,d]. Only few P,N-ligands bearing a cyclic amine have provided high enantioselectivities, mostly achieved only in the benchmark substrate [97,171].In 2011, a new family of cinchona-derived phosphino-amine ligands (L72–L73) was developed by Dixon and co-workers [172]. These ligands consist on a cinchona backbone bearing three different sites that allow a cooperative catalysis: a Brønsted (N) and a Lewis base (P) and a H-bond-donor group (NH). Ligands L72 and L73 were readily prepared from commercially available ortho-diphenylphosphino benzoic acids and the desired 9-amino(9-deoxy)epicinchona alkaloids (66a–b; Scheme 48 ).Phosphine-amino ligands L72 and L73 (R1 = Et; R2 = H; Ar = Ph) were mixed with Ag2O to yield cooperative Ag(I)-based Brønsted base/Lewis acid catalysts that effectively promoted the asymmetric aldol reaction of isocyanoacetate nucleophiles and aldehydes (Scheme 49 ) [172]. Both aromatic and branched aliphatic aldehydes could be used to provide oxazolines with high diastereo- and enantioselectivities (up to 98%) by using ligand L73 (R1 = Et; R2 = H; Ar = Ph). However, linear aliphatic aldehydes showed lower enantioselectivities.After its first application, Ag(I)/L72 catalyst has been efficiently used in the enantioselective catalytic addition of isocyanides to many other electrophilic compound such as aldehydes [173], aldimines [174], ketones [175] ketimines [176], allenoates [177], alkynyl ketones [178], other carbon–carbon double bond containing electron withdrawing groups (EWG) [179], and p-quinone methides (p-QMs) [180]. Besides isocyanide chemistry, quinine-derived ligands L72 have recently showed impressive results in asymmetric Cu-catalyzed cross-coupling reactions (Scheme 50 ) [181]. Ligand L72a (R1 = vinyl; R2 = OMe; Ar = 3,5-tBu2-C6H3) allowed the largely unexplored asymmetric Cu-catalyzed Sonogashira Csp3-Csp cross-coupling between a range ofalkyl halides and alkynes (>120 examples, up to 99% ee; Scheme 50a). To show the utility of this transformation, they performed the asymmetric Sonogashira C(sp3)C(sp) cross-coupling reaction of a mesogenic compound with the core structures of several bioactive molecules, such as estrone, biotin etc (see Section 6). L72b (R1 = vinyl; R2 = OMe; Ar = Ph) has also allowed the radical asymmetric oxidative C(sp3)C(sp) cross-coupling of unactivated C(sp3)H bonds on N-fluorocarboxamides with terminal alkynes (Scheme 50b) [182]. A range of chiral alkynyl amides were afforded in a highly regio-, chemo-, and enantioselective manner (up to 97% ee).Very recently, chinchona-derived ligands L72c–d (R1 = vinyl; R2 = OMe; Ar = 2,6-Me2-C6H3 or 9-phenanthryl) have been used in the Cu-catalyzed enantioconvergent radical Suzuki − Miyaura C(sp3) − C(sp2) cross-coupling of racemic alkyl halides with B(mac)-derived boronate esters (Scheme 50c) [183]. The reaction showed a broad scope regarding both coupling partners, including aryl- and heteroarylboronate esters, as well as benzyl-, heterobenzyl-, and propargyl bromides and chlorides furnishing high enantioselectivities.As ferrocene-based P-imine ligands L43–L45 (Scheme 21), the phosphine-imino ligands L74–L77 bearing central chirality (Scheme 51 ) were initially applied in Pd-allylic substitutions. All ligands could be easily prepared by mixing (diphenylphosphino)benzaldehyde 67 with a desired chiral amino scaffold (for ligands L74–L76) or a sulfinamide (for ligand L77).Thus, ligands L74, prepared from commercially available SAMP as the chiral amine, provided somewhat lower enantioselectivity (up to 92% ee) than related ligand L43 (up to 96% ee) [104] in the Pd-allylic substitution reactions [184]. Using available peptides as chiral amines, Hoveyda et al. prepared a large library of peptide-based ligands L75 and L76. Over the years, its modular nature has allowed to achieve excellent enantioselectivities in the Cu-catalyzed conjugate addition of a broad range of α,β-unsaturated substrates (Fig. 22 ) [185] For example, L75a (R1 = iPr, R2 = Bn, R3 = NHnBu) exhibited excellent enantioselectivities (up to 98% ee) in the conjugate addition of different alkylzincs to cyclic enones [185a,h], while L75b (R1 = iPr, R2 = p-OtBu-Bn, R3 = NHnBu) was preferred for the linear ones (Fig. 22) [185b,e]. Both ligands have been very useful also with unsaturated furanones, with ee’s up to >98% (L75a) and up to 97% ee (L75b) (Fig. 22) [185f]. When nitroalkenes were used as Michael acceptors, the best ligands were found to be L75c (R1 = tBu, R2 = p-OBn-Bn, R3 = NHnBu) for cyclic nitroalkenes (up to 96% ee) [185c], and L75d (R1 = tBu, R2 = p-OBn-Bn, R3 = NEt2) for trisubstituted linear nitroalkenes (up to 98% ee; Fig. 22) [185g,186]. The Cu/L75d was used by Carreira in the total synthesis of (+)-Daphmanidin E (see Section 6) [187]. The same ligand also allowed the tandem conjugate addition–nitro-Mannich reaction for the preparation of anti- and syn-β-nitroamines with three contiguous stereocenters (up to 96% ee; Fig. 22) [188].A ligand bearing only a peptidic fragment has been also efficiently used in the Cu-catalyzed conjugate addition of cyclic enones (up to >98% ee) [185d] and lactones (up to 96% ee) [185f]. Ligands L76 have been found to induce high enantioselection in a wide range of asymmetric CN bond forming transformations such as the aza-Diels-Alder [189] and Mannich type reactions [189b,190].P,N-sulfinyl imine ligands L77, in which the chirality is found in the sulphur atom, were obtained via Ti-mediated condensation of the corresponding sulfinamide with compound 67 (Scheme 51). With a tert-butyl substituent attached to the imine an enantioselectivity up to 96% ee in the allylic alkylation of the benchmark substrate was attained [191]. Ligands L77 were also used to the Ir-catalyzed hydrogenation of trisubstituted olefins, but with only moderate enantioselectivities [192].Other ligands bearing the chirality at the sulphur centre are the P-sulfoximine ligands L78–L81 (Scheme 52 ), which have been efficiently used in Ir-catalyzed hydrogenation reactions. Bolm et al. developed phosphinosulfoximine ligands L78 and L79 in few steps. The key step consists in the Cu-mediated coupling N-arylation of sulfoximines with bromo-aryl phosphine oxide 68. Reductive deoxygenation of 69 with trichlorosilane gave the corresponding ligands as solid, air-stable products in good yields Scheme 52 [193].Ligands L78 were screened to the enantioselective Ir-catalyzed hydrogenation of N-aryl imines, using iodine as a promoter. A ligand containing an isobutyl and a phenyl N-substituent provided the best catalytic performance. With the optimal ligand it was possible to reduce a range of imines with ee’s over 90% ee for most of the substrates [193]. Later, analogous bicyclic ligands L79 were prepared following a similar synthetic strategy starting from 1,8-diiodonaphthalene. These ligands showed 92% ee in the hydrogenation of 2-methylquinoline, albeit moderate enantioselectivities were obtained for other quinolone derivatives (55–87% ee) [194]. More important are the results attained in the olefin hydrogenation of β,β’-disubstituted enones [195]. Prior to this, the existing methods for preparing valuable optically pure ketones were mainly non-catalytic and with a limited substrate scope [196]. A range of enones could be reduced with enantioselectivities up to 97% ee (Scheme 53 ). However, the effectiveness of the catalyst is affected by the substitution pattern of the enone and the steric constraints of the olefin substituents. For instance, a low enantioselectivity was observed for α,β-disubstituted enones (55% ee for (E)-3-methyl-4-phenyl-3-buten-2-one) [197].To increase the scope of P-sulfoximine ligands in Ir-hydrogenation reactions, the phosphine group on L78 has been recently substituted by biaryl phosphite moieties, leading to ligands L80. The more rigid benzothiazine derivative L81 was also synthesized [198]. In contrast to L78–L79, the sulfoximine group was inserted prior to the P-group. The synthesis of L80–L81 starts with alcohol protection of the corresponding 1-Br-phenols 70 and 73 with methoxymethyl chloride. Then, MOM-protected intermediates were coupled with (S)-S-methyl-S-phenylsulfoximine (76), which upon deprotection of the MOM group in acid media, reacted with the desired phosphorochloridite. Finally, Ir-complexes [Ir(cod)(L80-L81)]BArF were prepared using the same methodology than for [Ir(cod)(L8)]BArF [198]. The authors found that Ir-complexes containing ligands L80 were obtained as a mixture of isomers. In contrast, complexes containing ligands L81 with a more rigid backbone were present as a single isomer, suggesting that in the case of ligands L80 two different stable conformations for the six-membered chelate ring are possible. One more time, having a biaryl phosphite moiety on the ligand scaffold improved the substrate versatility of the hydrogenation reaction. Thus, [Ir(cod)(L80–L81)]BArF complexes increased the scope attained in the reduction of α,β-unsaturated enones (95–97% ee), including α,β-disubstituted enones and with an exocyclic double bond. Furthermore, other olefins bearing poorly coordinative groups, such as lactones, diphenyl alkenylboronic esters among others, were also hydrogenated with ee’s up to 99% (Fig. 23 ).As mentioned previously (see Section 2) the phosphino-oxazoline ligands (PHOX) made a breakthrough in asymmetric catalysis due to their synthetic versatility and broad catalytic applicability [85c,199]. With the aim of exploring other five-membered nitrogen heterocycles, P,N-ligands bearing aromatic heterocycles such as oxazoles, thiazoles and imidazoles, as well as other non-aromatic rings (e.g. thiazolines and imidazolines) have been developed. The resulting P,N ligands have shown excellent results in asymmetric catalysis, especially in Ir-catalyzed hydrogenations and in Pd-catalyzed reactions.This field has been pioneered by Andersson's group with the aim to enlarge the substrate versatility in the challenging hydrogenation of unfunctionalized olefins. They started by rational design of the bicyclic oxazole-based P,N-ligands L82 (Scheme 54 ) [200]. Their synthesis starts with the transformation of diazodimedone in presence of a catalyst and benzonitrile. Catalytic enantioselective reduction of the obtained keto oxazole with (R)-Me-CBS-borane provided (S)-alcohol derivative, in which the desired phosphinite group was installed through conventional chemistry. These ligands met the criteria established from a computational study made by the same authors about the hydrogenation of (E)-1,2-diphenyl-1-propene with the Pfaltz Ir/PHOX-catalyst [201]. Thus, ligands L82 combines the presence of a P- and a N-donor atom, to achieve a significant trans effect, with a rigid bicycle to reduce conformational flexibility, and a six-membered chelate ring is generated upon complexation to Ir. All this resulted in Ir/L82 complexes that generate an appropriate chiral environment for asymmetric induction to the substrate (Scheme 54). The outstanding enantioselectivities (93–99% ee) achieved by the Ir/L82 complexes in the hydrogenation of 1,2-disubstituted styrenes corroborated the computationally derived selectivity model.Then, by systematic modification of L82 by replacing either the oxazole by a thiazole group or the phosphinite group by a N-phosphine moiety, phosphine-thiazole L83 [202] and N-phosphine-thiazole ligands L84 [64e] were obtained (Scheme 55 ). In the case of L83, 5- and 6- and 7- fused-rings were studied. Both ligand families are derived from ketoesters¸ which were transformed into thiazole esters 77 through condensation with benzothioamide. Next, for the preparation of phosphine ligands L83, the corresponding thiazole ester 77 was converted to alcohol (rac)-78. In contrast, when the target was the aminophosphine ligand L84, ester 77 was first converted to the amide 79 and then it was reduced to amine (rac)-80. Both, alcohol and amine derivatives were obtained as racemates and resolved by preparative chiral HPLC. Finally, they were converted to the corresponding phosphine or aminophosphine ligands through already reported chemistry.The catalyst precursors [Ir(cod)(L83-L84)]BArF were prepared using the same methodology than for [Ir(cod)(L8)]BArF, and were tested in asymmetric hydrogenation reactions. By selecting the appropriate ligand L82–L84 it was possible to increase the substrate scope considerable. For instance, phosphine-thiazole ligands L83 showed to be more suitable than the oxazole counterpart L82 in the Ir-hydrogenation of α,β-unsaturated esters. To note that the six-membered ring backbone showed the best catalytic performance. With a di-o-tolyl phosphine moiety and the appropriated substituent on the thiazole ring, derivatives of α- and β-methyl cinnamic acid ethyl esters were reduced with high enantioselectivities (80–98% ee for (E)-olefins and 95% ee for a (Z)-olefin) (Fig. 24 ) [202].Instead, for the reduction of vinyl fluorides the best enantioselectivities was achieved with [Ir(cod)(L84)]BArF [64e]. At this time, the hydrogenation of fluorine-containing olefins was little explored, probably, due to the ability of vinylic fluorine to be cleaved off [203]. In contrast, the Ir/L84 catalyst showed little defluorination. Although the substrate scope was limited, it was possible to hydrogenate a trisubstituted fluoroolefin bearing a hydroxy and an acetate group in 99% and >99% ee, respectively (Fig. 24). Ir-catalyst bearing ligand L84 was also very useful in the hydrogenation of 1,1′-disubstituted vinylphosphonates and carboxyethylvinylphosphonates (Fig. 24) [204]. It should be noted that for this last type of substrates, E- and Z-isomers as well as their mixtures could be hydrogenated in excellent enantioselectivities (up to >99% ee). Finally, with the use of both thiazole ligands L83 (R = Ar = Ph, n = 1) and L84, it was possible to hydrogenate a range of 1,1′-diaryl substituted olefins in excellent enantioselectivities (up to >99% ee) [205].The corresponding phosphite-oxazole and thiazole-based ligands L85–L86 were also prepared and applied in Ir-catalyzed hydrogenation of unfunctionalized olefins or with poorly coordinative groups (Fig. 25 ). Thiazole-based ligands L86 provided the highest enantioselectivities. The introduction of the phosphite moiety allows to extend the substrate scope. Ir/L86 catalytic system allowed the hydrogenation of both E- and Z-trisubstituted olefins along with 1,1′-disubstituted terminal alkenes, furnishing excellent enantioselectivities (ee’s up to 99%) [206]. The catalytic system also tolerated the presence in the olefin of some neighboring polar groups (e.g. esters, alcohols, phosphinates …) for which ee values up to 99% has been also attained.Useful, ligands L85 and L86 were also screened in the Pd-allylic substitution reaction [207]. After ligand screening and in contrast to the hydrogenation, it was found that oxazole ligands (L85) exhibited in general higher enantioselectivities than L86. With the proper selection of each ligand parameter it was possible to achieve high enantio- and regioselectivities (ee up to 96%) for a variety of cyclic and tri-, di- and monosubstituted linear substrates (Fig. 26 ). The study of Pd-allyl intermediates by NMR and DFT aided to understand the influence of the ligands parameters on the ratio of the Pd-allyl species and the electrophilicity of the allylic terminal carbon atoms. Ligands L85 provided also high enantio- and regioselectivities in Pd-catalyzed intermolecular Heck reactions of 2,3-dihydrofuran with several aryl triflates [208].To study the effect of the backbone in the catalytic activity, Andersson’s group developed the other two ligand families L87 [209] and L88 [210] (Scheme 56 ), in which the backbone was modified while keeping the thiazole unit. Ligands L87, which feature an open-chain-backbone, were synthesized from inexpensive methyl 3-oxobutanoate (Scheme 56a). The synthesis starts with the introduction of the thiazole ring. Next, Oppolzer sultam (81) was introduced in order to incorporate a chiral alkyl chain by using diverse alkyl halides in the presence of lithium hexamethyldisilazide (LHMDS). Each chiral ligand precursor 82 was obtained with high diastereomeric purities. After reduction with LiAlH4, the corresponding alcohols were transformed to the final phosphino-thiazole ligands L87. Unfortunately, the new ligands L87 were slightly less successful than the their more rigid counterparts L83 in the Ir-catalyzed hydrogenation of unfunctionalized trisubstituted olefins, allylic alcohols and imines [209].In contrast, the introduction of a bicyclic amine into the ligand resulted beneficial in terms of scope for the Ir-hydrogenation of poorly coordinative olefins [64f,h,210,211]. In addition, their synthesis was achieved in fewer steps than L87 (Scheme 56b). Ligands L88 were synthesized from (1S,3R,4R)-2-azabicyclo[2.2.1]heptane-3-car-boxylic acid [212]. This intermediate was transformed to N-Boc-protected thioamide in 3 steps, which was then cyclized with an α-bromoketone to provide the N-protected thiazole. After removal of the protecting group, the phosphorus fragment was incorporated to ligands L88. The resulting Ir-complexes [Ir(cod)(L88)]BArF were efficiently used in the hydrogenation of olefins bearing a variety of poorly coordinating groups, showing comparable results to those obtained with phosphine-thiazole L84 [210]. Moreover, the use of Ir/L88 was beneficial, widening the scope of the hydrogenated olefins (Fig. 27 a). Excellent enantioselectivities were obtained for vinyl boronates (up to 98% ee) [64f], vinylic, allylic and homoallylic sulfones (up to 99% ee) [211a], γ-substituted cinnamyl alcohols (up to 99% ee) [64h] and very recently to α,β-unsaturated α-fluoro aryl and alkyl ketones (up to >99% ee) [211c]. Importantly, the use of an Ir-complex containing the enantiomer of L88 (R = Ph, Ar = o-Tol) allowed the cooperative dynamic kinetic asymmetric hydrogenation of allylic alcohols (Fig. 27b), to produce a broad range of chiral alcohols containing two stereogenic centres with excellent diastereoselectivities (up to 95:5) and enantioselectivities (up to 99%) [211b]. Mechanistic studies supported that racemization of the substrate is achieved by cleavage and reforming of the oxygen–carbon bond.The combination of a biaryl phosphite and the nitrogenated bicyclic backbone of L88, lead to the family of ligands L89 (Fig. 28 ) [66]. [Ir(cod)(L89)]BArF were synthesized following the same methodology than for [Ir(cod)(L8)]BArF. Ir-complexes with a ligand bearing an (S)ax-binaphthol moiety helped to expand the substrate scope of the previous successful N-phosphane ligands L88 on the Ir-hydrogenation of unfunctionalized olefins. For example, E- and Z-tri- and 1,1′-disubstituted substrates, α,β-unsaturated enones, alkenylboronic esters and disubstituted olefins were reduced with excellent enantioselectivities (up to 99% ee) [66] From the results obtained with the oxazole- and thiazole-based ligands developed by Andersson's group it can been concluded that they showed a different substrate scope in the hydrogenation of olefins. Having this in mind, the group developed a new set of ligands with an imidazole ring (L90; Scheme 57 ), to investigate even further the substrate scope [213]. The imidazole moiety was formed through condensation of ester 83 (obtained from esterification of 2-aminonicotinic acid) with the corresponding α-bromoacetophenone. Regioselective hydrogenation of 84 with Pd/C in TFA and subsequent reduction of the ester moiety yielded the corresponding alcohol. After HPLC resolution, chiral alcohol was transformed to the phosphine-imidazole ligands L90.Following the same methodology than for [Ir(cod)(L8)]BArF, ligands L90 were complexed to [IrCl(cod)]2 and mixed with NaBArF to give the corresponding iridium/BArF complexes. Ir-complexes of imidazole ligands L90 showed a similar catalytic performance to its oxazole and thiazole analogues in the asymmetric hydrogenation of olefins with poorly-coordinative groups (up to 98% ee). In addition, imidazole ligands L90 showed good results in the Ir-catalyzed hydrogenation of the demanding vinyl fluorides. It was proposed that the reason was the decreased basicity of the imidazole ring, resulting in less defluorination. The Ir/L90 catalyst (R = Ph, Ar = Ph or 3,5-Me2-C6H3) were able to reduce various vinyl fluorides with up to 86% ee, including unsaturated esters that the previously commented ligands 84 could not hydrogenate (Fig. 29 ) [213].Interestingly, the tuning of the substituent on the imidazole ring of ligand L90 led to excellent ligands for the highly regio- and enantioselective reduction of 1,4-cyclohexadienes (Scheme 58 a-d) [214]. They showed an impressive substrate scope, providing excellent results for substrates having little functionality, but also for others bearing strongly coordinating substituents and heterocycles fused rings (up to 99% ee). Finally, ligand L90d has been recently used in the asymmetric hydrogenation of allylsilanes (ee’s up to 99%; Scheme 58e) [215]. The compounds were further subjected to the Hosomi-Sakurai allylation yielding the corresponding homoallylic alcohols with three stereogenic centers in excellent diastereoselectivities.Recently, Riera, Verdaguer and co-workers presented the synthesis of a novel class of P-chirogenic aminophosphine-imidazole ligands L91–L92 [216], (Scheme 59 ) which are analogues the MaxPHOX family of ligands [43]. The new ligands were prepared from N-Boc valine, as MaxPHOX ligands, but the imidazole moiety was introduced through condensation with ortho-phenylenediamine (ligand L91) or with phenacylbromide, followed by cyclization with ammonium acetate (ligands L92). Removal of the borane protecting group with neat pyrrolidine, followed by treatment with [IrCl(cod)]2 and counterion exchange with NaBArF yielded the corresponding MaxPHOX/Ir-complexes. A catalyst precursor based on benzoimidazole ligand (L91) was found to be superior than its imidazole counterpart in the hydrogenation of the model cyclic β-enamide (89% vs 72% ee). However, it showed a lower catalytic activity and enantioselectivity than its oxazoline analogue.Phosphite-thiazoline ligands L93 [18c] (Scheme 60 ) were designed to expand the substrate scope of successful phosphite-oxazoline ligand family L23, which despite their versatility there was still room for improvement in the Pd-catalyzed allylic substitution of cyclic substrates [48b]. To this aim, the oxazoline was replaced with a thiazoline (ligands L93), in order to reduce the chiral pocket created by the ligands and make it more appropriate for cyclic substrates. Ligands L93 were synthesized from (R)-cysteine methyl ester hydrochloride in only 3 steps (Scheme 60) [18c]. Then, the thiazoline group was formed by coupling with ethyl benzimidate hydrochloride. Next, the formed thiazoline ester was reduced with MeMgBr to afford hydroxyl thiazoline in 40% ee. Semipreparative chiral HPLC was used to give access to both enantiomers of 86. Finally, the phosphite group was introduced to provide ligands L93.As predicted, phosphite–thiazoline ligands L93 improved considerably the enantioselectivities of unhindered cyclic substrates (up to 94% ee; Fig. 30 a), while oxazoline analogues L23 worked better for the rest of substrates. The introduction of a thiazoline ring was also advantageous for the Ir-catalyzed hydrogenation of olefins [217]. Ir/BArF complexes bearing thiazoline ligands L93 allowed to increase the number of substrates efficiently reduced, including Z-trisubstituted olefins, trifluoromethylated olefins and enones (Fig. 30b).Busacca and co-workers patented the phosphine-imidazolines analogues of PHOX ligands, the so called BIPI (L94) (Boehringer-Ingelheim phosphinoimidazolines) [218]. The additional nitrogen atom provides an extra tuning site in the ligand scaffold by modifying the N-substituent group. Ligands L94 were synthesized in a modular way as shown in Scheme 61 [219]. The imidazoline ring was built by condensation of o-haloimidates with the desired chiral diamine, furnishing the haloimidazolines. Then, the phosphine group was installed via the aromatic SNAr reaction with phosphide nucleophiles. Finally, reaction of the resulting phosphine-imidazolines with the corresponding alkyl halides furnished BIPI ligands L94 (Scheme 61). In the cases where R2 was an electron-rich aryl group (e.g., R2 = p-OMe-C6H4) or with electron-deficient aryl groups on the phosphine (e.g. Ar = 3,5-F2-C6H3), the N-alkyl substituent was incorporated previous to the phosphine moiety.BIPI ligands (L94) were screened in Pd-catalyzed intramolecular Heck reactions that involve the formation of a chiral quaternary centre [219,220]. It was shown that the enantioselectivity increased with the use of more electron deficient phosphines (e.g. Ar = 3,5-F2-C6H3), which allowed to achieve the highest enantioselectivities reported at that time with BINAP and PHOX for challenging substrates (ee's up to 87%; Scheme 62 ). Over the years, BIPI ligands have also been used in the Rh- and Ir-catalyzed hydrogenation of imines [221], unsaturated ureas [222], urea esters, Boc-, Cbz-enecarbamates and enamides [223], and unfunctionalized olefins [56b]. The BIPI ligand was also used in the synthesis of a cathepsin S inhibitor (see Section 6) [224].A year later of the development of BIPI ligands, Pfaltz and co-workers reported a similar ligand family called PHIM ligands (L95, Fig. 31 ). In this case, easily accessible α-hydroxyamides were used instead of diamines, so one of the alkyl substituents on the resulting imidazoline ring was exchanged by a hydrogen atom. The correpsonding Ir-complexes were prepared by using the same protocol used for [Ir(cod)(L8)]BArF complexes and tested in the hydrogenation of various unfunctionalized olefins, exhibiting in some cases higher enantioselectivities than its PHOX analogues [225]. Later, Ir-complexes containing PHIM ligands were efficiently applied in the asymmetric hydrogenation of the poorly studied vinylsilanes [226].Then, Pfaltz also patented the SimplePHIM ligands (L96), which are a simplified version of PHIM ligands (Scheme 63 ) [227]. These ligands were obtained from oxalyl chloride, which was coupled with the corresponding aminoalcohols to yield intermediate 87. After formation of the oxazoline ring using standard procedures, the phosphine group was installed to afford ligands L96. [Ir(cod)(L96)]BArF complexes were evaluated in the hydrogenation of acetophenone N-phenylimine [228] and vinylsilanes [226]. Although they provided high enantioselectivities (up to 95% and 88% ee, respectively), they didn’t showed the best catalytic performance among all ligands tested. In contrast, they turned to be very efficient in the asymmetric hydrogenation of terminal boronic esters (up to 96% ee, Scheme 63) [52b].Heterodonor P,N-ligands containing a pyridine group have turn into popular alternatives to P-oxazoline ligands because of the pyridine robustness and their straightforward synthesis. However, few P-pyridine ligands have provided outstanding results in terms of reaction scope. The success of QUINAP and PHOX ligands pushed Katsuki and co-workers to develop phosphine-pyridine ligands L97 and apply them in Pd-catalyzed asymmetric allylic substitutions [229,230]. Next, other similar bicyclic P-pyridine ligands have been synthesized and used in many asymmetric transformations (L98–L104; Scheme 64 a). The synthesis of L97 starts with chloropyrindine or chlorotetrahydroquinoline compounds, which were transformed into the corresponding chiral intermediates 88 in 5 steps, via epoxide formation and its subsequent opening [231]. Next, Suzuki cross-coupling of compounds 88 with 2-hydroxyphenylboronic acid gave the corresponding pyridylphenols, which can be then converted into the desired 2-(phosphinoaryl)pyridines in a conventional manner (Scheme 64b) [229].A ligand with a 5-membered fused ring and an isopropyl substituent (L97, n = 1, R = iPr), provided the best enantioselectivities in the Pd-catalyzed allylic alkylation of linear substrates with dimethyl malonate (up to 98% ee) [229,230]. This ligand was also used in Pd-catalyzed intramolecular allylic amination [232], Baeyer-Villiger reactions [233] and tandem allylic substitution reactions for the formation of heterocycles [234], providing moderate to good enantioselectivities.A similar ligand bearing a pinene moiety was developed by Chelucci’s and Malkov-Kočovský’s groups, simultaneously (L98, Scheme 64). This ligand could be synthesized in fewer steps than L97 since it is derived from chiral and inexpensive (−)-β-pinene [235]. Ligands L98 were used in the Pd-catalyzed allylic alkylation of 1,3-diphenylprop-2-enyl acetate with dimethyl malonate (50% ee) [235a] and in the Pd-catalyzed Heck reaction of dihydrofuran and phenyl triflate (88% ee) [235b]. The use of ent-L98 with an isopropyl substituent on the pinene moiety increased the scope and enantioselectivity provided by L97 in the Pd-catalyzed Baeyer-Villiger reaction of cyclobutanones [236] Following the same synthetic methodology, Andersson reported phosphine and phosphinite ligands L99 and L100 with the phosphorous donor atom attached to the pinene chiral motif. The cis phosphine ligand L99 gave the highest enantioselectivities (up to 97%; Fig. 32 ) in the Ir-catalyzed hydrogenation of different trisubstituted olefins (e.g. methylstilbene derivatives, α,β-unsaturated esters …) [237]. However, activities were poor even at 100 bar of hydrogen pressure. Later, the same group developed ligands L101, in which the pinene element has been removed. The resulting ligands have indeed the same backbone than oxazole, thiazole and imidazole ligands L83, L84 and L90, but with a pyridine N-donor group. The correpsonding Ir-complexes were obtained using the same protocol described for [Ir(cod)(L8)]BArF. The new Ir-catalysts showed better activities and enantioselectivities than the ones with pinene-containing ligands L99 in the hydrogenation of several trisubstituted olefins (up to 99% ee; Fig. 32) [238]. In most of the cases, complexes bearing five-membered-ring ligands (n = 1) were better catalysts than six membered-ring, which is the opposite trend observed in the case of their oxazole, thiazole and imidazole counterparts.Ligands L102 are also analogues of L98 but with the phosphine group directly attached to the pinene ring instead of being at the phenyl scaffold. [Ir(cod)(L102)]BArF were applied to the reduction of different olefins although they only proved to be successful for some 1,1′-disubstituted enol phosphonates, providing 70–90% ee (L102, R = p-MeO-C6H4; Fig. 32) [239].Later, the pinene moiety was replaced by a (+)-camphor moiety to give ligands L103 and L104. [Ir(cod)(L103)]BArF was screened in the asymmetric hydrogenation of various substrates, although it only gave a high enantioselectivity in the reduction of trans-α-methylstilbene (94% ee, Fig. 32) [240]. [Ir(cod)(L104)]BArF were also applied in the Ir-catalyzed hydrogenation of trans-α-methylstilbenes. A complex containing a ligand without any substituent on the pyridyl moiety (L104, R1 = R2 = H) provided the best enantioselectivities (up to 96% ee, Fig. 32) [241]. However, the hydrogenation of trisubstituted alkenes functionalized with alcohol, acetate and ester groups only attained moderate to good enantioselectivities (58–80% ee). Notably, the same ligand allowed the hydrogenation of methyl (Z)-2-acetamido-3-phenylacrylate in 97% ee (Fig. 32), which made it the first ligand that showed a good performance in the Ir-hydrogenation of dehydroamino acids that have been traditionally studied using Rh and Ru catalysts [242].The Pfaltz’s group has prepared several phosphine- and phosphinite-pyridine ligands (L105–L108) specially designed for the Ir-catalyzed hydrogenation of olefins. Phosphine- and phosphinite-pyridine ligands L105–L106 (Scheme 65 ) were designed to be sterically similar to PHOX [243]. Phosphine ligands L105 were prepared from cheap ethyl picolinate, which was readily alkylated by lithiated BH3-protected methyldiphenylphosphane to yield ketone (Scheme 65). This intermediate was reduced enantioselectively by (−)-chlorodiisopinocampheyl borane [(−)-Ipc2BCl]. As the resulting pyridyl-alcohol was an oil, it couldn’t be recrystallized so the hydroxyl group was protected with tBuMe2SiOTf. After recrystallization, the hydroxyl group was deprotected and protected again with the desired silyl group. Finally, the phosphine was deprotected to give ligands L105. In the case of phosphinite ligands, the synthesis was even more straightforward (Scheme 65). Pyridyl-alcohols were synthesized by ketone reduction or by alkylation of aldehydes with 2-lithiopyridine. After HPLC resolution, the introduction of the phosphinite group lead to ligands L106.An extensive ligand screening of [Ir(cod)(L105-L106)]BArF catalyst-precursors in the hydrogenation of trans-α-methylstilbene, indicated that complexes bearing phosphinites were superior to those containing phosphines (L105) (up to 97% ee vs. 88% ee). High enantioselectivities (up to 96% ee) were also afforded with other interesting olefins. Among them, it should be highlighted that [Ir(cod)(L106)]BArF (R1 = R2 = tBu) allowed the hydrogenation of the acyclic tetrasubstituted alkene shown in Scheme 66 in 81% ee and >99% conversion, which has been usually hydrogenated with low conversion and poor enantioselectivity.Bicyclic phosphinite-pyridines L107 (Scheme 67 a) were designed to study whether the more rigid conformation due to an additional ring could increased their enantioselectivities than the provided with L105–L106 [244]. Later, related disubstituted pyridine ligands with a bulky aryl group on the 2-position of the pyridyl scaffold were also developed (L108, Scheme 67b). Ligands L107 are accessible from simple, available starting materials throught the pyridyl alcohols (Scheme 67a). The key step is the oxidation of the trisubstituted pyridines, which are prepared in three steps from the corresponding ketone, to the corresponding N-oxides. N-oxides were then subjected to a Boekelheide rearrangement to yield pyridyl alcohols, which could be resolved by preparative HPLC. Recently, the same group demonstrated that pyridyl alcohols can be easily resolved via enzymatic kinetic resolution [245]. Enantiopure pyridyl alcohols were transformed towards the desired phosphinite ligands using the same methodology than for previous ligand L105–L106. The next generation of bicyclic ligands (L108) were prepared from chiral silyl-protected chloropyridine precursor (Scheme 67b) [246]. The introduction of the bulky group on the 2-pyridyl scaffold was introduced through Suzuki–Miyaura cross-coupling using commercially available boronic acids.Ligands L107 and L108 improved the catalyst performance of previous L105–L106 in the Ir-catalyzed hydrogenation of many olefins, becoming one of the few privileged ligand family for this transformation (Fig. 33 ). Generally, 2-phenyl substituted ligands with a bulky tBu and o-Tol phosphinite group provided the highest enantioselectivities [244]. Ligand L107b was initially found to be successful in the reduction of dihydronaphthalenes, allylic alcohols and α,β-conjugated esters (Fig. 33). This ligand was also successfully employed in the hydrogenation of the dihydronaphthalene core of antitumoral (+)-mutisianthol in 90% ee (see Section 6) [247]. A ligand from the L107-family (R1 = Ph, R2 = Ph) allowed the total synthesis of (+)-torrubiellone C [248] and (−)-pyridovericin [249], by catalyzing the hydrogenation of the appropriate α,β-unsaturated ester motif enantioselectively (see Section 6). (S)-L107b was also found to promote the asymmetric hydrogenation of furans with unprecedented high enantioselectivities (Fig. 33) [244]. The scope of these substrates was later extended, furnishing excellent enantioselectivities for a range of monosubstituted furans with a 3-alkyl or 3-aryl group and for benzofurans with an alkyl substituent at the 2- or 3-position [250]. Other heterocycles such as 2- and 3-substituted indoles or benzo[b]thiophene 1,1-dioxides could be also hydrogenated with enantioselectivities up to >99% ee (Fig. 33) [251]. Another type of substrates where ligand L107b proved to be successful are pinacol derived boronic esters [52b]. While the previously mentioned SimplePHIM ligands were good for terminal boronic esters, L107b was the optimal choice for trisubstituted substrates (Fig. 33), providing up to >99% ee. More recently, it has been found that ligands L107 provided higher enantioselectivities than the related PHOX family in the hydrogenation of trisubstituted vinylsilanes (ee’s up to >99% ee) [246]. Besides olefins with poorly coordinating groups, it was also found that phosphinite-pyridine ligands (S)-L107b,e are able to hydrogenate purely alkyl-substituted olefins with outstanding enantioselectivities (up to 99% ee, Fig. 33) [252]. Its hydrogenation has been used to prepare different biologically relevant products, such as α- and γ-tocopheryl acetates, precursors of main components of vitamin E (see Section 6) [252b,253]. Other relevant natural products, such as macrocidin A and long-chain polydeoxypropionaes have been synthesized through hydrogenation of long-chain molecules with the ligand family L107 [254].Later, it was found that the addition of bulkier substituents at the 2-pyridino position (e.g. R1 = 9-Anth) led to even more enantioselective ligands for the hydrogenation of dihydronaphthalene substrates (with (S)-L107d) [246]. It was also found that disubstituted ligands (S)-L108 exhibited excellent enantioselectivities in the hydrogenation of some challenging substrates. In particular, a ligand with a 9-anthracene group ((S)-L108a) showed an excellent enantioselectivity of >99% ee and 95.2:4.8 dr in the hydrogenation of (E,E)-farnesol (Fig. 33), even higher than the afforded with ligands L107 (99% ee). Furthermore, unprecedented enantioselectivities (96–99% ee) were attained in the hydrogenation of α-substituted α,β-unsaturated esters ((S)-L108b) [246,255]. β-Methyl-substituted esters (Fig. 33), which resulted to be more problematic than expected, could be hydrogenated in also high enantioselectivities (up to 98% ee) [254]. The reduction of a broad range of maleic and fumaric acid diesters was also achieved with the use of ligands (S)-L108c [256]. Finally, it has been showed once again the applicability of ligands L107-108 by the successful hydrogenation (ee's up to 98% ee and TON > 9300) of a dihydroquinoline core of agrochemical importance [257].As for other P,N-ligands, Diéguez's group replaced the phosphinite fragment of ligands L106 with a biaryl phosphite moiety, yielding the library of ligands L109–L110 (Fig. 34 ) [18a]. Note that despite the broad substrate scope of P-pyridine ligands developed by Pfaltz et al., the catalysts were suitable mainly for trisubstituted olefins. With the incorporation of a flexible biaryl phosphite moiety (L109–L110), it was possible to increase the scope of the hydrogenated substrates to 1,1′-disubstituted terminal alkenes, successfully furnishing both enantiomers in up to 99% ee. Moreover, the catalytic performance was preserved for a range of E- and Z-trisubstituted olefins (Fig. 34). The system showed high tolerance to neighboring functional groups (such as alcohols, esters, silanes …), leading as well to excellent enantioselectivities (up to >99% ee) [258].The new phosphite-pyridine ligands were also successfully applied in the Pd-allylic substitution of tri- and disubstituted allylic substrates with C-, O- and N-nucleophiles (Fig. 35 ) [18a]. The system was highly efficient in the substitution of trisubstituted substrates (up to >99% ee). A range of carbocyclic compounds were easily attained by combining in a sequential manner the Pd-allylic alkylation with Ru-catalyzed ring closing metathesis. Furthermore, the reaction could be performed in environmentally friendly solvents, concretely 1,2-propylene carbonate and ionic liquids (1-butyl-3-methyl imidazolium hexafluorophosphate and N-butyl-N-methyl pyrrolidinium bis(trifluoromethylsulfonylamide). To note is that catalyst reuse could be fulfilled up to 5 runs by using the latter ionic liquids, while maintaining the excellent enantioselectivities. Studies of the Pd-1,3-diphenyl, 1,3-dimethyl and 1,3-cyclohexenyl allyl intermediates by NMR spectroscopy showed that in general, for enantioselectivities to be high the ligand parameters need to be correctly combined so that the isomer that reacts faster with the nucleophile is predominantly formed [18a].Qu and co-workers prepared a series of air-stable P-chiral pyridyl-dihydrobenzooxaphosphole ligands, which were called BoQPhos (L111). These ligands could be readily obtained by a diastereoselective nucleophilic aromatic substitution of sulfonyl pyridines with P-stereogenic intermediates (Fig. 36 ) [259]. Sulfonyl pyridines were obtained from 2-chloropyridine or 2,6-dichloropyridine, while the P-stereogenic fragment was prepared in 8 steps from methyldichlorophosphine [260].[Ir(cod)(L111)]BArF complexes were tested in the hydrogenation of some tri- and tetrasubstituted unfunctionalized alkenes. Although the ee 's obtained were moderately good (76–90% ee), it should be highlighted that it was possible to hydrogenate the more challenging tetrasubstituted indene and dihydronaphthalene in high conversions (90–>99%) and promising enantioselectivities (76–80% ee) with [Ir(cod)(L11a)]BArF (Scheme 68 ) [259]. More recently, it has been showed the utility of methoxy-substituted L111b for the Ir-catalyzed hydrogenation of pyridinium salts [261]. Ir/L11b allowed the preparation of piperidines bearing 2-alkyl and 2-aryl substituents of different nature, with enantioselectivities up to 86% and 98% respectively (Scheme 68).Finally, Fan and co-workers have accounted the most recent family of P-pyridine ligands, consisting in aminophosphine-pyridine ligands L112 (Scheme 69 ). These ligands contain 2-(pyridin-2-yl)-substituted 1,2,3,4-tetrahydroquinoline backbone that were achieved via Ru-catalyzed asymmetric hydrogenation of 2-(pyridin-2-yl)quinolines (Scheme 69a) [262]. Ligands L112 were screened in the Ir-catalyzed hydrogenation of the model substrates trans-β-methylcinnamate and E-methylstilbene, which were reduced in >99% ee. More important are the high enantioselectivities obtained in the reduction of challenging 7-membered ring imines, concretely benzazepines and benzodiazepines. A ligand with a 2-isopropyl-pyridine moiety (L112a) proved to be the most active and enantioselective ligand, showing good diastereomeric ratios and ee’s up to 99% for both types of substrates (Scheme 69b). It should be noted that the dihydrogenation of benzazepines was performed as a two-step one-pot process, otherwise only partially reduction was observed, being the CC bond the most reactive one.Chiral P,O-ligands have traditionally played a less important role than P,N-ligands in enantioselective catalysis. The hemilability of the P,O-ligands, owing to the occurrence of both a hard (O) and a soft (P) bases on the same metal center, facilitates several transformations at the metal center, such as oxidative addition, ligand exchange, isomerization, etc., that often has a positive effect on catalytic activity. However, this hemilability, can at the same time cause a detrimental effect on enantioselectivity, since the ligand can be coordinated in a monodentate fashion in the enantiodiscriminating transition state. MOP ligands constitute one of the early examples in which a P,O-ligand acts as monodentate ligand (Fig. 37 ) [263,264]. Despite this, MOP-ligands have been used with great effectiveness in various asymmetric reactions [263]. In all these cases, the ligand acts as a chiral monophosphine with the ether group that produces a secondary interaction with the incoming nucleophile/reagent. Such secondary interaction, as early demonstrated by Hayashi and coworkers, is key to maximize enantioselectivities.Heterodonor phosphine-phosphine oxide ligands have been applied with great effectiveness in different asymmetric transformations (Fig. 38 ) [265]. These ligands can be prepared via Pd-catalyzed monooxidation of the corresponding bidentate phosphines [266].In this respect, Faller and coworkers found that BINAP(O) ligand provided high enantioselectivities in the Ru-catalyzed Diels-Alder reactions (Scheme 70 a) [267]. More recently, the groups of Oestreich [268] and Hou [269] independently published the application of BINAP(O) in the Pd-catalyzed asymmetric intermolecular Heck reaction (Scheme 70b). They found an important change in the regio- and enantioselectivity of the arylation of cyclic alkenes when BINAP(O) ligand was used instead of BINAP. In this respect, the arylation of 2,3-dihydrofuran behaves as perfectly regiodivergent; while the use of BINAP favors the formation of the thermodynamically more stable 2-aryl-2,3-dihydrofuran, the use of BINAP(O) led to the preferential formation of 2-aryl-2,5-dihydrofuran. In addition, little alkene migration was observed with BINAP(O) [268]. The effectiveness of Pd-BINAP(O) in asymmetric Heck reaction was also demonstrated in the efficient kinetic resolution of 2-substituted-dihydrofurans providing optically enriched trans-2,5-disubstituted-dihydrofurans and 2-substituted dihydrofurans in high yield and ee’s (S factor of up to 70; Scheme 70b) [269]. Zhou’s group also reported the application of BINAP(O) in the preparation of chiral fused carbo- and heterocycles through a domino reaction involving an asymmetric intermolecular Heck reaction followed by a diastereoselective cyclization (Scheme 70c) [270].Another relevant example of heterodonor phosphine-phosphine oxide ligands can be found in the work of Charette’s group that reported the application of BozPHOS, a monoxide version of the Me-Duphos (Fig. 38), in the Cu-catalyzed 1,2-addition of diorganozinc reagents to N-phosphinoylarylaldimines (ee’s up to 99%; Scheme 71 ) [271]. However, its application was in part hampered by the accessibility and stability of the N-phosphinoylalkylaldimines. To solve this, they demonstrated that the reaction also worked well using the sulfinic adduct of N-phosphinoylimines (Scheme 71) [272].The latest heterodonor phosphine-phosphine oxide design with an outstanding applicability can be found in the work of Zhou's group, that reported the successfully application of spirocyclic SDP(O) (Fig. 38) in Pd-catalyzed intermolecular Heck reactions (Scheme 72 ). Useful, SDP(O) ligands not only allowed the arylation of standard substrates such as 2,3-dihydrofuran or cyclopentene [273], but also of 5 substituted-2,3-dihydrofurans [274] that led to the preparation of a chiral quaternary carbon center (Scheme 72a). More interestingly, the use of SDP(O) ligands was further extended to the desymmetrization of 4-substituted-cyclopent-1-enes and other bicyclic olefins via asymmetric Heck reaction and hydroarylation, respectively (Scheme 72b) [275]. The use of spirocyclic SDP(O) ligands also allowed the unique asymmetric intermolecular Heck reaction with aryl halides as coupling partners (Scheme 72c) [276]. A wide range of aryl bromides and chlorides, including examples with heteroaromatic groups, were efficiently introduced in the Heck reaction of various cyclic olefins (35 examples with ee’s typically >95%).Hemilabile amido-phosphine ligands have also shown their effectiveness in asymmetric catalysis. Tomioka and coworkers early demonstrated that the effectiveness of this type of ligands was mainly consequence of the coordination of the amide carbonyl oxygen to the metal [277]. Among all the amido-phosphine ligands, we should highlight ligands L113–L119 (Fig. 39 ).The synthesis of such ligands turned to be quite straightforward (Scheme 73 ). Thus, proline-based ligands were easily attained from (S)-tert-butyl-2-(bromomethyl)pyrrolidine-1-carboxylate through a nucleophilic substitution with a range of metallated phosphines (Scheme 73a). Elimination of the N-Boc protecting group, led the corresponding free amine, which reacted with the desired acetyl chlorides, isocyanates or carbamoyl chlorides. Oxygen-sensitive phosphine moieties, such as di-tert-butyl- or dicyclohexylphosphines, should be protected as borane adducts to avoid oxidation through the ligand synthesis. The synthesis of non-biaryl atropoisomeric ligands L118 and L119 starts from the corresponding aromatic tertiary amide, which is transformed to the corresponding amido-phosphine compound by lithiation followed by reaction with ClPPh2 (Scheme 73b). For ligand L118, the racemic amido-phosphine was chemically resolved by using (–)-camphanic chloride, which allowed the separation by flash chromatography of the corresponding diastereoisomers. Alkaline saponification, followed by etherification gave ligand L118. For ligand L119, the 2-(diphenylphosphaneyl)-N,N-diisopropylbenzamide was then formylated to give the corresponding aldehyde, which allowed the introduction of the chiral N-tert-butanesulfinyl imine, by condensation with (R)-tert-butanesulfinamide and subsequent 1,2-addition of PhMgBr. The diastereomeric amido-phosphines were then separated by flash chromatography.The simple proline-based ligand L113, prepared by Tomioka and coworkers was applied with success in several asymmetric transformation: the Rh-catalyzed 1,4-addition of arylboronic acids to cycloalkenones (ee’s up to 97%) [277], the Cu-catalyzed conjugate addition of dialkylzinc reagents to nitroalkenes (ee’s up to 80%) [278] and in the highly regio- and enantioselective (regio’s up to >99% and ee’s up to 91%) allylic substitution of Grignard reagents to cinnamyl-type allylic bromides [279] (Scheme 74 ).Later, they also studied the introduction of an extra stereogenic center at the N-Boc amido group of ligand L113 (ligands L114). The use of N-Boc-L-valine-connected amido-phosphine ligands L114 allowed to extended the applicability of ligand L113 to the Rh-catalyzed arylation of N-tosyl- and N-phosphinoyl aldimines (ee’s up to 99%; Scheme 75 ) [280] and in the Cu-catalyzed conjugate addition of dialkylzincs to β-aryl-α,β-unsaturated sulfonylaldimines [281] (ee’s up to 91%; Scheme 75).Later, they further studied other peptidic modifications of ligand L113 and found that ligand L115, involving a small D-Phe-D-Val dipeptide, was useful in the asymmetric conjugate addition of organozinc reagents to cycloalkenones (ee’s up to 98%; Scheme 76 ) [282]. Advantageously, the Cu/L115 catalyst was also applied in the kinetic resolution of (rac)-5-substituted cycloalkenones to yield trans-3,5-disubstituted alkanones with excellent ee’s (up to 90%) and excellent trans/cis ratios (up to >98/2 dr; Scheme 76) [283]. Then, they also reported the useful of Cu/L115 system in the conjugate addition of 6-substituted cyclohexenones to give disubstituted cyclohexanones, albeit with marginally selectivity (cis/trans ratio close to 1). Epimerization of these mixtures with DBU (1,8-diazabicyclo[5.4.0]undec-7-ene) favored the formation of the most stable trans-cyclohexenones as the major product in high yields and ee’s (up to 96%; Scheme 76) [284].Later, the Pfaltz’s group further modified Tomioka’s proline-based ligand L113 to include several dialkyl and dialkyl phosphino groups as well as urea groups and bulky amide at the pyrrolidine N-atom (ligands L116; Fig. 39) [285,286]. Ligands L116 containing bulky phosphine moieties (R1 = Cy or tBu) shown its effectiveness in the Ir-hydrogenation of trans-methylstilbene olefins and olefins with poorly coordinative groups (e.g. α,β-unsaturated ketones and carboxylic esters; Fig. 40 a). These results were pretty unexpected having in mind the lability of the Ir-O bond, and unambiguously demonstrated that ligands L116 remain coordinated in a bidentate fashion during the catalytic cycle, most likely due to the highly acidic character of the IrIII/IrV intermediates.More recently, related pyrrolidine-based P,O ligands derived from cheap carbohydrates (D-ribose, D-mannose and D-arabinose; ligands L117) were also screened in the Ir-catalyzed hydrogenation of olefins [287]. The presence of a rigid bicyclic skeleton in ligands L117 had a positive effect on enantioselectivity, enabling to successful hydrogenate also 1,1′-disubstituted allylic acetates (Fig. 40b). As a further advantage and in contrast to L116, ligand L117 does not require the presence of less stable, bulkier phosphine substituents for optimal performance.Another relevant class of amido-phosphine ligands are the non-biaryl atropoisomeric ligands L118 and L119 (Fig. 39). Ligand L118 provided an enantioselectivity of 95% in the Pd-catalyzed allylic substitution reaction of 1,3-diphenylallyl acetate with dimethyl malonate as nucleophile [288]. The Pd/L118 catalyst was applied in the asymmetric Heck reaction, albeit with low ee’s (up to 55%) [289]. More recently, Cia and Xu groups applied the amidophosphine ligand L119, which proved to be highly efficient in Ag-catalyzed [3 + 2] cycloaddition reactions. Catalyst Ag/L119 mediated the [3 + 2] cycloaddition of aldiminoesters with nitroalkenes to yield optically enriched nitrosubstituted pyrrolidines (dr’s up to >99:1 and ee’s up to 99%; Scheme 77 ) [290]. The same catalyst was also used in the preparation of imide-containing pyrrolidines by reaction of iminoesters with maleimides (dr’s up to >98:2 and ee’s up to 99%; Scheme 77) [291].Sulfinamido-phosphines L120–L123 are another type of heterodonor P,O ligands that has recently shown its applicability in asymmetric catalysis (Fig. 41 ) [292,293]. Their synthesis is again fairly straightforward, as can be seen in the short synthesis of ligands L120 (Scheme 78 ). Thus, condensation of (R)-tert-butanesulfinamide with the corresponding 2-phosphinobenzaldehyde led to imino-phosphine intermediates. Stereoselective addition of R2Li followed by reaction with R3Cl (for R3 ≠ H) yielded L120 ligands.Ligands L120 (R1 = Ph, Cy, Ad; R2 = Ar, Me, tBu; R3 = H, Me, CH2-9-Anth) have proved to be useful in several asymmetric transformation, such as the Cu-catalyzed [3 + 2] cycloaddition of azomethine ylides with a range of β-trifluoromethyl β,β-disubstituted enones and α-trifluoromethyl α,β-unsaturated esters [292a,b], the Pd-catalyzed Suzuki reaction for the preparation of axially chiral phosphonates and phosphine oxides [292c], the Pd-catalyzed intramolecular Heck reaction of allyl aryl ethers [292d] and the Pd-catalyzed intermolecular Heck reaction of alkynes [292e] (Scheme 79 ).Ligands L121–L123 have also shown its usefulness but due to their latest development the range of reactions where they have been applied is still limited. Thus, the Xantphos-inspired ligand L121 has been applied with success in the arylation of sulfenate anions (ee’s up 99%) [293a], and ligands L122 and L123 in the boroacylation of 1,1′-disubstituted allenes [293b] and 1,3-dipolar cycloadditions [293c-d], respectively.The application of anionic P,O ligands is less developed than its neutral analogous despite of the early successful application of phosphine-carboxylate ligands L124 and L125 (Scheme 80 a) [294]. These ligands attained high enantioselectivities in the Pd-catalyzed allylic substitution reactions (ee’s up to >99%; Scheme 80a). More recently, phosphine-sulfonate ligands L126 allowed the asymmetric copolymerization of polar vinyl monomers with carbon monoxide to yield highly head-to-tail isotactic γ-polyketone polymers (Scheme 80b) [295]. In 2017, Zhou’s group accounted a series phosphine-carboxylate ligands (SpiroCAP) related to phosphine-oxazoline ligands L32, which demostrated to be efficient in the hydrogenation of terminal unsaturated carboxylic acids (Scheme 80c) [79].P-thioether ligands have a strong preponderancy among the P,S-ligands [3f,g,296]. In the early 90 s, researchers realized that P–thioether compounds could be of great use in asymmetric catalysis with the growth of a quite important number of P-thioether ligands. Even so, they found that the presence of diastereoisomeric mixtures of catalytically active species that arise from the sulfur coordination, which becomes an stereogenic center after coordination, made difficult to achieve high enantioselectivities. Thus, very few of the early developed P-thioether ligands had found an important impact in asymmetric catalysis. In the last decade, many efforts have been made to understand how to control the sulfur coordination. For this, mechanistic investigations had played a relevant role and they have revealed that sulfur coordination can be controlled, which have led to a new push on the use of P-thioether ligands in asymmetric catalysis. This section focuses on these new P-S ligands and the correlation between their ligand architecture and catalytic results.BINAP and ferrocene-based ligands have shown they widely prominent useful in asymmetric catalysis. The development of their heterodonor versions was therefore an expected step. In 1994, Gladiali and coworkers used BINAP-based phosphine-thioether ligands L127 (R = Me and i Pr; Fig. 42 ) in hydroformylation and transfer hydrogenation. Albeit its moderate success, this pioneering work spread the way for the use of P-S ligands in asymmetric catalysis [297]. Then, other groups reported the application of ligands L127 (R = Me, Ph, 2-iPr-C6H4, 2-Naph, 3,5-Xyl and Cy) in the Pd-catalyzed allylic alkylation using as nucleophiles the model dimethyl malonate and the less studied indoles (R = 2-iPr-C6H4; Fig. 42) with more success (96% ee), although the high ee's are limited to the benchmark linear substrate rac-1,3-diphenylallyl acetate [298]. More recently, a chiral biphenyl version of ligands L127 has been developed, providing similar high ee’s in the Pd-catalyzed allylic substitution with indoles [299].In contrast to binaphtyl- and biphenyl-based P-thioether ligands, since the early development of ferrocene-based P-thioether ligands in 1996 by the groups of Pregosin [300] and Togni [301], a broad range of this type of P-thioethers have been developed [3f]. Among them, it should be underlined: the phosphine-thioether Fesulphos ligands L128, the N–phosphine-thioether FerroNPS-type ligands L129–L130, and the phosphine-thioether ThioClick-Ferrophos L131 (Fig. 43 ). The synthesis of these ligands can be achieved in few steps (Scheme 81 ). Fesulphos ligands L128 were prepared from R-tert-butylsulfenylferrocene in two steps. The first step involves the diastereoselective introduction of the desired phosphine moiety via ortho-lithiation. The second consists in the reduction of the sulfoxide to the thioether group with HSiCl3·NEt3. Ligands L129–L130 were achieved from Ugi’s amine, which facilitates the diastereoselective introduction of the thioether group. The tertiary amine was then transformed to the secondary amine or to the benzimidazole. Finally, treatment with ClPPh2 led to ligands L129–L130. ThioClick-Ferrophos ligand L131 was synthesized following an analogous method starting from (S,R p)-ortho-bromo-(1-azidoethyl)ferrocene.Fesulphos ligands L128 (R = Ph, 4-FPh, 4-CF3Ph, 2-furyl, Cy, o-Tol, 1-Naph; Fig. 43), developed by Carretero's group, have become one of the most useful P–S families in asymmetric catalysis (Scheme 82 ). They have been used with great success in various CC bond forming reactions, in combination with both Pd and Cu catalyst precursors [302]. Thus, in their first report they demonstrated the usefulness of Pd/L128 catalyst in the Pd-catalyzed allylic substitution of the benchmark substrate rac-1,3-diphenylallyl acetate using several malonates and amines as nucleophiles (ee’s up to 98%; Scheme 82) [302a,b]. Pd/L128 also demonstrated to be efficient in the asymmetric ring opening of heterobicyclic alkenes with diorganozinc reagents (ee’s up to >99%; Scheme 82) [302a,c,f]. Since then, Carretero and coworkers have also demonstrated the usefulness of Cu/L128 catalyst precursors in a range of other metal mediatedreactions, such as Mannich-type reactions of N-sulfonylimines with several electrophiles [302h,l], (aza)-Diels-Alder reactions of electron rich alkenes with aldimines [302d,g] and 1,3-cycloaddition reactions of azomethine ylides with an extensive variety of activated olefins [302e,i,k,m,o-r] (ee’s up to >99%; Scheme 82).Later, Chan’s group developed ligands L129 [303] and L130 [304] (Fig. 43), which combines both planar and central chirality, and demonstrated their versatility in Pd-catalyzed allylic substitutions. Thus, Pd/L129 was successfully applied in the allylic substitution of benchmark rac-1,3-diphenylallyl acetate with dimethyl malonate, a range of amines and several aliphatic alcohols (ee’s up to 98%; Scheme 83 ) [303a,b], being one of the first successful examples of allylic etherifications with non-aromatic alcohols. More recently, the use of ligand L130 containing a benzimidazole unit, extended the nucleophile scope to indoles (ee’s up to 96%; Scheme 83) [304a] and to the alkylation of some cyclic substrates with ee’s up to 87% [304b].ThioClick-ferrophos ligand L131 (Fig. 43), developed by Fukuzawa’s group, was screened in the Ag-catalyzed Mannich reactions of N-tosylimines with a glycine Schiff base [305a] with moderate diastereoselectivities and high enantioselectivities (dr’s up to 7:3 and ee’s up to 98%) (Scheme 84 ). Ag/L131 also efficiently catalyzed the 1,3-cycloadditions of azomethine ylides using a variety of activated alkenes providing similar enantioselectivities than those achieved with the Cu/Fesulphos L128 catalyst (ee’s up to 99%) [305b-g]. Later, the same group also found that Ag/L131 catalyzed the conjugate addition of several types of Michael acceptors to different imino esters, oxazoline-esters and related substrates (ee’s up to >99%; Scheme 84) [305i-l]. Finally, by switching the positions of the thioether and the phosphine moieties high enantioselectivities can also be reached in Pd-catalyzed allylic substitution of rac-1,3-diphenylallyl acetate with dimethyl malonate, benzylamine and some benzylic alcohols (ee’s up to 90%) [305m].An additional group of powerful P–S ligands are those having the two donor functionalities linked by two carbon atoms [296,306]. One of the first successful examples of such a group of ligands can be found with phosphinite-thioether ligands L132, with a very simple ligand backbone (Scheme 85 ) [307] These ligands are prepared from the corresponding chiral epoxide via epoxide ring opening with the corresponding thiol or from the corresponding Evan’s N-acyl carboximide in few steps to provide the corresponding thioether-alcohol. Treatment of the later compounds with the desired chlorophosphine gives access to ligands L132 (Scheme 85).The Evans' group demonstrated that by optimizing the different ligand parameters (thioether and phosphinite substituents as well as the ligand backbone) it is possible to control the thioether coordination to the metal. As a result, ligands L132 constitutes one of the early-developed P–thioether ligands that provided excellent results to several asymmetric reaction, such as the Rh-catalyzed hydrosilylation of ketones and hydrogenation of alkenes [307c] and the Pd-catalyzed allylic substitutions [307a,b] (Scheme 86 ). However, the substrate/reagents scope was still low.Much later our group decided to replace the phosphinite moiety in the Evans' ligands L132 by the benefits of biaryl phosphite groups (ligand L133; R1 = tBu, 2,6-Me2Ph; Fig. 44 a). Air stable ligands L133 were fruitfully applied in the hydrogenation of unfunctionalized alkenes, including terminal olefins and olefins with poorly coordinative groups (ee’s up to 99%; Fig. 44b) [308]. The catalysis were carried out using the preformed catalyst precursors [Ir(cod)(L133)]BArF, prepared using the same methodology than for [Ir(cod)(L8)]BArF. One single isomer was found except for complexes containing ligands with a flexible biphenyl moiety, due to the tropoisomerization of these units. The X-Ray structure confirms the coordination of the ligand throught the P and S atoms to the metal, whith a twist-boat conformation of the chelate ring and a pseudoaxial disposition of the thioether moiety. It should be highlighted that the use of Ir/L133 allowed the pioneering reduction of 1,1′-disubstituted aryl-substituted boronic esters. Interestingly, the enantioselectivity is mainly determined by the thioether substituent and the configuration of the phosphite group. The replacement of the phosphite moiety by several phosphinite groups provided lower ee’s. The study of the reaction intermediates by HPNMR (high pressure NMR) spectroscopy and DFT calculations allowed to found the active Ir-dihydride alkene species, which follows the classical Halpern-mechanism, in which the minor species are the most active ones [309]. In addition such mechanistic study provided helpful insights to understand the influence of the different ligand elements on enantioselectivity.Other relevant families of P–S ligands having a two carbon atoms linker are the families of P-thioethers ligands L134 and L135 and the phosphoroamidite-thioether ligands L136–L138 (Fig. 45 ). They resemble very much to the Evans' ligands in the simplicity of the ligand backbone, which aids the recognition of key intermediates by NMR as well as speeds up the ligand optimization by DFT calculations. Ligands were prepared from the corresponding epoxides (ligands L134 and L135) or aziridine (compounds L136–L138) following the same synthetic strategy as for the synthesis of Evan’s ligands.The arylglycidol-based phosphinite-thioether ligands L134 (R1 = Ar, tBu, Ad, Cy; R2 = Ph, Tol, Cy, Mes, R3 = Me, Tr, Bn) (Fig. 45) have found to be useful in allylic substitutions and hydrogenation reactions [310]. A practical advantage offered by L134 is the fact that they are made in three steps from accessible arylglycidols [310a]. In addition, both enantiomers of these P,S-ligands can be reached by simple selection of the tartrate ester used in the Sharpless epoxidation leading to the arylglycidol. Pd/L134 catalytic systems provided similar high enantioselectivities than Evans' ligands L132, working under milder reactions conditions, room temperature and shorter reactions times, in the allylic substitution of di- and trisubstituted linear allylic acetates with a range of malonate-type nucleophiles and amines, and also extended the nucleophile scope to the less studied aliphatic alcohols [310a]. Ligands L134 were also used in the Rh-catalyzed hydrogenation of dehydroamino acids, albeit with lower success than related ligands L132 [310b]. Much more notable are the results reached in the Ir-catalyzed hydrogenation of unfunctionalized olefins [310c], in spite of the known difficulties of enantiocontrol associated to substrates lacking metal-coordinating functionalities. [Ir(cod)(L134)]BArF complexes were therefore able to reduce a large number of olefins, with similar ee’s than the best ones attained with Ir–P,N catalysts (43 examples,  ee 's up to 99%; Fig. 46 ). Unlike ligands L133 with a cyclohexane-based backbone, the use of phosphite analogues led to lower enantioselectivities than with phosphinite-thioether ligands L134. The crystal structures of these Ir-catalyst precursors showed that while ligands with a phosphite moiety had the thioether group in equatorial, in the related phosphinites the thioether was in axial. This contrasts with the pseudoaxial arrangement of the thioether substituent in Ir-structures with cyclohexane-based phosphite–thioether commented above, that also form a six-membered chelate ring. This behaviour seems to show that the disposition of the thioether substituent (in this case, axial disposition) is important to obtain high enantioselectivity. DFT calculations indicated that the reaction proceeds via an Ir(III)/Ir(V) catalytic system in which the enantioselectivity-determining step is the migration of a hydride to the coordinated alkene. In addition, the analysis of the transition states allowed to develop a quadrant model system that facilitates rationalization of the catalytic results. These DFT studies were also crucial to guide the ligand optimization process towards high enantioselectivities. They indicated the need of ligands with a mesityl group at the carbon next to the thioether group (Ar = 2,4,6-Me3-C6H2) and a bulky aromatic thioether groups (2,6-dimethylphenyl or 1-naphthyl moieties, depending on the substrate). The application of mesityl-containing ligands L134 are therefore crucial to achieve the highest ee’s for a range of olefins including examples containing poorly coordinative groups and terminal alkenes (ee’s up to >99%; Fig. 46). Remarkably, the catalytic systems could be also recycle up to 3 times with 1,2-propylene carbonate.Recently, phosphite–thioether ligands L135 (R1 = iPr, nPr, tBu, Ph, 2,6-Me2Ph, 4-CF3Ph, 4-MeOPh, 9-Anth) (Fig. 45), prepared in three steps from indene, were designed to maximize the substrate range in Pd-catalyzed allylic substitution reactions [18d]. The simple indene backbone facilitated both DFT and NMR studies of Pd-allyl key intermediates, which were used to optimize the thioether and phosphite substituents in the search of the best catalyst. As a result, catalyst Pd/L135b (R1 = 9-Anth) is one of the very few catalysts able to afford excellent enantioselectivities (typically >95% ee) for a large number of unhindered and hindered allylic acetates with an array of C, N and O nucleophiles (Scheme 87 ; 40 compounds in total). Notably, the excellent performance of L135b was maintained using 1,2-propylene carbonate as solvent. Mechanistic investigations provided an elucidation into the exceptionally rare wide substrate scope. Enantioselectivity is therefore controlled by the relative stability of the Pd-η3-allyl intermediates and the electrophilicity of the allylic terminal carbons. More concretely, Pd/L135b catalytic system not only favors the preferential formation of one of the possible Pd-allyl intermediates, but also speeds up the nucleophilic addition at the terminal allylic carbon atom trans to the phosphite moiety of most stable Pd-allyl intermediate. In addition, the authors took advantage of the great diversity of the allylic substitution products arising from the introduction of malonates having allyl and propargylic groups for the preparation of chiral functionalized carbo- and heterocycles as well as polycarbocyles. The former compounds were prepared by means of ring-closing metathesis, while the latter were prepared via Pauson-Khand reaction (Scheme 88 ).Phosphoroamidite-thioether ligands L136–L138 (Fig. 45), easily prepared in three steps from (2S,3S)-2,3-diphenylaziridine, have also been effectively applied in many asymmetric transformations [311]. Thus, ligand L136 been fruitfully applied in the Pd-catalyzed allylic substitution of 1,3-diarylallyl acetates with a collection of indoles and hydrazones (ee’s up to 98; Scheme 89 a) [311a,b]. Similar high ee’s were also achieved in the allylic substitution of rac-1,3-diphenylallyl acetate with benzyl amine and benzyl alcohol [311a]. Ligand L136 also evidenced to be highly competent in both Cu- and Pd-catalyzed cycloaddition reactions. Thus, for instance, catalyst Cu/L136 afforded a range of polysubstituted endo pyrroles in high diastereo- and enantioselectivities via 1,3-cycloaddition of azomethine ylides and nitroalkenes (Scheme 89b) [311c]. Interestingly, the use of related H8-Binol-derived ligand L137 (Fig. 45) led to the formation of the exo pyrroles (Scheme 89b) [311c. Catalyst Pd/ L136 was successfully used in inverse-electron demand decarboxylative [4 + 2] cycloaddition reactions. Thus, highly functionalized dihydroquinol-2-ones were produced with excellent selectivities (d.r. > 20:1 and ee’s up to 95%; Scheme 89c) [311d]. Pd/L136 has recently found to be beneficial in the visible-light-driven [5 + 2] cycloaddition of vinylcyclopropanes with α-diazoketones. This new methodology provides facile access to highly functionalized 7-membered ring lactones (d.r.’s up to 16:1 and ee’s up to 96%; Scheme 89d) [311e]. Similarly, a series of quinolinones were synthesized via Pd-catalyzed light-driven decarboxylate [4 + 2] cycloaddition of tosylated vinyl carbamates with in situ generated ketenes (ee’s up to 96% ee; Scheme 89e) [311g]. It should be mentioned that some Pd-catalyzed decarboxylative cycloaddition reactions do not requires the presence of a chiral biaryl phosphoroamidite moiety (ligand L138; Fig. 45). Thus, a range of tetrahydroquinolines bearing three contiguous stereocenters were efficiently prepared using Pd/L138 by reaction of benzoxazinanones with activated alkenes (d.r.’s typically >95:5 and ee’s up to 98% ee; Scheme 89f) [311f].Carbohydrates have also been used as platforms for preparing P-thioether ligands. The use of carbohydrates is advantageous since they are cheap and readily available. Moreover, they have a well-established chemistry and they are highly functionalized, which favor the synthesis of highly modular ligand libraries and enable an easy ligand optimization for each particular substrate and reaction [312]. Khiar’s group were the first to apply carbohydrate P–thioether ligands in catalysis. They used pyranoside ligands L139 and L140 (Fig. 47 ) in the Rh-catalyzed hydrogenation of some enamides (ee's up to 98%) and in the Pd-catalyzed allylic substitution of benchmark 1,3-diphenylallyl acetate (ee's up to 96%) [313]. The use of pseudo-enantiomeric ligands L139 and L140 allowed the preparation of both isomers of the products, without having to prepare the enantiomeric ligands from the expensive L-sugar series.Furanoside phosphite-thioether ligands L141 and L142 (R = Ph, Me, iPr, tBu, 4-MePh, 4-CF3Ph, 2,6-Me2Ph) (Fig. 47) were prepared from D-xylose in multigram scale. Treatment of D-xylose with I2 in acetone followed by deprotection of the more reactive isopropylidene group led to 1,2-O-isopropylidene-α-D-xylofuranose (key for the synthesis of ligands L141), which was easily transformed to the ribofuranoside anologue (key for the synthesis of L142). From both xylo- and ribofuranoside diols, ligands L141 and L142 were prepared by introducing the thioether group at the primary alcohol, via an SN2 reaction, followed by treatment with the desired phosphorochloridite (Scheme 90 ).Phosphite-thioether ligands L141 and L142 (Fig. 47) represented the first use of P,S-ligands in the asymmetric hydrogenation of unfunctionalized olefins or with poorly coordinative groups [314]. The catalysis were carried out using the preformed catalyst precursors [Ir(cod)(L141–L142)]BArF, prepared using the same methodology than for [Ir(cod)(L8)]BArF. The X-Ray analysis indicated that in contrast to previously commented Ir/P-S complexes, such as [Ir(cod)(L133)]BArF, the thioether substituent adopts an equatorial disposition. In this context, the use of ribofuranoside ligand L142k (R = 2,6-Me2-C6H3) provided high ee 's in the hydrogenation of methyl stilbene-type olefins, Z-trisubstituted olefins and triarylsubstituted olefins (Fig. 48 ). The latter provides a feasible entry point to valuable compounds containing diarylmethine chiral centers. Ir/L142k catalytic system also led to high ee 's in the reduction of many 1,1′-disubstituted olefins. In addition, both enantiomers of the reduced products can be easily attained by changing the configuration of the biaryl phosphite moiety. Another interesting feature of this ligand is that the furanoside scaffold allowed to restrict efficiently the tropoisomerization of conformationally flexible biphenyl phosphite moieties. For most of the substrates studied, similar high enantioselectivities have therefore been reached with the cheap achiral bulky biphenyl phosphite moiety (ligand L142b). Again, these catalysts work well in 1,2-propylene carbonate, helping the catalysts to be recycled several times. Finally, the use of phosphinite or phosphine analogues led to lower enantioselectivities.When ligands L141 and L142 were tested in allylic substitution of linear hindered 1,3-disubstituted allylic acetates with a range of C-nucleophiles (e.g. α-substituted malonates, diketones, cyano esters …) and a selection of O- and N-nucleophiles, high enantioselectivities were attained using Pd/L142k (R = 2,6-Me2-C6H3) catalyst (ee’s up to >99%; Scheme 91 ) [18b,315]. To achieve high enantioselectivities for more demanding cyclic and unhindered linear substrates, the use of xylofuranoside ligand L141h (R = 1-Naph) was required (ee’s up to >99%; Scheme 91). This feature was rationalized with the aid of NMR studies of the Pd-allyl intermediates and DFT calculations of the TSs using cyclohex-2-en-1-yl acetate as model substrate. These studies demonstrated how the size of the chiral pocket in the catalytic species is affected by the configuration at C-3 of the furanoside backbone. Thus, by using catalyst Pd/L141h, only one of the two possible syn/syn diastereomer Pd-1,3-cyclohexenyl-allyl intermediate is predominantly formed (dr’s > 20:1) [315].Finally, Taddol-type phosphite-thioether ligands were made from L-tartaric acid (ligands L143) and D-mannitol (ligands L144) and screened in several asymmetric transformations (Fig. 47) [316]. Several positions of the ligands (R1 = Me, 1-Ad, Ph, tBu, 2,6-Me2Ph, 1-Naph, 2-Naph; R2 = Me, H, Ph; R3 = Me, H, Ph) can be easily varied through highly efficient methods. This methods also allowed to generate at will new stereogenic centers next to the donor functionalities (R4 = H, Me, CH2OTBDMS, CH2OTBDPS, CH2OTIPS, CH2OTr; R5 = H, Me). [Ir(cod)(L144f)]BArF (R1 = 1-Naph, R4 = (R)-CH2OTBDMS and R5 = H) provided ee’s up to 95% in the hydrogenation of trans-methylstilbene-type substrates, β,β’-disubstituted unsaturated esters, α,β-disubstituted enones, lactones and lactams bearing an exocylic double bond [316b]. Note that for most of these substrates the selenoether version of the ligands attained slightly higher enantioselectivities than the thioether analogues [316b]. The use of catalyst [Ir(cod)(L143h)]BArF (R1 = 1-Naph; R2 = R3 = H) was necessary to maximize ee’s in the reduction of terminal olefins (ee’s up to 99%) [316b]. Interestingly, by using [Ir(cod)(L144g)]BArF (R1 = 1-Naph, R4 = (S)-CH2OTBDMS and R5 = H) enantioselectivities up to 99% were attained in the hydrogenation of cyclic β-enamides (Scheme 92 a). Interestingly, the exchange of the metal from Ir to Rh led to the preferential formation of the opposite enantiomer (Scheme 92a) [316a]. This is one of the rare examples of enantioswitchable metal-catalyzed transformation. This allows, for instance, access to both enantiomers of the precursors for the synthesis of rotigotine (used in the treatment of Parkinson's disease) [41a] and alnespirone (a selective 5-HT1A receptor full agonist) [41d]. In addition, the use of [Rh(cod)(L144f)]BF4 (R1 = 1-Naph, R4 = (R)-CH2OTBDMS and R5 = H) and [Rh(cod)(L144g)]BF4 (R1 = 1-Naph, R4 = (S)-CH2OTBDMS and R5 = H) catalyst proved also be useful in the asymmetric hydrogenation of functionalized olefins, such as dehydroamino acids (Scheme 92b) [316b]. This is again a quite unique feature, since the reduction of both unfunctionalized and functionalized alkenes follows very different catalytic cycles, and each type of substrate has been shown to require a particular catalyst type (Rh/PP-catalysts for functionalized and Ir/PN-catalysts for unfunctionalized) for optimal results [35e].Interestingly, the use of ligand L144g (R1 = 1-Naph, R4 = (S)-CH2OTBDMS and R5 = H) and its derivatives containing different silylated protecting groups (R4 = (S)-CH2OTBPMS and (S)-CH2OTIPS) also furnished high enantioselectivities in the Pd-catalyzed allylic substitutions (ee’s up to 99%) [316c]. Interestingly, for cyclic substrates it is possible to select the enantiomeric series of the substitution product, as in the case of the hydrogenation of terminal alkenes, by swapping the configuration of the biaryl phosphite moiety.Another strategy to overcome the problem of controlling the configuration of the S-thioether group after coordination to the metal center is the exchange of the thioether group by a chiral sulfoxide. Fig. 49 shows the most successful P-sulfoxide ligands developed to date. Ligands L145 and L146 were prepared from the corresponding bromo and 1,3-dibromobenzenes, which reacts with (R)-tert-butyl tert-buthanethiosulfinate to yield the corresponding sulfoxides. From the latter, the desired phosphine moiety was diastereoselectively introduced via ortho-lithiation (Scheme 93 ). Similarly, reaction of two equivalents of (R)-(tert-butylsulfinyl)benzene with 1-(dichlorophosphaneyl)piperidine led to ligand L148. Ligand L147 was efficiently prepared from (2S, 3S)-2,3-diphenyl aziridine. Aziridine ring opening with 4-Br-thiophenol followed by diastereoselective oxidation led to the corresponding amino-sulfoxide compound. Reaction of the latter with 2-(diphenylphosphanyl)benzoic acid gives access to ligands L147 (Scheme 93).Ligands L145 were the first successful application of P-sulfoxide ligands to several asymmetric reactions [317]. Thus, Rh/L145 provided high enantioselectivities in the Rh-catalyzed 1,4-addition of arylboronic acids to a large number of electron-deficient olefins (up to 98% ee; Scheme 94 a) [317a]. The presence of a second sulfoxide moiety at the other ortho position of the phosphine group (ligand L146; R = Ph) allowed the construction of chiral γ,γ-diarylsubstituted carbonyl compounds via the same reaction, that led to the preparation of bioactive compounds such as sertraline (ee’s up to >99%; Scheme 94b and Section 6) [318]. Ligands L145 also allowed the first Cu-catalyzed formation of α-aryl-β-borylstannanes by means of a three-component borylstannation of aryl-substituted alkenes (Scheme 94c) [317b]. Such transformation relies in the efficiency in controlling the stereochemistry of B-Cu addition as well as its ability to facilitate the transmetallation of enantioenriched alkyl-Cu species with retention of configuration. More recently, an efficient cooperative Cu/Pd-catalyzed asymmetric allylboration of alkenes has been reported (Scheme 94d) [317c]. The application of CuOAc/L145 (R1 = OiPr, R2 = iPr) in combination with Pd(dppf)Cl2 catalysts allowed the three-component reaction of styrenes, B2(pin)2 and allyl carbonates to be done with ee’s as high as 97%.Phosphine-sulfoxide ligand L147 (Fig. 49) was fruitfully applied (ee’s up to 99%) in the Pd-catalyzed allylic substitution with a very broad nucleophile scope (various malonates, including examples with different functionalities at the α-position, as well as ketoesters, amines, alcohols and indoles (Scheme 95 ) [319].The N-phosphine-bis(sulfoxide) ligand L148 (Fig. 49) have recently been used in the allylic etherification and amination of 1,3-diarylallyl acetates with benzylic alcohols and amines (ee’s up to 99%; Scheme 96 a) [320a]. Even more interesting are the excellent results achieved in the allylic alkylation of puzzling unsymmetrically 1,3-disubstituted allylic acetates. Pd/L148 catalyzed the dynamic kinetic resolution of this class of substrates with indoles (up to 84% yield with up to 95% ee; Scheme 96b) [320b]. The bifunctional character of this ligand is responsible for this favorable stereocontrol. The authors postulate that the two sulfoxide moieties play a different role, while one of them tightly coordinates to Pd, the other one directs the nucleophilic attack via a hydrogen bond interaction. Finally, ligand L148 also provided excellent enantioselectivities in the Rh-catalyzed 1,4-addition of arylboronic acids to cyclic enones (up to 98% ee) and of sodium tetraarylborates to chromenones (up to >99% ee; Scheme 96c) [320c].A ligand acquires an even greater value if, in addition to showing a high reaction and substrate applicability, it can be used in the total synthesis of relevant chiral compounds. We here compile some applications of previously discussed heterodonor ligands to total synthesis (Table 1 ). Different P-oxazoline/pyridine/amine/imine/thiazole/imidazoline ligands, five families of P,O and three families of P-thioether/sulfoxide ligands have found applications. The examples clearly illustrate the potential of these ligands in total synthesis. However, as can be see below, only a small set of the ligands commented has found suitable application to date. Among these we have phosphine-oxazoline L1, with electronic withdrawing groups in the phenyl ring and the phosphine moiety of the PHOX ligand, and the spiro based-SIPHOX ligand L32 with many applications each. The other three groups of ligands are the phosphine-amine L75, the phosphinite-pyridine L107, developed by Pfaltz, and the family of N-phosphine-thiazole ligands L84 and L88 developed by Andersson's group. The low application or no application for the rest of ligands may be due, in part, to the fact that research has focused mainly on finding the right ligand/s for specific asymmetric catalytic reactions. Specifically, in finding the correlation between the structure of the ligand and the catalytic capacity. Another factor to take into account is the availability of a chiral ligand. Still many of these ligands are prepared through multi-step syntheses, from costly starting material and/or toxic reagents and with low yields, which reduces their commercial interest. The progress made in the recent years to obtain ligands that are made in fewer steps and that can be manipulated in the air will hopefully widen the spectrum of chiral ligands that are commercially available.The success of phosphine-oxazoline ligands (PHOX) inspired the development of many new P-oxazoline ligand families that cover modifications in the ligand scaffold and/or the steric /electronic characteristics of the phosphine group to the change of the phosphine group with a phosphinite or phosphite group. New development arrived also by replacing the oxazoline group by several other N-donor groups and O- and S-donor groups. The structures of these chiral heterodonor ligands have gained in diversity and new families of very efficient ligands have emerged, which have allowed to improve catalytic performance in some asymmetric transformations. In addition, the majority of these ligands maintain the short and efficient synthetic route developed with PHOX ligands. The utility of the most successful heterodonor ligand families has been described in this review. We have shown how a suitable ligand design, aided by mechanistic studies, these ligands have become versatile ligands for metal-mediated asymmetric reactions, with superior catalytic performance in many reactions than the best C2-symmetric N,N and P,P-ligands reported so far. Their excellent results together with its easy synthesis and tailor-made modularity spread the way to new generations of heterodonor ligands and to further enlarge the range of processes catalyzed by them. This will help the progress and will therefore drive the growth of asymmetric catalysis as a vital element to achieve the sustainable production of enantiopure compounds in the coming years.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.We gratefully acknowledge financial support from the Ministerio de Economía y Competitividad (CTQ2016-74878-P), Ministerio de Ciencia e Innovación (PID2019-104904GB-I00), European Regional Development Fund (AEI/FEDER, UE), the Catalan Government (2017SGR1472) and the ICREA Foundation (ICREA Academia award to Montserrat Diéguez).

The success of phosphine-oxazoline ligands (PHOX) inspired the progress in P-oxazoline ligand families by modifying either the ligand backbone, the electronic and/or steric properties of the phosphine group or by exchanging the phosphine to a phosphinite or a phosphite group. In this respect, the structures of chiral P-oxazoline ligands have become more diverse and new families of very efficient ligands have emerged, which have improved catalytic performance in some asymmetric transformations, with an increased versatility, both in the range of reactions and in the range of substrates/reagents. In addition, most of ligands are synthesized from easily accessible chiral amino alcohols, maintaining the short and efficient synthetic route developed for PHOX ligands. New ligands have been developed by replacing the oxazoline functionality with several other N-donor groups, e.g. imidazole, thiazole, oxazole, pyridine, etc., and O- and S-groups. This review offers a critical overview of the utility of these most successful bidentate heterodonor P-N, P-O and P-S ligand families applied in metal-mediated processes. We illustrate how, through proper ligand design, these heterodonor bidentate ligand families can be an excellent source of ligands, with superior catalytic performance in many asymmetric reactions than the best C2-symmetric N,N and P,P-ligands reported so far.

Data will be made available on request.Modern society is heavily dependent on the chemical industries for most of its needs. This dependency comes with a huge environmental impact as a result of the constant emission of greenhouse gases from their chemical processes. In this regard, the development and integration of green technologies, such as the solid oxide electrolyzer cell (SOEC) in the industries, will play a big role in reducing these emissions. For example, as advanced high-temperature electrochemical devices, SOECs can produce green chemical intermediates (such as CO) from industrial gaseous waste like CO2 through the use of renewable energy sources [1,2]. In addition to providing chemical intermediates, the difficulty in the storage of renewable energy at peak seasons can be solved by utilizing the renewable energy in SOEC operations (Power-to-X) [1–3]. Green hydrogen, for example, can be produced from renewable energy sources through electrochemical water splitting (Equation (1)). Furthermore, synthesis gas (CO + H2), an important synthetic fuel in downstream processes, can also be produced through the co-electrolysis of H2O and CO2 [4–6]. From thermodynamics perspective, the simultaneous electrolysis of CO2 and steam is possible as the reduction of CO2 during co-electrolysis is facilitated by the reverse water gas shift reaction (RWGS) [3,7]. Eqs. (1) and (2) represent the separate steam and CO2 electrolysis reactions with their energy demands. The reactions during co-electrolysis operations involve a much more complex process due to the catalytic RWGS reaction (Eq. (3)) taking place at the same time with the steam and CO2 electrolysis [4,8]. [1] H 2 O → H 2 + 1 2 O 2 , Δ H r ( 900 ° C ) = 249 k J ‧ m o l - 1 . [2] C O 2 → C O + 1 2 O 2 , Δ H r ( 900 ° C ) = 282 k J ‧ m o l - 1 . RWGS [3] H 2 C O 2 → H 2 O + CO , Δ H r ( 900 ° C ) = 33 k J ‧ m o l - 1 . The high operating temperatures of SOEC (700–900 °C) eliminates the need for a noble-metal catalyst as well as allows for higher efficiency and increased production rate [9,10]. However, the thermodynamic advantage offered by the high-temperature operation presents a challenge for electrode material selection, especially for the fuel electrode which is in contact with different types of fuel. So far, Ni-YSZ (yttria stabilized zirconia) cermet has been the conventional electrode for the fuel electrode [11,12]. Ni acts as an electron conductor as well as an electro-catalyst for the reduction of fuel gases. YSZ, on the other hand, provides the pathway for ionic conductivity [13]. Extensive research in the electrochemical characteristics of Ni-YSZ electrode has already been performed, indicating its good electro-catalytic activity under different operating conditions [11–15].However, during long term operation, a significant degradation is observed especially under electrolysis mode [16,17]. Microstructural degradation as a result of Ni-migration and Ni coarsening was identified as significant factors causing the degradation at the Ni-YSZ fuel electrode [16]. Furthermore, different microstructural optimization techniques have been applied to reduce the Ni-migration [18,19]. For example, Ovtar et al. [18] studied the influence of GDC infiltration on the long term durability of Ni-YSZ cermet in steam electrolysis conditions. They reported that the infiltration of GDC nanoparticles into the Ni-YSZ cermet significantly decreased the voltage degradation rate by more than 900%. Recently, efforts to completely replace the YSZ oxide phase with the GDC have also been made [20–22]. Unlike YSZ, GDC is a mixed ionic and electronic conducting material under reducing conditions. In a reducing atmosphere, the Ce4+ is reduced to Ce3+ which generates the electronic property in addition to the ionic property. A positive consequence of this reduction is the extension of the triple phase boundary (TPB) beyond the electrolyte/electrode/gas interface to almost the entire surface area of the electrode exposed to the gas phase. Hence, the kinetics of the electrochemically active species increases leading to a lower polarization [23,24]. Also, the catalytic activity of ceria-based materials has been linked to the reversible transition between Ce4+ and Ce3+ [25–27].With regards to the electrochemical processes and performance, most literature [28–31] has reported the presence of two dominant arcs in the Nyquist plot of the Ni-GDC electrode obtained in fuel cell mode by electrochemical impedance spectroscopy (EIS). For example, Fu et al. [29] have reported the presence of two arcs in the Nyquist plot of a Ni-GDC fuel electrode and LSCF-GDC oxygen electrode, a dominant low frequency arc and high frequency arc. Lomberg et al. [30] and Macedo et al. [28] have also reported two dominant arcs from Ni-GDC electrodes cells fabricated with a commercial Ni-GDC powder and a Ni infiltrated GDC cell respectively. The two dominant Nyquist arcs were attributed to a diffusional process in the low frequency and a charge transfer process in the high frequency. A Ni-GDC symmetrical cell was also analyzed by Aravind et al. [32] in humidified hydrogen (40:60, H2:N2), and they identified three processes at 1123 K while Sumi et al. [33,34] identify about 5 processes through DRT deconvolution of the spectra for the single cell containing Ni-GDC fuel electrode. In steam electrolysis conditions, Athanasiou et al. [25] applied the analysis method of differences in the deconvolution of their impedance spectra obtained from the Ni-GDC fuel electrode and LSM oxygen electrode. They identified four different electrode processes; two low frequency processes (0.04–0.20 Hz) followed by two intermediate frequency processes (5–30 Hz).Following the inconclusive literature reports on the electrochemical processes in SOEC mode, this study aims to shed more light on the electrochemical processes occurring in Ni-GDC fuel electrodes as well as to investigate the long-term stability behaviour in both steam and co-electrolysis conditions. To this aim electrolyte-supported single cells (Ni-GDC//8YSZ//GDC//LSCF) were fabricated, followed by electrochemical characterization using electrochemical impedance spectroscopy (EIS) at different temperatures. The impedance data were also obtained at different partial pressures of steam, CO2 and oxygen at OCV. Furthermore, measurements under polarization (0.7–1.4 V) were recorded in order to investigate the charge transfer process. Equivalent circuit models, as well as distribution of relaxation times (DRT), were used to investigate the dominant electrochemical processes present in single cells.Electrolyte supported single cells were prepared for the electrochemical measurements. The fuel electrode consists of commercial NiO–Ce0.9Gd0.1O0.95 (GDC) powder from Marion Technologies (NiO:GDC, 65:35) while the LSCF (La0.58Sr0.4Co0.2Fe0.8O3-δ) oxygen electrode powder was synthesized through the modified Pecheni method [35]. The fuel electrode paste was made by mixing NiO-GDC in 6% ethyl cellulose (binder) and α-terpineol. Afterward, the slurry was mixed using a planetary vacuum mixer (THINKY Mixer ARV-310) and further homogenized by roll milling for about 30 min. The button cells were fabricated by using dense 8YSZ electrolyte supports from Kerafol® (d = 20 mm, thickness ∼ 250 μm). A thin layer (∼ 4–5 μm) of GDC was screen printed (EKRA screen printing Technologies) on one side of 8YSZ substrates as a barrier layer (sintered at 1350 °C for 1 h under air). The fuel electrodes (15–18 μm) were afterward screen printed on the opposite side of the electrolyte. Four different sintering temperatures i.e. 1200, 1250, 1300 and 1350 °C for 2 h (with 2 °C/min ramp) were considered. Among them, 1200 °C for 2 h was selected as an optimized sintering condition based on the polarization resistance (Supplementary Information Fig. S1) and good adhesion. The pure LSCF layer was screen printed on the side containing the GDC barrier layer (sintered at 1080 °C for 3 h). Finally, a thin NiO layer (∼ 8 μm) was also screen printed on the fuel electrode side as a current collector. The single cell configuration before reduction of the NiO in the electrode is represented as NiO-GDC//8YSZ//GDC//LSCF, with an effective area of 0.785 cm2.The single cells were electrochemically characterized using a NorEcs Probostat™ set-up. For the electrochemical measurements, the cell was heated up to 900 °C (at 1°/min) under N2. After reaching 900 °C, the Nickel oxide in the fuel electrode was gradually reduced to Nickel metal by systematically replacing N2 with H2 as described here [36], with a total flow rate of 9 Nl‧h−1. On the oxygen electrode side, a flow rate of 9 Nl‧h−1 of compressed air was used. After the reduction, the voltage against current density (I–V) characteristics and electrochemical impedance measurements were recorded using an IviumStat potentiostat/galvanostat. Three different measurements were carried out; the first is impedance measurement as a function of temperature from 900 to 750 °C with 25 °C steps at the open-circuit voltage (OCV). The frequency range during the measurement was from 0.11 Hz to 110 kHz with an AC amplitude of 50 mV and 21 frequencies per decade. The second set of impedance measurements was carried out at a different gas composition (psteam, and pCO2) as well as different partial pressures of oxygen at OCV conditions. Each of the measurements was carried out in both steam (H2O:H2, 50:50) and co-electrolysis (H2O:CO2:H2, 40:40:20) conditions. Finally, the long-term stability tests of the single cells were carried out under steam electrolysis (H2O:H2, 50:50, 3.6% fuel utilization) and co-electrolysis conditions (H2O:CO2:H2, 40:40:20, 2.3% fuel utilization) at 900 °C with −0.5 A‧cm−2 current density for 500 h.The EIS was utilized in obtaining the impedance spectra while the complex nonlinear least-square (CNLS) fit analysis procedure, as well as the distribution of relaxation times (DRT) procedure, were applied to analyze the acquired impedance spectra. For the CNLS analysis, an equivalent circuit model (ECM) which best describes the impedance spectra was proposed. The ECM was subsequently fitted to the measured data by using a commercially available CNLS-fit program (RelaxIS® software, RHD-Instruments).The DRT analysis technique was applied on the measured impedance spectra [37–40]. Common electrochemical equivalent circuit elements; RQ, Warburg and Gerisher elements have their corresponding DRT analytical forms. The RQ represents a constant phase element (CPE) in parallel to a resistance, wherein the CPE models the behaviour of an imperfect capacitor. In the DRT transformation, each RQ-equivalent circuit element gives a peak at specific relaxation frequencies in the DRT plot. On the other hand, asymmetric equivalent circuit elements, such as the finite length Warburg element and the Gerisher element, have complex analytical equations, thereby exhibiting more than one peak over a range of frequencies; one large peak at a characteristic frequency followed by additional smaller peaks at higher frequencies [37,38,41]. As a consequence, the additional peaks can overlap with other electrochemical processes and impede effective deconvolution of the measured spectra. Also, it is worthy to note that the presence of noise in the measured impedance spectra can generate additional artificial peaks in the DRT [42].The microstructure of the measured cells was analyzed with Quanta FEG 650 (FEI©) scanning electron microscope. The samples were immersed in resin. After polishing, a copper tape was looped around and the sample was sputtered with gold in order to reduce the charging effect. A detailed description of the sample preparation is found here [43].Electrochemical measurements of single cells were performed in the temperature range of 750–900 °C, after the reduction of the fuel electrode. The current-voltage characteristics, as well as polarization resistance (Rp), were then obtained as previously mentioned [44]. Experimental open circuit voltages of the cells under each measurement condition were well within 15 mV of the theoretical Nernst potential which implies adequate cell sealing. The OCV of the co-electrolysis conditions is seen to be slightly lower than those of steam electrolysis which agrees with the theoretical calculations from the Nernst equation (Supplementary table T1). Fig. 1 a and Fig. 1b illustrate the I–V curve of the cells under both steam (50:50, H2:H2O) and co-electrolysis conditions (40:40:20, H2O:CO2:CO). In general, decreasing the operating temperatures resulted in a decrease in current density and an increase in polarization resistance. For example, in the steam electrolysis mode, the current density reached at 1.5 V decreased from 1.31 A‧cm−2 at 900 °C to 0.41 A‧cm−2 at 750 °C while the Rp increased from 0.061 to 0.31 Ω‧cm2. A similar trend was observed for the co-electrolysis conditions. The decrease in electrochemical performance is attributed to a decrease in the cell kinetic activities with decreasing temperatures. Furthermore, the I–V curves in both electrolysis modes illustrate good continuity across the OCV implying that these Ni-GDC cells can work in reversible SOCs in fuel cell mode [11]. The effect of the difference in the percentage of the fuel gas composition (50% steam and 80% H2O/CO2) can be seen in the I–V curves. For example, in EC mode (at 1.5 V, 900 °C), the steam electrolysis conditions, despite having a lower fuel percentage, exhibits a similar current density (1.31 A‧cm−2) to that of co-electrolysis (1.37 A‧cm−2). Fig. 1c and d compares the Rp as well as the Arrhenius plots of ohmic resistance (Rs) and polarization resistance (Rp) under steam and co-electrolysis conditions. The Rs corresponds to the high frequency intercept with the real axis on the Nyquist plot while Rp represents the difference between the low and high frequency intercept with the real axis in the Nyquist plot. The steam electrolysis, even with a lower fuel percentage (50% steam and 80% H2O/CO2 at 9 Nl‧h−1), shows lower Rp values, especially at higher operating temperatures, than the co-electrolysis condition. For example, at 900 °C, the steam electrolysis shows a Rp value of 0.061 Ω‧cm2 while 0.089 Ω‧cm2 is obtained in co-electrolysis mode, thereby implying a higher activity towards H2O reduction than CO2/H2O reduction. At lower temperatures (750 °C), the steam and co-electrolysis modes exhibit similar Rp values of 0.319 Ω‧cm2 and 0.313 Ω‧cm2 respectively. This could be attributed to the domination of the electrode processes and resistances by lower electrode conductivities at lower operating temperatures. Furthermore, no difference is observed in the ohmic resistance of both electrolysis modes. This suggests that the ohmic resistance contribution is not affected by the difference in the fuel gas. The activation energies for the resistance contribution Rs and Rp were calculated from the slope of the Arrhenius plot, according to the equation [4]. [4] ln R = − ln σ 0 + E A R g T The steam and co-electrolysis conditions show similar activation energies of 77 ± 4 kJ‧mol−1 and 75 ± 4 kJ‧mol−1 respectively. To explain the similar activation energies; in the presence of sufficient steam percentage in co-electrolysis conditions, the CO2 is converted to CO through the reverse water gas shift [8,45], hence the electrochemistry of both conditions tend to show similar behaviour. A similar activation energy of 90.54 kJ‧mol−1 was obtained by Grosselindermann et al. [46] for the symmetrical cell of Ni-GDC under steam electrolysis conditions. A separate analysis of the LSCF symmetrical cell showed that the oxygen electrode has only a little contribution to the total Rp, indicating that the fuel electrode is the major contributor to the Rp [46].For the ohmic resistance, an activation energy of 69 kJ‧mol−1 (0.71 eV) is obtained, which is close to values reported for the ionic conductivity in YSZ electrolytes in literature [47,48]. This indicates that the ohmic resistance is controlled by the resistance of the thick electrolyte and therefore the same for steam- and co-electrolysis in this study.The quality of the measured impedance data was investigated by applying the Kramers-Kronig transformations test, as a mathematical validation tool [49]. For the impedance evaluation, the measured spectra data were evaluated with both DRT and CNLS fitting. The DRT analysis allowed for the deconvolution of physical processes while also highlighting the frequency ranges associated with each process. To accurately identify the electrode processes, measurements were taken under varying conditions; temperature variations allow the identification of relaxation frequencies of thermally activated processes, and measurement under varied oxygen partial pressure and fuel composition allows for the identification of oxygen and fuel electrode processes respectively. As a consequence of extensive DRT analysis, an equivalent circuit model (Fig. 2 a) consisting of three time constants (RQ) in series with a finite length diffusion (Warburg short element) was developed to evaluate the impedance spectra via a CNLS fit procedure. P1, P2 and P3 in the DRT plot (Fig. 2b) are modelled as 3 RQ-element, while P4 corresponds to a finite length Warburg short element. P4a peak is attributed to a possible satellite peak of the Warburg short element. The inductive effect in the high-frequency region of the Nyquist plot, originating from the wiring, is accounted for by an inductance (I) element, and the ohmic losses are accounted for by the ohmic resistor (Rs) in series with the R//CPE and Ws.To evaluate the quality of the fit and the chosen ECM, a simulated impedance spectrum and DRT plot were generated from the proposed ECM model. Fig. 2a shows an exemplary comparison between the experimental and simulated Nyquist plot as well as the DRT (with a lambda value of 10−6) for the steam electrolysis at OCV and 900 °C. The plots show a qualitative agreement between the chosen model and experimental results based on the impedance features and characteristics relaxation frequencies. The small peak at the low frequency region (0.1 Hz) is regarded as a possible artifact in the spectrum from the experimental measurement set-up. Furthermore, the residuals obtained from the fitting (Supplementary info S2) are uniformly distributed around the frequency axis with a relative error value of around 2%. This is an indication that the applied ECM can adequately reproduce the measured impedance spectral over the recorded frequency range [50]. However, since residuals are only a mathematical quantity, the physical correctness of the proposed model in comparison to the measured impedance must be validated by additional experiments by measuring impedance spectra under different operating conditions.To investigate thermally activated processes in the electrode resistance, the cells were measured under different operating temperatures from 750 to 900 °C at OCV and constant fuel gas composition (50:50, H2O:H2 and 40:40:20, H2O:CO2:CO). Fig. 3 a and b illustrate the Nyquist plot (without the ohmic resistance) of the impedance spectra in both steam- and co-electrolysis conditions as a function of temperature. A similar trend in temperature dependence was observed in both electrolysis mode, i.e., a shift in the spectra towards higher resistance values as well as an increase in the magnitude of the arcs with a decrease in temperature. Furthermore, the obtained impedance spectra exhibit two distinct arcs; a low-frequency arc and a high-frequency arc. The high-frequency arc demonstrates a significant temperature dependence, shifting towards higher values with a decrease in temperature. At lower temperature values of 775 °C and 750 °C, this arc dominates the Nyquist plot. On the other hand, the low-frequency arc shows little or negligible dependence on temperature. Similar features were also observed in the literature [25,46]. Fig. 3c and d show the DRT plots of the impedance spectra as a function of temperature. In general, an increase in the peak area, which corresponds to an increase in resistance, is observed with a decrease in temperature. In both electrolysis conditions, the high frequency P1, P2 and P3 processes show strong temperature dependence, indicating they are thermally activated processes. The P1 doesn't show any shift in the frequency while peaks P2 and P3 shift towards lower frequencies with a decrease in the temperature. At lower temperatures, the peak P3 dominates and overlaps with suspected satellite peaks of a finite length Warburg process. Such overlap limits effective and complete deconvolution of the processes. The P4 peak, which occurs within the frequency range of 1 Hz–10 Hz, is found to show the least dependence on temperature. This is particularly true in steam electrolysis. In co-electrolysis, an inconsistent temperature dependence is observed in the DRT. Fig. 3e and f show the Arrhenius plots of the individual resistances obtained from the fitting of the ECM. It is observed that the high frequency resistances R1 and R2, corresponding to process P1 and P2 in the DRT plot respectively, have the most pronounced dependencies on temperature. Calculation of the activation energies from the Arrhenius plot shows that R1 has an activation energy of 127 (1.32 eV) and 130 kJ‧mol−1 (1.35 eV) for steam and co-electrolysis respectively. Such a high frequency process with high activation energy could be attributed to charge transfer processes in the bulk/TPB of the electrodes [42,51,52]. The fit also illustrates that the low-frequency Ws element exhibits very small activation energies, especially in steam electrolysis. In the co-electrolysis, the inconsistency observed at the low frequency of the DRT appears to show a slight negative activation energy in the fitting. However, the trend illustrates a virtually independent temperature variation. Such a small activation energy is usually attributed to gas diffusion process [11,51,53]. In general, the Arrhenius plot, which shows good agreement with the DRT, highlights that the high frequency regime exhibits significant temperature dependence while the low-frequency regime is almost temperature independent. The notable remark is in the mid-frequency regime (R3), which shows low activation energy in the fit while the DRT plot suggests a thermally activated process. This discrepancy is attributed to the limitation in effective deconvolution of a possible temperature-dependent process due to the presence of satellite peaks of the Ws, occurring within the same frequency range.The impedance measurement under different partial pressures and fuel gas compositions was also performed to identify the impedance contributions from the fuel electrode. In steam electrolysis, the steam content was systematically reduced from 50% to 10% steam, while in co-electrolysis, the composition of steam and CO2 was simultaneously varied from 10% to 50%. Fig. 4 a and b show the variation of impedance spectra as a function of change in the fuel gas at 900 °C. In both electrolysis modes, a decrease in resistance with an increase in the steam partial pressure is observed. In the co-electrolysis mode, this implies that the cell shows higher kinetics towards H2O reduction than CO2 reduction [11]. The Nyquist plots exhibit the characteristics high and low frequency arcs. The high-frequency arc shows a mild increase in magnitude while the low-frequency arc exhibits a remarkable increase in the magnitude with a decrease in the steam partial pressure. This indicates a pronounced dependence of the low-frequency arc on steam partial pressure. Fig. 4c and d show the DRT spectra of the impedance spectra as a function of partial pressure and composition. The four distinct peaks, P4, P3, P2 and P1, observed in the DRT plots correspond to the Ws, R3, R2 and R1 of the equivalent circuit fitting, respectively. In both electrolysis modes, P4, P3 and P2 exhibit an observable dependence on the partial pressure variation, increasing in magnitude with a decrease in steam partial pressure. The P1 peak exhibits only a slight dependency to fuel partial pressure variation. The low-frequency peak (P4) shows the most drastic dependence on the fuel gas composition variation in both electrolysis modes. However, in co-electrolysis mode, the P3 peak is virtually unchanged by the change in gas composition. Fig. 4e and f show the plots of resistances R1, R2, R3 and Ws as a function of H2O partial pressure in steam and co-electrolysis modes. It can be clearly seen that Ws has the most significant dependencies in steam partial pressures. R3, on the other hand, illustrates mixed dependence, showing slight dependence in steam electrolysis and virtually unaffected in co-electrolysis. Hence it is uncertain if this process is partly a fuel electrode or exclusively an oxygen electrode process. R1 and R2 exhibit observable dependence on the partial pressure variation.The characteristics behaviour of the low-frequency process, both in steam and temperature variation, is usually attributed to a diffusion process [53–55]. The observed diffusion process appeared to be dominant in the obtained impedance spectra. In such a case, Bessler et al. [54,55] proposed the term, gas concentration impedance for the impedance caused by gas-phase activities. Generally, the impedance contribution from the gas diffusion process in electrolyte supported cells are considered to be reduced due to the smaller thickness of the fuel electrode in comparison to the thicker substrate in anode supported cells. However, the fuel electrode fabrication method, the fuel gas channel and the contact mesh also contribute significantly to the gas diffusion impedance [46,54,55]. Also, surface diffusion of species could add some contribution to the low frequency impedance peak. Unlike the EIS of Ni-YSZ [44,51], the low frequency process of Ni-GDC electrodes has been majorly associated with either a finite-length Warburg or finite-length Gerischer impedance behaviour rather than the conventional RQ time constant [25,30,46]. This is possibly due to the interplay between gas diffusion processes and other surface exchange processes associated with MIEC electrodes [25,46]. However, this low frequency process exhibits an almost independent temperature variation with a low activation energy. This indicates that this process is most likely dominated by gas diffusion process.Also, the electrochemical activities of GDC-containing electrodes have been reported to exhibit a low frequency peak due to the chemical capacitance of the GDC phase [30,46,56]. For example, in the analysis of Ni-infiltrated GDC electrodes, Lomberg et al. [30] reported a possible coupling of the gas diffusion process and chemical capacitance at the low frequency regime. The chemical capacitance of a specimen refers to its ability to store chemical energy via oxygen vacancies in response to changes in the local oxygen chemical potential. In the Ni-GDC electrode, the chemical capacitance is suspected to originate from the variation of the oxygen nonstoichiometry of the GDC phase [30,56]. However, chemical capacitance is reported to show pronounced dependence on temperature variation [30,46,56]. On the contrary, there is almost no temperature dependency in the low frequency process in our analysis, especially in steam electrolysis. This is in agreement with the results obtained by Watanabe et al. [57], wherein they observed insignificant temperature dependence for the low frequency process. They, however, attributed the low frequency process to a gas conversion impedance that has a non-stoichiometry capacitance. To further clarify this, an analysis of the DRT plot (Supplementary Information Fig. S3) from an electrolyte supported Ni-YSZ on the same test rig show the occurrence of the characteristic low frequency peak within the same relaxation frequency range. This suggests that the low frequency process is rather dominated by the gas diffusion process, with possible additional contributions from the chemical capacitance due to the non-stoichiometry of the GDC phase. Similar results were observed by Aravind et al. [32]. In their study of symmetrical Ni-GDC cells, the low frequency process, fitted with a Ws, was assigned to a gas phase-diffusion process. Grosselindenmann et al. [46] attributed the low frequency process to a coupled process of gas diffusion and activation polarization.In general, the controversy in the attribution of the low frequency process could be attributed to the different fabrication methods, composition (Ni:GDC ratio) as well as cell configurations employed by different groups. For example, a 50:50 ratio of Ni:GDC was used by Grosslindenman et al. [46] for a cell with a configuration Ni-GDC//GDC//YSZ//GDC//LSCF. Meanwhile, in our work, a 65:35 Ni: GDC ratio was used without a GDC sandwich between the fuel electrode and the electrolyte. While the percentage of the GDC in the cermet may have played a little role, the presence of a GDC layer between the fuel electrode and the electrolyte may have significantly affected the electrochemistry at the electrolyte/electrode interface. To further shed more light to this, single cells consisting of GDC fuel electrode and LSCF oxygen electrode was analyzed using impedance spectroscopy. Preliminary results of the impedance spectra indicates that the low frequency region shows significant temperature dependence, which points to the chemical capacitance of the oxygen nonstoichiometry of the GDC phase [30,46,56]. However, the details of this result as well as a comprehensive study of the influence of the Ni-GDC composition, microstructure and cell configuration would be the scope of another paper.The P2 (R2) process shows significant temperature and pH2O dependency. On the other hand, P3 (R3) only exhibited slight steam partial pressure dependence, especially in steam electrolysis while also showing pronounced temperature dependence. This means that these two processes are fully (or partly) fuel electrode processes. Investigations of the response of surface oxygen vacancies, electrons and adsorbates on the ceria surfaces-gas interface confirm that the desorption/adsorption, as well as the electron transfer between the species, is a rate-determining step in the water-splitting reaction [58,59]. Generally, molecular water adsorption on a bare ceria surface is energetically more favourable than complete dissociative adsorption. However, the presence of oxygen vacancies on reduced ceria greatly enhances water dissociation over molecular adsorption [60]. This causes the electrochemical reduction of the fuel gas on the GDC electrodes to be dominated by surface reactions [24,60]. The detailed mechanistic steps for water splitting reaction on ceria electrodes is still unclear, however, it has been confirmed that the OH− is an existing specie on doped ceria and that it participates as an active intermediates species for both water-splitting reaction and hydrogen oxidation reaction [25,58,60]. Hence, the mid frequency processes are therefore, associated with the surface electrode reaction, most likely desorption/adsorption of species of the fuel gas species in addition to a charge transfer process.To investigate the oxygen electrode contribution to the impedance spectra, the impedance measurements were performed on the single cells by varying the oxygen partial pressure from 0.1 to 1 atm. Fig. 5 a shows the DRT plot of the impedance spectra as a function of oxygen partial pressure at 900 °C. The fuel gas composition and partial pressure of steam (H2O:H2, 50:50) were kept constant. From the plot, the low and mid-frequency peaks of P4, P3 and P2 exhibit a partial increase in peak values with a decrease in pO2. This indicates that these peaks are partly oxygen electrode processes. Similar trends were also obtained for the resistance and Warburg elements from the fitting of the impedance with the equivalent circuit in Fig. 5b with R2 exhibiting the most pronounced dependence. The R1 and R3 resistances on the other hand suggest a slight dependency on pO2 variation. A notable observation from the electrochemical investigation is that the high frequency resistance R1 exhibits a similar (slight) dependence in both pH2O (Fig. 4e and f) and pO2 (Fig. 5b). However, the DRT (Fig. 5a) shows that the P1 peak is unaffected by pO2. Therefore, considering the contradictory trend between the P1 peaks of the DRT (Fig. 5a) and the R1 of the ECM fitting (Fig. 5b), it is, therefore, arguable to assign this process as partly an oxygen electrode process. To resolve this difficulty, a symmetrical LSCF cell was fabricated and measured in a two-electrode measurement in air at 900 °C at 0 V. Fig. 5c illustrates the DRT of the impedance spectrum obtained from symmetrical LSCF electrodes as compared to those of the single cells.The DRT obtained from the symmetrical cell was normalized by dividing the resistance contribution by two. Four major peaks are observed in the DRT of the LSCF oxygen electrode, a low frequency peak (P4), two intermediate frequency peaks (P3 and P2) and a high frequency region (P1). The electrochemical behaviour of LSCF has been extensively studied, hence the electrochemical processes were inferred from literature [51,52,61]. Literature reports on similar LSCF electrodes attributed the small low frequency peak (P4) to diffusion on the oxygen electrode, while the intermediate frequency processes of P3 and P2 are assigned to oxygen surface exchange kinetics and oxide ion diffusion respectively. Lastly, the P1 process is attributed to a charge transfer process [42,51,61,62].The comparison of the DRT plot between the symmetrical cell and single cell indicates that the high frequency P1 peak is a contribution from both the oxygen electrode and fuel electrode, most likely a charge transfer process at the oxygen electrode and the charge transfer process at the TPB of the fuel electrode. Since GDC is a MIEC material, charge transfer reaction can occur both at the DPB (of GDC and gas) and the TPB (of Ni, GDC and gas). The intermediate frequency peaks of P2 and P3 are an interplay between the oxygen electrode and the fuel electrode. The P2 peak is mainly a fuel electrode process while P3 is mainly an oxygen electrode process. The low frequency oxygen diffusion process is seen to be shifted to a lower frequency from the diffusion process of the fuel electrode. This illustrates that the little peak around 1 Hz (P4a in Fig. 5a) is part of the gas diffusion process of the oxygen electrode. A separate time constant was not assigned to this process since it overlaps with the fuel electrode contribution. Attempts to fit the data with an added time constant to account for this DRT peak did not result in significant improvement in the fit results.In summary, the electrochemical analysis of the impedance spectra highlights that the electrochemical activities of single cells of Ni-GDC fuel electrodes and LSCF oxygen electrode are dominated by four electrode processes; a low frequency peak, two intermediate peaks and a high frequency peak. Table 1 summarizes the possible electrochemical reaction steps of the Ni-GDC with the LSCF oxygen electrode.To investigate the long-term stability of the cell, degradation measurements were carried out for 500 h at 900 °C and −0.5 A‧cm−2 in both electrolysis modes. Two different cells were measured for each electrolysis mode. The first cell was operated without any intermediate measurements during the degradation while for the second cells, I–V and impedance measurements at OCV as well as under polarization (from 0.7 V to 1.4 V) were taken every one hundred hours to investigate the evolution of the degradation. Fig. 6 a shows the degradation behaviour in steam (50% H2O) and co-electrolysis conditions (80% fuel) in the cell operated without interruptions. In steam electrolysis conditions, two different degradation regions were observed; a low degradation region observed within the first 150 h followed by a progressive degradation region. A smaller degradation rate is obtained in the co-electrolysis conditions than in steam electrolysis. The co-electrolysis exhibits a degradation rate of 308 mV‧kh−1 while 499 mV‧kh−1 is observed in the steam electrolysis. Direct comparison of the obtained degradation rates with those of Ni-YSZ is challenging due to the different operating conditions (fuel gas mixture, operating temperature and current density) used in the literature [11,17,63]. For the second cell operated with intermediate I–V and impedance measurements, a steady decrease in the current density is observed with degradation time. For example, in steam electrolysis, the current density at 1.5 V decreased from 1.11 A‧cm−2 at 100 h to 0.41 A‧cm−2 after 500 h (Supplementary Fig. S4). A similar trend was observed for the co-electrolysis mode, but to a lesser extent.Analysis of the degradation behaviour using the impedance spectra recorded every 100 h (Fig. 6b and c) shows that the ohmic resistance plays the most significant role in the degradation of the cell. Table 2 summarizes the obtained ohmic and polarization resistances as a function of degradation. In steam electrolysis, for example, the ohmic resistance increased from 0.47 Ω‧cm2 at the start of degradation to 1.14 Ω‧cm2 after 500 h. Similarly, in co-electrolysis, the ohmic resistance increased from 0.53 Ω‧cm2 at the start of the degradation to 1.05 Ω‧cm2 after 500 h. Comparing the ohmic resistance degradation between steam and co-electrolysis, it is observed that the steam electrolysis (0.67 Ω‧cm2) showed a higher ohmic degradation than co-electrolysis conditions (0.51 Ω‧cm2). In literature, the increase in ohmic resistance is attributed to the formation of the resistive SrZrO3 phase at the GDC/electrolyte interface [64,65]. Also, a gradual surface oxidation of the Ni mesh has been reported [20,66] to cause an increase in the ohmic resistance. Such current collector oxidation increases the contact resistance of the electrode and the currect collector. Furthermore, the higher ohmic resistance in the steam electrolysis could also be attributed to increased pore formation near the electrode/electrolyte interface due to a possible increase in Ni migration away from the electrolyte [16]. In addition to the ohmic resistance, the Nyquist plots illustrate that the high/mid-frequency regime exhibits a pronounced contribution to the degradation in both electrolysis modes. However, this is more pronounced in steam electrolysis. Evaluation of the Rp with degradation time indicates that in steam electrolysis, the Rp increased from 0.07 Ω‧cm2 at the start of degradation to 0.28 Ω‧cm2 after 500 h (300% increase). While in co-electrolysis, the Rp increased from 0.10 to 0.16 Ω‧cm2 (60% increase). This implies that, in addition to the ohmic contribution, the Rp also shows a pronounced contribution to the degradation rate, especially in steam electrolysis, while in co-electrolysis the Rp contribution is less severe. Fig. 6d and e shows the variation of the individual resistances as a function of degradation time. In both electrolysis modes, R1 and R2 exhibit the most significant effect on the degradation rate. However, the contribution of these two resistances as a function of degradation time is more pronounced in steam electrolysis than in co-electrolysis. The higher contribution of these resistances in steam electrolysis degradation could be attributed to greater Ni particle growth and depletion due to higher p(H2O) in steam electrolysis [16,67]. Furthermore, the impedance spectra under polarization obtained during the degradation time shows that the high frequency region is also mostly affected by degradation (Supplementary Fig. S5).For the LSCF oxygen electrode, R1 and R2 which are partly oxygen electrode processes, significantly contributed to the degradation. Reported FIB-SEM and ICP-OES studies on aged LSCF electrodes illustrate that the degradation phenomena is not due to any microstructural changes but rather due to the presence of Sr-rich surface phases after cell aging [68–71]. Hence, the contribution of the oxygen electrode (majorly in R1 and R2) can be attributed to Sr surface segregation on the LSCF oxygen electrode (Supplementary Fig. S6). The low frequency processes were less affected by the degradation time. Overall, the observed degradation is attributed to both the fuel and oxygen electrodes.The microstructure of the degraded cells was analyzed with a scanning electron microscope and compared with a reduced cell that was not electrochemically operated. SEM images were obtained with both the secondary and backscattered electron detectors. ImageJ® software was used to analyze the approximate particle size distribution of the fuel electrode. A comparison of the microstructure of the degraded cells under different electrolysis modes was performed. Fig. 7 a–c shows the microstructure of the reduced fuel electrode while Fig. 7e–g and Fig. 7i–k represent the microstructure of the cell after degradation in steam and co-electrolysis conditions respectively. Fig. 7(d, h and l) represents the microstructure of the corresponding oxygen electrode. Firstly, the microstructures illustrate an increase in the Ni particle size of the degraded fuel electrode as compared to the reduced cell. Based on the fuel electrode SEM images covering a cross section of 13 μm at 3 different locations on the electrode, an approximate mean particle size was determined. The reduced cell has an average Ni particle size of 1.37 μm while 2.19 μm (indicating a 62% increase) and 2.86 μm (exhibiting a 109% increase) were obtained for the co-electrolysis and steam electrolysis mode respectively. The particle size distribution is shown in Fig. 8 . This points to depletion and agglomeration of Ni particles during cell operation. A similar Ni particle growth of 140%/100 h was observed by Holzer et al. [72] for a Ni-GDC electrode operated in humid atmosphere (60% H2O, 40% N2/H2). The Ni particle agglomeration is, however, observed to be more severe in steam electrolysis than in co-electrolysis. The 47% higher Ni particle growth observed in the steam electrolysis as compared to the co-electrolysis conditions could be attributed to the difference in p(H2O). The higher steam partial pressure could have facilitated more Ni particle growth [72–74] as well as depletion of volatile Ni species, most notably Ni(OH)2 [16,67]. Such Ni particle growth will invariably reduce the surface area for electrochemical reaction. Furthermore, the backscattered electron images of the reduced cell show an even distribution of the GDC particles as compared to the degraded cells, which show an uneven GDC particle distribution. This points to a loss of GDC percolation in the measured cells. The increased Ni particle growth and the higher loss of GDC percolation in steam electrolysis as compared to the co-electrolysis could have contributed to the significant impact of the R2 process on the degradation test in steam electrolysis when compared to the co-electrolysis conditions (Fig. 7b and c). The GDC is observed to cover parts of the Ni particles as could be seen in Fig. 7(g and k). The extent to which such coverage impacts the electrochemical performance of the electrode is still unclear, however, the loss of GDC percolation and the coverage of Ni particles could further reduce the electrochemical reaction zones leading to an increase in the resistance of the underlying process. Overall, the post-test analysis of the cells highlights three observations; Ni depletion and agglomeration, loss of GDC percolation and lastly, coverage of the Ni particles with GDC [75]. These observations could be (partly) responsible for the observed reduction in cell performance during operation. Loss of GDC percolation and Ni depletion could lead to loss of connectivity between the Ni and the oxide phase leading to decrease in the electrochemical reaction zone. With regards to the microstructural changes observed and reported in literature [16,63,67] for Ni-YSZ after degradation, pronounced Ni coarsening and Ni depletion at the electrode/electrolyte interface have been reported as a major degradation phenomenon. However, a direct comparison of the extent of the Ni coarsening reported in the literature with those of the Ni-GDC could not be made due to the different operating conditions (temperature, gas pressure, test rig design) used by different working groups.For the oxygen electrode, no significant change is observed in the microstructure of the three cells. However, as the GDC barrier layer is not completely dense, migration of volatile SrO to the electrolyte interface cannot be entirely hindered. As a consequence, the SrO reacts with zirconia to form the insulating SrZrO3 at the electrolyte interface (Supplementary Fig. S6). Furthermore, Sr segregation and formation of cobalt oxide particles have also been reported to cause LSCF degradation [68–71,76].In this study, a comprehensive electrochemical impedance analysis of electrolyte-supported single cells comprising Ni-GDC fuel electrode and LSCF oxygen electrode was performed by CNLS fitting to an Equivalent Circuit Model (ECM) and Distribution of Relaxation Times (DRT), both under steam and co-electrolysis conditions. The impedance spectra were obtained in the 750–900 °C temperature range. Further measurements were also carried out at different steam compositions as well as different H2O/CO2/CO compositions for co-electrolysis at OCV. The observed processes were further modelled using 3 R//CPE elements in series with a finite length diffusion element (Ws). The low frequency process, modelled with a finite length Warburg short, is attributed to diffusion and surface processes, the two intermediate frequency processes are attributed to an overlap of (surface exchange and ion diffusion processes in the) oxygen electrode and (surface electrode reactions with charge transfer in the) fuel electrode processes. The high frequency process corresponds to a charge transfer at both electrodes. Long-term stability tests of the single cells were carried out under steam electrolysis (H2O: H2, 50:50) and co-electrolysis (H2O:CO2:CO, 40:40:20) conditions at 900 °C with −0.5 A‧cm−2 current density for 500 h. Steam electrolysis conditions exhibit the highest degradation rate of 499 mV‧kh−1, while a lower degradation rate of 308 mV‧kh−1 is observed under co-electrolysis conditions. The post-test analysis of the operated cell shows Ni depletion and agglomeration, loss of GDC percolation as well as coverage of the Ni particles with GDC. Ifeanyichukwu D. Unachukwu: Methodology, Investigation, Formal analysis, Validation, Conceptualization, Data curation, Software, Visualization, Writing – original draft, Writing – review & editing. Vaibhav Vibhu: Methodology, Formal analysis, Validation, Conceptualization, Software, Supervision, Visualization, Writing – review & editing. Izaak C. Vinke: Methodology, Supervision, Validation, Project administration, Conceptualization, Resources, Software, Visualization. Rüdiger-A. Eichel: Supervision, Funding acquisition, Project administration, Resources. L.G.J. (Bert) de Haart: Methodology, Supervision, Validation, Funding acquisition, Project administration, Conceptualization, Resources, Software, Visualization.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.The authors gratefully acknowledge funding from the German Federal Ministry of Education and Research (BMBF) through the iNEW 2.0 Project: incubator for sustainable and renewable value chains, under grant agreement number 03SF0627A.The following is the Supplementary data to this article: Multimedia component 1 Multimedia component 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.jpowsour.2022.232436.

The present study investigates the electrochemical performance and degradation behaviour of a Nickel - Gd2O3 doped CeO2 (Ni-GDC) electrode containing single cell under steam electrolysis and co-electrolysis modes. The cell consists of the Ni-GDC fuel electrode, an 8 mol% Y2O3 stabilized ZrO2 (8YSZ) electrolyte layer, a GDC barrier layer and a (La,Sr)(Co,Fe)O3 (LSCF) oxygen electrode. Firstly, the electrolyte-supported single cells were fabricated and characterized using DC- and AC-techniques in the 750–900 °C temperature range. Distribution of relaxation times (DRT) analysis was employed to resolve frequency-dependent electrode processes. The observed processes were further modelled using an equivalent circuit model (ECM) with 3 R//CPE (resistor//constant phase element) in series with a finite length diffusion element (Warburg short - Ws). Long-term stability tests of the single cells were carried out under steam electrolysis (H2O:H2, 50:50) and co-electrolysis (H2O:CO2:CO, 40:40:20) conditions at 900 °C with −0.5 A‧cm−2 current density for 500 h. Steam electrolysis conditions exhibit the highest degradation rate of 499 mV‧kh−1, while a lower degradation rate of 308 mV‧kh−1 is observed under co-electrolysis conditions. The post-test analysis of the operated cell shows increased Ni particles size, suggesting Ni agglomeration in both electrolysis modes.

We, as humans, consume energy in different forms in our day to day life. Most of it is in the form of electrical energy that comes from the burning of finite fossil fuels. Combustion of fossil fuels harms our environment due to large scale carbon emission and alternative solutions with clean and sustainable energy are required to overcome our dependency on fossil fuels. Solar energy and wind energy are the most exploited among renewable energies. The amount of solar energy reaching the surface of the earth is several folds greater than that required for human progress so even a low conversion performance would be satisfactory to fulfill the energy requirements. [1] However, wind and solar energy generation are not constant and need efficient and economical storage facilities and conversion to better fuels. Energy conversion to hydrogen releases no carbon dioxide (CO2) and it can also generate clean discharge (water) on combustion. [2,3] This has attracted much attention due to its green nature and high energy density. Water splitting is amongst the many promising technologies that enable the production of hydrogen and oxygen, which converts sustainable renewable energy into ideal chemical energy using electricity or sunlight. [4,5] This energy conversion technology has amassed worldwide attention owing to its high energy conversion efficiency, a potentially wide range of applications, and negligible environmental pollution. [6–8] Integrated solar water-splitting systems that integrate light-capturing semiconductors with electrocatalysts to effectively split water display specific promise as a means of direct sunlight to fuel production. [9] Thus generated fuels can be used in various applications to reduce carbon emission dramatically. For example, the hydrogen used in a hydrogen driven engine does not burn the fuel like in the conventional internal combustion engine. Instead, it fuses with the oxygen from the air to form H2O instead of CO2. No primary wastes are produced in the electrochemical splitting of water when the electricity is used from solar cells. As a result, it is considered to be a clean process for producing H2 and O2 gases.Electrochemical water splitting can be classified into two half-reactions and they are OER and HER. These reactions are kinetically inactive and require an overpotential to occur at a practical rate. The overpotential loss due to OER is generally much greater than HER. Therefore, OER is generally viewed as the bottleneck of water splitting. [9,10] As a consequence, water electrolysis requires effective catalysts to expedite these reactions smoothly and unhindered. Presently, carbon-based platinum is the most efficient catalyst for HER [11–14] while the standard catalysts for OER are ruthenium- and iridium- based oxides. [15] The first reported work on the OER catalyst was by Meyer and co-workers [16] on the so-called “blue dimer” a binuclear oxo-bridged ruthenium complex. Even though these are excellent electrocatalysts, scarcity in nature and high cost make it difficult to administer them into large scale applications. This gave rise to the widespread study of different non-precious earth-abundant catalysts that have high efficiency for both OER and HER. Considerable progress has been made in using transition metals for better efficiency in OER. Intellectual studies of Co, Ni, Fe, and Mn-based oxides in OER dates back to more than half a century ago. [17–20] Among these metals, special mention goes to cobalt-based compounds for OER and HER reactions. Although cobalt has no biological relevance in water splitting, and although it is significantly less abundant than Fe, Mn, or Ni, it is now emerging as a fascinating metal due to its catalytic power for OER and HER. [21] Cobalt compounds, either as molecular species or as three-dimensional materials, seem to be attractive multi-electron catalysts for both HER and OER in the water-splitting process.Since OER is considered to be kinetically sluggish, researchers have done extensive studies to overcome this drawback in overall water splitting. Cobalt seems to be an effective catalyst for enhanced and fast OER. Herein, we discuss recent developments regarding characterization, design, and evaluation in the field of cobalt-based electrocatalyst based on noble metals, noble metal alloys, transition metal compounds, perovskite, and functional nanocarbon structures. Further, we share the challenges faced and our outlook on this topic. We also consider future aspects as well.A classic electrolyzer consists of three components, i.e. anode, cathode, and an electrolyte. The electrocatalysts are coated on the electrodes for a speedy activity. In such a system the electrocatalysts used are different on the anode and the cathode which facilitates OER and HER, respectively. But recent studies have come up with new electrocatalysts which are bifunctional for both OER and HER. Ruthenium- and iridium- based compounds had been considered as the novel catalysts for OER due to their high performance in a wide range of pH. [15] Since these precious metals are scarce and expensive, earth-abundant transition metal catalysts are studied to replace them. Generally, electrolyzers operate under high conductive media, i.e. either acidic or alkaline. One drawback of these transition metals is that they degrade in an acidic condition due to high oxidative potential which, in turn, hinders the activity of the catalyst. So, to overcome this, most of the transition metal-based OER catalysts are studied under alkaline medium.We can find many proposed mechanisms for OER in literature. [22–25] Let us denote “M” to be the active site of a catalyst. In the first step, a hydroxyl radical (OH−) is adsorbed onto the active site (M) to give M–OH by an electron (e−) oxidation on OH−. Then another OH− reacts with M–OH to give M–O by removing a proton and electron pair. M–OH and M–O are common intermediates in most of the proposed mechanisms. The dissimilarities start after the formation of M–O. Generally, there are two reaction mechanisms to form oxygen gas (O2) from M–O. In the first (green pathway), it is the direct combination of two M–O intermediates to produce oxygen gas and free active site M. In the second (white pathway), there is a formation of M–OOH intermediate (hydroperoxide intermediate) by the nucleophilic attack of OH− on M–O, paired with an e− oxidation. A further proton paired electron transfer results in the decomposition of M–OOH into O2(g) and the free active site M. This forms the basis for the majority of the mechanisms proposed with the change in the number of electron transfer in individual steps (single/multiple electron reaction). The bonding interaction within the intermediates (M–O, M–OH, and M–OOH) is decisive for the overall electrocatalytic ability. [26] A brief schematic explanation is given in Fig. 1 .The mechanisms are proposed based on certain electrocatalytic kinetic parameters like overpotential (η) and Tafel slope (b). These are also used to evaluate the electrocatalyst’s catalytic performance.Overpotential is defined as, the difference between the potential required to practically run the reaction and the theoretically found out equilibrium potential of the reaction. It is one of the important factors for assessing the performance of an OER catalyst. But, it is difficult to get the exact value of the overpotential. So, the potential value at a constant current density is taken as the overpotential. Normally, a 10 mA cm−2 current density value is set to find the overpotential of the target reaction.For OER, the overpotential is calculated as the potential difference between the potential reaching a current density (i) of 10 mA cm−2 and the equilibrium potential of 1.23 V. Generally, an electrocatalyst with an overpotential in the range of 0.3–0.4 V is considered to be of excellent catalytic activity for OER.Tafel slope is usually drawn to evaluate the reaction kinetics and mechanism. It is also used to contrast the catalytic activity of different catalysts. As stated by the Butler-Volmer equation, (1) i = i O e x p α a n F E RT + exp α c n F E RT where “i” is the current density, “io” is the exchange current density, “αa” the anodic charge transfer coefficient, “αc” the cathodic charge transfer coefficient, “R” the universal gas constant, “F” the Faraday’s constant (96485C mol−1), “n” the number of electrons involved in the electrode reaction, “E” applied potential, and “T” absolute temperature (K).When there is very high overpotential for the anodic electrode, in the above equation the overall current is largely due to the anodic electrode. Therefore the equation can be simplified as, (2) i ≈ i O e x p α a F η RT The above equation is known as the Tafel equation. [27] This can be further reduced as, (3) log ( i ) = log i O + η b (4) b = σ η σ l o g ( i ) = 2.303 R T α F Tafel slope (b) defines how swiftly the current increases with the overpotential applied. It also helps in finding the rate-determining step (RDS) and formulating a mechanism for the reaction. In an OER, the RDS can be either a single electron reaction or multiple electron reaction. In a single electron reaction mechanism, the transfer coefficient, designated by “α”, becomes the symmetry factor (β) which is given by the equation below, (5) α = β = 1 2 + η λ where λ is the re-organization energy. From this, the Tafel slope shows a value of 120 mV dec−1. In other words, if the Tafel slope of an electrocatalyst is 120 mV dec−1 then the RDS is a single electron reaction and accordingly, a mechanism can be proposed. Whereas in multiple electron reaction, according to Bockris and Reddy, the transfer coefficient is formulated as (6) α a = η b ν + η r β where η b is the number of electrons that transfer back to the electrode before RDS, ν is number of RDS that have occurred in the overall reaction and η r is the number of electrons that are involved in the RDS. The Tafel slope and transfer coefficient are therefore related to the number of electrons participating in the reaction. Therefore, different Tafel slope defines different RDS and hence different mechanisms for the reaction. In order for an OER catalyst to be considered good in catalytic activity, it should possess a low Tafel slope.To obtain a low Tafel slope, various cobalt-based catalysts have been reported in the literature by different teams of researchers. Mainly on noble metals, transition metals, perovskites, and carbon-based cobalt compounds. We will discuss the recent developments in the above-mentioned categories.In an electrolyzer, the OER reaction is kinetically sluggish and requires higher overpotential than the HER reaction. Therefore, catalysts with enhanced activity are required to overcome this drawback. Noble metals are among the catalysts that show high activity as well as durability in a vast range of pH for OER reactions. Iridium, ruthenium, gold, and silver are some commonly used noble metals for this purpose. However, because these elements are scarce and expensive, researchers have tried to use them optimally or replacing them in many ways by synthesizing catalysts consisting of cost-effective alternatives while maintaining equal or better performance. Here we discuss mainly the cobalt-based noble metals used for improved OER activity. Since cobalt has high activity for OER, it can be doped with another active catalyst to better the performance.Iridium has been used as a novel catalyst for OER due to its high stability and activity in both acidic and alkaline medium. Cobalt linked with iridium for OER has been reported by many research groups. Eunju Lee Tae et al. have studied crystalline cobalt oxides nanoparticles (nc-CoOx) on ITO (indium tin oxide) glass substrate doped with ~ 5 mol % crystalline iridium oxide nanoparticles (nc-IrOx). [28] It displayed a much lower overpotential (η) and Tafel slope (b) in comparison to the nc-CoOx electrode and nc-IrOx electrode. Under a buffer solution of 0.1 M phosphate, the η at 1 mA cm−2 values of [nc-CoOx]ITO were 0.37 V at pH 7 and 0.34 V at pH 13, whereas for [nc-IrOx/nc-CoOx]ITO were 0.22 V at pH 7 and 0.19 V at pH 13. A drop of 0.15 V can be seen regardless of the pH condition. The Tafel slope for [nc-IrOx/nc-CoOx]ITO was 29 to 34 mV dec−1 and for [ncIrOx]RDC (rotating disc carbon electrode) was 38 to 44 mV dec−1. A work performed by Youkui Zhang et al. saw the development of a self-assembled 3-dimensional Cobalt-Iridium (CoIr-x) hierarchical structures using a one-step reduction path with NaBH4 as a reducing agent. [29] The Ir species were incorporated into the irregular surfaced cobalt-based hydroxide nanosheets (3D CoIr-x) as clusters and single atoms. It showed enhanced activity in the neutral and alkaline medium compared to the commercially available IrO2 electrocatalyst. [29] In the neutral media (1.0 M phosphatic buffer solution, pH@7), the CoIr-0.2 (0.2 is the molar ratio of Ir to Co in precursor or CoIr with 9.7 wt% Ir content) showed a η of 0.373 V@10 mA cm−2 current density, which is relatively lower when compared to the IrO2 (0.431 V@10 mA cm−2). Regarding the Tafel slope, CoIr-0.2 exhibited a value of 117.5 mV dec−1, which is also below the value for IrO2 of 132.1 mV dec−1. Whereas in the basic media (1.0 M potassium hydroxide (KOH), pH@14), CoIr-0.2 needed an η of 0.235 V to achieve 10 mA cm−2 and Tafel slope of 70.2 mV dec−1. From the above-mentioned electrocatalyst studies, we can understand that the OER activity is favored more by the alkaline medium. This is possible because the rate-determining step of the OER reaction is the accelerated OH− discharge process for the aforementioned catalysts. [30] Recent studies indicate that by enabling electrochemical oxidation of metallic electrocatalysts in basic media one can considerably enhance the behavior of OER in neutral electrolytes as compared to the direct activation in neutral media without pre-oxidation of metallic electrodes. [31] Taehyun Kwon and co-workers fabricated hollow octahedral nanocages of Co-doped on IrCu alloy (Co-IrCu ONC/C) on carbon support [32], which showed excellent OER activity in addition to prolonged stability in acidic medium. The catalyst exhibited an η value of 0.293 V at a current density of 10 mA cm−2, as compared to Ir/C catalyst that showed an η of 0.315 V. A Tafel slope value of 50 mV dec−1 was reported for Co-IrCu ONC/C, and the catalyst also demonstrated high durability with a reduction of only 3 percent after 2000 cycles of CV analysis. In comparison, the Ir/C catalyst dramatically deactivated with a 50% decrease in current density. The outstanding performance of Co-IrCuONC/Cis credited tothe 3Dintegratedstructure. In another work, Waqas Qamar Zaman and the team constructed a multimetallic IrO2 catalyst by co-doping with two different 3d transition metals (nickel and cobalt) to atomically replace 50% of the precious metal. [33] It was developed by the hydrothermal method to make sure composite homogeneity is achieved and later crystallized at 400 °C. The lower crystal formation energy (calculated using DFT (density functional theory) studies) for co-doping was the main factor for the significant dopant penetration. The synthesized co-doped IrO2 (Ir-NC-50) demonstrated η at 10 mA cm−2 current density of 0.285 V which was significantly lower than individually doped IrO2 by cobalt and nickel. The Tafel data of Ir-NC-50 was 53 mV dec−1 lower than the classic IrO2 (65 mV dec−1). Therein, they also establish a linear correlation between the decreasing onset potential and the decreasing iridium concentration. This helps in reducing the iridium doping by replacing it with transition metals without any decrease inactivity. Stability studies were carried out by mounting the prepared material (Ir-NC-50) on the Ti plate in an acidic medium at a constant current of 10 mA cm−2 for 5 hrs. It showed negligible differences in activity, pre and post chronopotentiometry tests. Areum Yu et al. fabricated a crystalline one-dimensional (1D) tubular nanocomposite of iridium and cobalt (IrxCo1-xOy) through electrospinning and successive calcination. [34] The Nanocomposite with various Ir to Co ratios was created to test the efficiency and find the optimum loading of Ir to maximize the activity for OER reaction. Ir0.46Co0.54Oy nanotubes demonstrated the best activity for OER as well as high stability in alkaline medium. Scanning electron microscopy (SEM) images showed that Co3O4 had a smooth tubular structure and IrO2 a fiber morphology. The simple creation of Co3O4 shell in IrxCo1−xOy probably offers a prototype of the tubular structure, promotes the development of IrO2 by combining precursors with Ir, and ultimately creates the mixed IrCo oxide nanotubes. To assess the electrochemical properties of the optimum catalyst (Ir0.46Co0.54Oy), cyclic voltammetry (CV) was carried out in 1 M KNO3 aqueous solution. The CV curves possessed a rectangular shape, showing the swift charging and discharging processes. Electrocatalytic measurements were conducted on a rotating disc electrode (RDE) voltammetry in an aqueous solution of 1 M NaOH at 1600 rpm electrode rotation speed. Ir0.46Co0.54Oy nanotubes had a η value of 0.310 V at a current density of 10 mA cm−2 and a Tafel slope of 58.6 mV dec−1, which was lower compared to IrOy nanofibers and Co3O4 nanotubes. The Stability test did not show much change in the potential used to generate 10 mA cm−2 after 1000 repetitive scans. Transition-metal oxides could demonstrate greater stability during OER when compared to metals, possibly due to metal oxides being already present at higher oxidation state and further oxidation shifts is less likely to take place. [35] In a recent work by Yingjun Sun et al., they reported a new material of Pt-rich PtCo/Ir-rich IrCo trimetallic fishbone like nanowires, denoted as PtCo/Ir FBNWs. [36] The optimized Pt62Co23/Ir15 FBNWs only needed an overpotential of 0.308 V, much lower than the commercial Ir/C (0.380 V), to reach a current density of 10 mA cm−2. The catalyst also displayed great activity in a broad pH range. DFT calculations reported that the catalyst’s high activity was because of the modulation of highly electron active Ir-5d orbital on the Pt-based hetero-d-band-junction. It also acts as an outstanding trifunctional (ORR, OER, and HER) catalyst in a wide range of pH levels. Alloys containing five or more metal elements in a single phase are categorized as high entropy alloys (HEAs) [37], which provides ample possibilities to alter alloys’ catalytic activities and surface electronic properties. [38–41] Recently, Zeyu Jin and coworkers worked on nanoporous HEA (np-HEA) of AlNiCoIrMo alloy, which showed high activity towards both OER and HER in acidic medium. [42] They were synthesized by incorporating Ir with other four metals into one single nanostructure phase of de-alloyed Al-based precursor alloys with merely 20 atomic % of Ir. The composition effect showed that the as-synthesized np-HEA had a nano ligament size of about 2 nm. An overpotential value of 0.233 V was needed to attain the current density of 10 mA cm−2 and a mass current density of about 115 mA mg−1 in an acidic medium. Seung Woo Lee et al. achieved a highly active and stable 3D mesoporous binary oxide of Ir and Ru (MS- IrO2/RuO2) with an optimum Ir to Ru molar ratio of 1:10 via nano-replication followed by Adams method. [43] When compared with conventional IrO2/RuO2 (0.340 V), MS- IrO2/RuO2 had a lower overpotential of 0.300 V at 10 mA cm−2. After the accelerated stability test for over 2 h, the increase in overpotential was found to be 0.022 V for MS- IrO2/RuO2, whereas 0.044 V for IrO2/RuO2, demonstrating a better stability for MS- IrO2/RuO2 (Fig. 2 ).A ternary compound with iridium, ruthenium, and cobalt was studied by J.L. Corona-Guinto and co-workers. [44] They developed a RuIrCoOx powder by chemical reduction method, followed by a thermal oxidation process. Cyclic and linear voltammetry at 20 mV s−1 scan rate and 900 rpm in 0.5 M H2SO4, was utilized to analyze the electrochemical properties of the electrocatalysts. The RuIrCoOx specimens revealed a Tafel slope value of 70 mV dec−1 at low current densities and 108 mV dec−1 at high current densities. When compared to RuIrOx, Tafel slopes of RuIrCoOx were lower in both the high and the low current densities. Whereas, the overpotential of RuIrCoOx was found to have a value of 0.410 V at a current density of 18 mA cm−2. Chronopotentiometry experiments were performed in a current pulse of 0.25 mA cm−2 to 75 mA cm−2 in 300 s. The pulse length of the currents was found to be long enough to keep the voltage constant. This result shows that the use of mixed metals can contribute to synergetic effects that may increase the stability and selectivity of OER kinetics. Lei Wang et al. reported an active hollow Ru-modulated CoxP polyhedral structure (Ru-RuPx-CoxP) by adopting a facile solid–liquid-phase chemical method. [45] Catalytic activity measured the value of η@10 mA cm−2 as 0.291 V and a Tafel slope of 85.4 mV dec−1 in an electrolyte of 0.1 M KOH. The obtained value was low compared to RuO2 (η10 = 0.312 V) and IrO2 (η10 = 0.393 V). They found that Ru modulation can cause unstable surface termination, improve the electron transfer and promote the reaction kinetics by enhancing the density of states at the Fermi level to improve the electron transfer which further reduces the adsorption energy gap between the intermediates. Youngmin Kim et al. synthesized RuO2/Co3O4 nanowires by electrospinning process [46] , which showed an η of 0.410 V at 10 mA cm−2 when subjected to 1600 rpm in 0.1 M KOH solution saturated with oxygen. In comparison to Co3O4 and Ketjenblack (KB), RuO2/Co3O4 NWs showed a relatively greater OER current density and lower onset and overpotentials. The study established that the addition of highly active and conductive RuO2 onto the 1-D Co3O4 nanowire enhances the bifunctional performance of catalyst to a larger extent when compared to pure 1-D Co3O4. Caiyan Gao et al. reported ruthenium-cobalt nanoalloys encapsulated in carbon layers doped with nitrogen fabricated through incipient wetness impregnation pyrolysis process (RuCo@NC-750 °C, 1.56 wt% Ru). [47] It showed great stability and activity towards OER. ICP-AES test of RuCo@NC-750 °C after 10 h of CA test showed unchanged Ru and Co content compared to the initial sample. It exhibited an η value at 10 mA cm−2 of 0.308 V and a TOF value of 0.35 s−1 at η300 for OER. They related the enhanced activity to the introduction of Ru into the Co lattice matrix, which can significantly increase the transfer of an electron from the alloys to the carbon surface and increase the defects on the carbon surface. In a recent work by Pengsong Li et al., a monoatomic Ru attached on the surface of CoFe-LDH (Ru/CoFe-LDH), with an optimized wt% of 0.45 Ru, displayed an excellent OER activity with a η@10 mA cm−2 as low as 0.198 V, significantly lower Tafel slope of 39 mV dec−1 and high stability in basic solution when compared to CoFe-LDH and commercial RuO2 catalysts. [48] In-situ and operando XAS measurements along with DFT + U calculation indicated an electronic coupling that is strong between LDH and Ru support, which acts as a cocatalyst to reduce the kinetic energy barrier to form *OOH from *O intermediate and avoided the formation of the high oxidation state of Ru. A comparison of catalytic activity of different transition metal LDH as cocatalysts revealed a trend as η10 (Ru/CoFe-LDHs) (~0.198 V) < η10 (Ru/NiFe-LDHs) (~0.220 V) < η10 (Ru/NiCo-LDHs) (~0.240 V) < η10 (Ru/MgAl-LDHs) (~0.290 V). Shaoyun Hao and coworkers modified the electronic properties of NiCo-LDH by incorporating Ru cations by one-step chlorine (Cl−1) corrosion of Ni foam (NiCoRu-LDH@NF). [49] Due to the high adsorption energy and improved active sites, the catalyst showed η@100 mA cm−2 of 0.270 V, a low Tafel slope of 40 mV dec−1 (Fig. 3 ), and continued stability (55 Hours at 100 mA cm−2) in basic solution. As in the previous results, here also, Ru reduced the energy barriers from *OH to *OOH which accelerated the reaction kinetics of OER. Juan Wang and the team fabricated a Co-doped RuO2 NWs (molar ratio, Ru:Co = 19:1) by combining a facile wet-chemical process and post-annealing treatment. [50] They demonstrated an η@10 mA cm−2 as low as 0.2 V under acidic medium. First-principle calculations estimating the adsorption free energy of intermediates indicate a modulation in d-band center after metal doping, which is thought to be the cause of the enhancement in the OER activity. Jieqiong Shan et al. performed the generation of an alloy of RuIr nanocrystal with Co metal (Co-RuIr). [51] Addition of Co gave way to an enhancement of electronic structure of Co-RuIr alloy resulting in 0.235 V overpotential at 10 mA cm−2 and a Tafel slope value of 66.9 mV dec−1 in 0.1 M HClO4 media. This improvement was due to the expected Co leaching which results in increased concentration of O2− species, which further favors OER catalytic activity.Ruohao Dong et al. designed a 2-D layered double hydroxide (LDH) structure comprising of silver (Ag) decorated Co(OH)2 nanosheets (Ag@Co(OH)2) via a selective reduction–oxidation technique from metal nitrates. [52] In the field of electrocatalysis, cobalt compounds are granted close attention, among the 2-D materials community. Ag@Co(OH)2 showed excellent performance for OER and a yield of gram-scale, possibly because the near-surface oxygen vacancies in the Co(OH)2 nanosheets have the capacity to improve the electrophilic ability of O being absorbed, promote the adsorption of –OH on active sites, and shape the –OOH adsorbed species. [53–54] In addition, the Co(OH)2 interlayer spacing is greater than the typical transition metal hydroxides, resulting in a high ion transfer tendency and improved OER kinetic capability. [55] The electrochemical measurements were performed in an electrochemical workstation that has a three-electrode system in a solution of 1 M KOH (pH ~ 13.85) at room temperature. CV measurements were performed by sweeping the potential from a value of 1.3 to 1.7 V (vs RHE) at 10 mV s−1 sweep rate and LSV with a 5 mV s−1 scan rate. From the LSV curves η@10 mA cm−2 of Ag@Co(OH)2 was found to be as low as 0.270 V, whereas a pure Co(OH)2, synthesized using the same technique showed an η@10 mA cm−2 of 0.350 V. [56–57] They also prepared ternary materials comprising of cobalt, silver, and other transition metal elements (Ag@Co-Ni LDH and Ag@Co-Fe LDH), which showed better performance than the original Co LDH. This explained how silver doping on the LDH can enhance its efficiency by intensifying the catalyst’s electrical conductivity. During the chronopotentiometry analysis (about 8 hrs.), the catalytic performance of Ag@Co(OH)2 faded only faintly with time. Ag@Co(OH)2 showed the greatest value of current density of 37.6 mA cm−2 @ 0.350 V that was threefold more than that of pure Co(OH)2 nanosheets. In the case of Tafel slope Ag@Co(OH)2 exhibited a value of 67 mV dec−1. Kai-Li Yan et al. synthesized an Ag-doped Co3O4 nanowire array reinforced FTO (Ag-Co/FTO) via a facile electrodeposition-hydrothermal process in acidic media (0.5 H2SO4), which presented 1.91 V vs RHE onset potential and 0.680 V overpotential. [58] The results indicate that the vertical growth of Co3O4 nanowire was a result of Ag film deposition on the FTO substrate, and the formation of the mesoporous nanostructure can be attributed to Ag2O in Ag-Co hydroxide precursors. In a similar work, B. Jansi Rani et al. reported Ag-doped Co3O4 nanorods via hydrothermal method at different temperatures (90, 120, 150, and 180 °C). [59] The sample synthesized at 150 °C showed a good activity towards OER with 0.344 V overpotential to achieve a current density of 9.45 mA cm−2. Conducting an EIS study and following the Nyquist plot, they concluded that the product synthesized at 150 °C had a low resistance of 13 Ω and a narrow arc radius indicating quicker interfacial electron transfer. The enhancement in activity could have been due to the incorporation of Ag into Co3O4 crystal lattice which enhances the electrical conductivity and reduces the catalyst’s internal resistance. It is also reported that Ag metal enhances the kinetic reversibility, specific capacitance, and redox reaction of pristine Co3O4. [60] In a different work, Xiaoyun Li et al. fabricated Ag loaded Fe-Co-S embedded on nitrogen-doped carbon composite (CISC-Ag) via calcination method and subsequent simple dipping route (Ag). [61] When they compared the electrocatalytic activity with pristine CISC, they found that CISC-Ag-3% (0.329 V) exhibited a lower overpotential at 10 mA cm−2 than pristine CISC (0.366 V). The measured photocurrent density of Hematite photoanode covered with CISC-Ag-3% (with catalyst loading mass of 0.06 mg cm−2) was found to be 0.527 mA cm−2 that was 7 times that the bare hematite photoanode (0.075 mA cm−2) at a bias voltage of 1.23 V vs. RHE. They elucidated that this activity was the result of the presence of metallic Ag in the catalyst, which suppresses the surface electron-hole recombination and endorses electron-hole formation. The electrical conductivity of Ag is mainly attributed to the loosely bonded outer electrons of Ag atoms. Therefore, adding Ag to any catalyst increases its electrical conductivity. Xu Zhao and coworkers worked on engineering the electrical conductivity of lamellar Ag-CoSe2 nanobelts (width 300–500 nm) via partial cation exchange method. [62] About 1% of Ag+ cations enhanced the catalyst’s OER stability and activity when compared to CoSe2 nanobelts. 1% Ag-CoSe2 demonstrated 0.320 V overpotential at 10 mA cm−2, 22.36 mA cm−2 current density at an η of 0.350 V and low Tafel slope of 56 mV dec−1. They interpreted that the introduction of Ag+ had two effects on the catalyst. Firstly, they found that Ag+ caused a 5.6% decrease in the electrochemical active surface area (ECSA) which hinders the activity; and secondly, due to the improved electron transport, there was an increase in Co4+ sites which promotes the OER activity. The small decrement in ECSA was compensated by a more active generation of Co4+ active sites.Aolin Lu and team worked on a core–shell structure of gold (Au)-cobalt nanoparticles, with the core being Au nanoparticle and shell being Co3O4 supported by carbon (AuCo/C). [63] They developed the structure by two different methods i.e. a facile one-pot synthesis and an injection synthesis. Due to the synergistic interaction between the shell and the core, core–shell nanocrystals (NCs) exhibit exceptional catalytic activity in comparison to single-component NCs. [64–65] Oxygen and nitrogen saturated solutions were used for OER activity measurements. Electrochemical polarization curves were analyzed at 5 mV s−1 scan rate after the CV test with the working electrode subjected to a rotation of 2000 rpm. They found that the optimal ratio of Au:Co to be 2:3 (Ag40Co60/C), which showed the highest value of current density of 428.6 and 366.9 mA mg−1 respectively, at 1.67 V, for OER in N2 and O2 saturated solutions when prepared by one-pot synthesis. A Tafel slope value of 65 mV dec−1 was exhibited by Au40Co60/C, which was lower than bulk Au and bulk Co3O4. Boon Siang Yeo et al. reported cobalt oxide (CoOx) deposited onto the surface of rough gold (Au) substrate as monolayers for better performance for OER reaction. [66] They reported the OER activity of 0.4 monolayer (ML) of CoOx on Au to be approximately 3 times greater than that of bulk iridium and 40 times greater than that of bulk CoOx under the same electrochemical conditions. The turn over frequency (TOF) of 0.4 ML CoOx on Au was found to be 1.8 s−1. This was mainly due to the growth in the surface CoIV population confirmed by Raman spectroscopy. Xunyu Lu et al. reported Au nanoparticles-doped mesoporous Co3O4 structure (Au/mCo3O4) synthesized via a nano casting process using mesoporous silica KIT-6 as a hard template. [67] An enhanced OER activity was obtained with 2.5% Au doped mCo3O4 which displayed a lower onset potential of 1.53 V (vs. RHE) and overpotential of η10 = 0.440 V when compared to mCo3O4 (onset potential = 1.56 V (vs. RHE), η10 = 0.520 V). This superior performance was ascribed to the doping of electronegative Au nanoparticles, highly exposed active sites, and large Brunauer–Emmett–Teller (BET) surface area. In a different work, Xu Zhao and associates developed CoSe2 nanobelts decorated with a trace amount (0.1 wt%) of isolated Au atom (Au1-CoSe2). [68] They probed from the study that the trace amounts of isolated Au atoms enhanced the exposure of cobalt active sites, limited the use of Au, shifted up the d-band center, and thereby reduced the H2O adsorption energy of active sites. An η of 0.303 V at 10 mA cm−2 and a Tafel slope of 42 mV dec−1 was calculated from the LSV curve of Au1-CoSe2 in O2 saturated alkaline medium (Fig. 4 ).To elevate a catalyst’s activity, synthesis of specific catalyst morphologies and nanostructures has been practiced for a long time. [69] Based on Mott–Schottky's effect, Zhong-Hua Xue and his team created a Janus particle based on cobalt. [30] A Janus Co/CoP nanoparticle was created by a controllable vacuum-diffusion method for continuous phosphidation of carbon-coated metallic cobalt nanoparticle. In Janus particle, one plane was metallic (Co) and the other was semi-conductive (CoP). Co/CoP had a high ECSA than that of pure Co sample which resulted in an enhanced activity for Co/CoP. Co/CoP-x showed OER activity in a wide range of pH which makes it different from other conventional electrocatalysts. Overpotential of Co/CoP-5 (η = 0.340 V, where 5 indicates the weight ratio of NaH2PO2 and Co elements) @ 10 mA cm−2 was significantly less than Co (0.430 V) and CoP (0.380 V) species in 1.0 M KOH basic media. In the neutral media (1.0 M PBS), Co/CoP-5 attained a current density of 2.6 mA cm−2 at a potential of 1.8 V (vs. RHE) and in acidic media (0.5 M H2SO4) it achieved a current density of 1.3 mA cm−2 at 1.8 V (vs. RHE) potential. The preparation of aligned cobalt-based Co@CoOx nanostructures was performed by Qi and team that was achieved by the pyrolysis of cobalt oxalate precursors. [70] The 2-dimensional alignment of the derived cobalt nanoparticles was guided by the precursor’s 2-dimensional morphology. To obtain a current density of 10 mA cm−2 the resulting electrocatalyst needs a significantly low overpotential value of 0.298 V. They performed the OER experiment in an aqueous solution of 1 M KOH on a simple glassy carbon electrode. The nanoparticles’ compact alignment as well as the metallic nature of the bulk of the catalyst is capable of assisting with the inner- and inter-particulate transfer of charge. In addition, the particles’ 2-dimensional alignment can, in general, intercept the dissolution and ripening of cobalt metal nanoparticles, as compared with isolated nanoparticles that is produced via traditional techniques. [71] The resulting electrocatalyst could also demonstrated a superior stability for electrolysis of 24 h at 1.55 V vs RHE. Jing et al. performed the synthesis of an oriented assembly constructed by hexagonal Co(OH)2 nanosheets. [72] A solvothermal technique in a mixture of methanol, water, and ethanol were first carried out to achieve single-crystalline CoC2O4 micro rods having ultra-high aspect ratio. The CoC2O4 was then converted into Co(OH)2 from anion-exchange by immersing the CoC2O4 precursor in a solution of alkali. The 1-dimensional microrod single crystals were converted into 2-dimensional nanosheets assembly and 3-dimensional structural voids were generated in the oriented assembly. The resulting materials combined the qualities of both 2-dimensional and 3-dimensional arrangements, displaying improved water oxidation activity in comparison to the freely isolated nanosheets. The assembly’s magnified performance is primarily due to two reasons. The first is that the anisotropic 2-dimensional nanosheet substructure possesses intrinsic high charge transfer capacity. [73–74] The fabricated ultra-long structure provides a directed pathway for the transfer of charge. The second reason is that the 3-dimensional structural voids elevate the accessible surface area and assist with mass transport. [75] The assembled arrangement can also prevent the aggregation and ripening that takes place during electrolysis for enhanced durability. [71] Relative to the isolated nanosheet counterpart the nanosheet assembly demonstrated a greater activity for OER at similar mass loading. The nanosheet assembly could achieve 10 mA cm−2 current density at an overpotential of 0.359 V, and to achieve the same current density, the isolated nanosheet displayed an overpotential of 0.394 V. In addition, the nanosheet assembly has a Tafel slope of 76 mV dec−1 when determined by LSV at 5 mV s−1. This value is comparatively smaller than that shown by the isolated nanosheets (95 mV dec−1). Wu’s group carried out a facile preparation of cobalt metal thin films through physical vapor deposition (PVD) on different nonconductive substrates. [76] These substrates include polyimide, mica sheet, regular and quartz glass, and polyethylene terephthalate (PET). When the group performed surface electrochemical modification using cyclic voltammetry, the films became active for electrocatalytic water oxidation. This activity was demonstrated by the film as it achieved 10 mA cm−2 current density at a low overpotential value of 0.330 V in a solution of 1 M KOH. This is also the value required for a photovoltaic equipment to achieve 12.3% solar-to-hydrogen efficiency. [77] The electrodes were also sturdy, and their activity remained unchanged during chronopotentiometry analysis of long durations. Furthermore, no other energy or time consuming treatments (e.g., annealing, aging, or anodic oxidation, are necessary to reach the achieved activity with the resulting catalyst film. Wan and team developed a new electrocatalyst of Co(OH)F for OER. [78] The 3-dimensional Co(OH)F microspheres were constructed by building blocks of 2-dimensional nanoflake, which are then woven by 1-dimensional nanorod foundations. Weaving and constructing the substructures of 1-dimensional nanorods and 2-dimensional nanoflakes could deliver high structural void porosity with abundant interior space in the 3-dimensional material synthesized. The Co(OH)F material’s hierarchical structure merges the merits of all material dimensions in heterogeneous catalysis. The advantages being possessed by the anisotropic low-dimensional (1-dimensional and 2-dimensional) substructures include swift charge transport and high surface-to-volume ratio. Furthermore, the nanorods’ interconnectivity is valuable for charge transport. The 3-dimensional arrangement generates adequate number of active sites per surface area and is useful for efficient mass diffusion during catalysis. With the synthesized material a low overpotential value of 0.313 V was required to drive an OER current density of 10 mA cm−2 in 1 M aqueous solution of KOH. Guo’s team employed a facile 2-phase protocol to synthesize an α-Co(OH)2 by utilizing sodium oleate as a phase-transfer surfactant. [79] The team regulated the structure and crystallinity of the α-Co(OH)2 by heat treatments toward improved electrocatalytic OER activity. Calcination of the synthesized α-Co(OH)2 at a temperature of 200 °C produced a networked and well-dispersed nanoparticles of CoO (Co-200). The CoO sample synthesized displayed an OER current density of 10 mA cm−2 under a low overpotential value of 0.312 V in a 1 M KOH aqueous solution. The improved activity could be described by the presence of ultra-small particle size and ample open spaces, both of which can deliver many surface catalytic sites. In addition, the onset potential for OER was 0.290 V and the Tafel slope was 75 mV dec−1. Zhao et al. successfully produced a unique hollow and porous CoO tetragonal prism-like structure through performing a facile and efficient co-precipitation technique. [80] With this technique, Co3(OAc)5OH particles of uniform size were synthesized using cobalt acetate in presence of polyvinylpyrrolidone (PVP K30). PVP itself possesses a strong coordination capability to metal ions through the N and/or CO functional groups. The Co3(OAc)5OH being produced has a highly uniform and discrete tetragonal prism-like system. The produced material was then calcinated at a temperature of 200 °C for a duration 3 h, in the presence of argon, and at a heating rate of 2 °C/min to obtain the porous CoO structure (CoO-200). High activity as well as high stability could be demonstrated by the porous and hollow CoO microprisms in 1 M KOH solution. A low overpotential of 0.280 V was needed to achieve 10 mA cm−2 current density. A Tafel slop of 70 mV dec−1 was also displayed by the Co-based catalyst that indicates a fast water oxidation kinetics. The high performance observed could be due to the synergistic effect that exists between 2 different but finely-distributed CoO crystalline phases, ameliorative crystallinity, uniform particle size, low mass transfer resistance, and high surface area exploited from the unique porous arrangement. Liang’s research team synthesized β-Co(OH)2/Co(OH)F hierarchical hexagrams with a six-fold symmetrical arrangement. [81] During the synthesis, hexagonal β-Co(OH)2 plates were first produced that behave as templates for the growth of Co(OH)F nanorods. An intermediate of β-Co(OH)2/Co(OH)F hybrid was then generated that consists of plate-like β-Co(OH)2 hexagonal cores appended with 6 rod-like CO(OH)F branches. Long reaction durations could lead to the complete conversion of β-Co(OH)2 hexagons that resulted in the formation of authentic six-branched Co(OH)F nanorods. As a result, nanorods of Co(OH)F were ordered into a six-fold symmetry. Another point to note is that along β-Co(OH)2 hexagon edges the growth of Co(OH)F nanorods could be observed as lateral branches in place of perpendicular to hexagons. The unusual epitaxial growth mechanism is regarded to be because of the matching between a-axis of β-Co(OH)2 crystals and the b-axis of Co(OH)F crystals, which is advantageous for electrocatalysis. Relative to pure Co(OH)F and β-Co(OH)2, the hybrid material could demonstrate enhanced water oxidation activity such as lower overpotential of 0.329 V to deliver 10 mA cm−2 current density. Liang et al. performed a polyvinylpyrrolidone (PVP)-assisted pyrolysis to carry out the transformation of ZIF-67 into meso/microporous cobalt-embedded nitrogen-enriched carbon (Co-NC) material for both OER and ORR. [82] During pyrolysis, PVP was enclosed within ZIF-67 in one-pot and remained in it. With this technique, the breakdown of the porous structure at low temperatures could be avoided. The group chose PVP for a number of reasons. The first one was that PVP could be encapsulated because of the strong coordination interaction between the CO groups in PVP and the metal ion sites in metal–organic frameworks. [83] The second reason was due to the findings by Lai et al. in which they found that a PVP derived carbon//ZIF derived carbon interfacial structure could be generated in PVP/ZIF nanocomposites. [84–85] The interfacial arrangement may lead to the improved electrocatalytic activities. Therefore, the group proposed that the meso/microporous Co-NC material could modify the pyrolysis functioning of ZIF-67, producing large electrochemical surface area. In addition, the presence of PVP could cause an increase in N content and the generation of the interfacial structure that could further contribute to the enhanced OER and ORR electrocatalytic activities. The group synthesized a number of materials and amongst them P-Co-NC-4 (4 being the synthetic PVP/Co2+ molar ratio) demonstrated the best activity. Analysis of the sample using LSV showed that it displayed an onset potential of 0.90 V. Its overpotential value at 10 mA cm−2 current density was 0.315 V and it displayed a Tafel slope of 75.7 mV dec−1. Liang and group prepared quasi-single-crystalline CoO hexagrams that was characterized by structural long-range ordering and plentiful oxygen vacancies as defects. [86] The material was synthesized at β-Co(OH)2/Co(OH)F hexagrams’ critical phase transition point. The matching between the a-axis of β-Co(OH)2 crystals and the b-axis of Co(OH)F crystals is vital for the generation of CoO hexagram single crystals. The resulting material, specifically P-400 (400 = pyrolysis temperature, P = pyrolysis step), possessing abundant defects were very efficient for the oxidation of water as it demonstrated a low overpotential value of 0.269 V to deliver a current density of 10 mA cm−2 in 1 M KOH aqueous solution. Liang and team used a simple preparation technique to produce 2-dimensional ultrathin α-Co(OH)2 nanosheets. [87] The technique involved mixing cobalt salt aqueous solution with a methanolic solution of 2-methylimidazole at room temperature. The products synthesized were of nanosheets form that were micrometer in size and possessed an average thickness of approximately 2.5 nm. The ultrathin structure provided the α-Co(OH)2 nanosheet with the ability to perform greatly for OER. An overpotential value 0.267 V was displayed by the resulting material at j = 10 mA cm−2. With this strategy it is also possible to synthesize other 2-dimensional cobalt-based layered double or triple hydroxides. Furthermore, the α-Co(OH)2 nanosheets demonstrated a Tafel slope of 64.9 mV dec−1. This value is comparatively lower than that of commercial RuO2 (78.7 mV dec−1) and hexagonal α-Co(OH)2 plates (81.2 mV dec−1). Quentin Daniel et al. established that when cobalt porphyrins were deposited on FTO glasses (via spin coating), they decompose into thin film of CoOx on the surface during electrochemical water oxidation under borate buffer (pH 9.2, 0.1 M). [88] The thin film was only detectable by XPES using low photon energies (1000 eV). The newly formed catalyst showed advanced activity for OER with a high TOF value of the order of 10 s−1. Huiling Sun et al. reports the synthesis of four Cobalt corroles attached with different acid/base pendants in neutral aqueous solution. [89] The working electrode used was catalyst loaded on FTO glass. Complex LCH [2]PO(OH) [2] –Co showcased higher performance for both OER and HER compared to other complexes. The LCH [2]PO(OH)2 –Co complex showed an overpotential of 0.45 V. Samaneh Sohrabi and team worked on a composite 3D porous coordination network (PCN) with 3D nanochannels via solvothermal method with the help of Zr6 clusters and tetrakis (4-carboxyphenyl) porphyrin cobalt (MOF). [90] Because of the presence of porphyrins and ultrastable Zr6 clusters on the backbone of the MOF, it was an easy access for the reactants. The catalyst developed showcased an overpotential of 0.4 V.Bangan Lu and co-workers synthesized a nanowire array of nickel and cobalt oxides freely standing on nickel foam substrate (NixCo3-xO4-1:1). [91] They reported that Ni doping on the Co3O4 increased the roughness and in turn increased the activity of the catalyst. The Ni doped Co3O4 (Ni:Co = 1:1) and pure Co3O4 had the same nanowire array structure when SEM image analysis was conducted. When the atomic ratio of Ni:Co was larger than 1:1, a change in morphology into nano flake structure was seen. Electrochemical studies showed an overpotential value at 5 mA cm−2 mg−1 of 0.56 V and 0.65 V for NixCo3-xO4-1:1 and Co3O4 respectively. At 0.6 V the current density of NixCo3-xO4-1:1 was about 5 times in magnitude than that of pure Co3O4. There was also a 5-fold difference in the roughness factor between NixCo3-xO4-1:1 and Co3O4, which proves that an increase in roughness increases the activity of the catalyst. Ni doping is assumed to increase the electrocatalytic activity of Co3O4, either by increasing its roughness factor and surface area (geometrical effect) or by enhancing its conductivity (electronic effect) or both. [92] The stability test studied showed a negligible change in potential after 10 h in an alkaline medium. Small Tafel slopes, lower overpotential, and high current density are due to the presence of large active electrochemical surface area (ECSA) as well as 1-D morphology for better charge conduction. Siwen Li et al. had a similar approach to develop a Co-Ni based 1-D nanotube with adjustable Co:Ni ratios by a cation-exchange method to build hydroxides (CoNi(OH)x) grown on a conductive Cu substrate for OER. [93] The nanotube morphology of the catalyst aided in forming a conductive structure with a large surface area, and therefore producing ample catalytic reaction sites. It is also reported that the Co2+ at octahedral sites (Oh) gives a better result for OER than Co2+ at tetrahedral sites (Th). [94–95] When X-ray absorption spectroscopy (XAS) was carried out, a peak at 781.1 eV in the Co l-edge XAS spectra correlates to the characteristic peak of Co2+ ions at Oh. [96] The LSV showed an onset potential value of 1.48 V (vs. RHE) and an η as low as 0.280 V at 10 mA cm−2 with a Tafel slope of ≈ 77 mV dec−1. EIS analysis revealed that CoNi(OH)x nanotube retains a considerably lower charge transfer resistivity which is due to the reaction between the O* intermediate and different hydroxides on the surface of the catalyst. [97] Xuehui Gao et al. synthesized hierarchical NiCo2O4 hollow micro cuboids constructed by 1-D porous nanowire subunits. [98] It exhibited a small onset potential of 1.46 V (vs. RHE) to reach 1 mA cm−2, 1.52 V (vs. RHE) at 10 mA cm−2, and a Tafel slope of 53 mV dec−1. OER activity of NiCo2O4 is accredited to its unique hollow mesoporous structure composed of 1-D nanowires, which provides easy access for electrolytes to the active sites. Substituting a second metal into monometallic phosphides could efficiently alter the electronic structure of the parent compounds and further enhance the OER activity. Lei Han et al. prepared Ni-Co mixed oxide nanocages from Ni-Co Prussian blue analog (PBA) cubes metal–organic framework (MOF) precursors through an anisotropic chemical etching route. [99] Due to their complex 3-D cage-like hollow structure in addition to the high surface area to volume ratio, they exhibited low overpotential (0.380 V@10 mA cm−2) and Tafel slope (50 mV dec−1) under basic medium. Wook Ahn and coworkers fabricated a multivoid nanocuboidal MOF catalyst with multiple mesosized and microsized pores synthesized from a ternary Ni-Co-Fe MOF (NCF-MOF) by a facile co-precipitation and post heat treatment method. [100] Altering ion exchange rates of the transition metals in the MOF are used to produce heteroatom doping, interconnected internal voids, and favorably tuned electronic structure by combining the outer electrons of active Co and Fe metal ions, which leads to reduced adsorption strength with the intermediates. All these aids in bringing about enhanced activity of OER with low overpotential (η10 = 0.320 V) and Tafel slope (49 mV dec−1) for the catalyst. Bocheng Qiu et al. reported Ni-Co bimetallic phosphide nanocages (NiCoP) with constant dispersion of Ni and Co atoms by using Cu2O cubes as sacrificial templates. [101] Ni0.6Co1.4P nanocages derived from Ni0.6Co1.4(OH)2 nanocages exhibited notable activity towards OER (η10 = 0.3 V, 80 mV dec−1) when compared to Ni2P(η10 = 0.420 V, 128 mV dec−1) and CoP (η10 = 0.370 V, 100 mV dec−1). The stability test at 1.53 V for 10 h proved that Ni0.6Co1.4P had the least current density loss (10%) compared to CoP(20%) and Ni2P(30%) nanocages. The authors elucidated that appropriate doping of Co atoms can significantly lower the activation barrier of the catalyst and increase the density of states (DOS) at the Fermi level, resulting in low intermediate adsorption energy and high charge carrier density. Enlai Hu et al. presented a template-assisted strategy to organize 2-D nanosheets of Ni-Co precursors into an oriented stacking of 3-D anisotropic Ag2WO4 cuboid particles. [102] After successive heating, etching, and phosphorization treatments, Ni-Co precursors are converted to open and hierarchical Ni-Co–P hollow nano bricks (HNBs). Overpotential value of 0.270 V to achieve 10 mA cm−2 current density and a Tafel slope of 76 mV dec−1 was observed. The extended stability was tested by a CA measurement and only about 6.5% of the initial current was lost in 20 h time period. Micropores and mesopores among the oriented stacking and macropores due to the open and hollow interior promote exposure of active sites as well as penetration of electrolytes into the catalyst which further eases the OER activity. [98,103] Xin Liang and coworkers formed Ni2P-CoP bimetallic phosphides via low-temperature phosphorization of Ni-Co organic frameworks. [104] Enhanced catalytic activity was achieved by controlled formation of interfaces of Ni2P-CoP, which reduced the bandgap and promoted faster electron transport. LSV curves showed onset potential (1.50 V (vs. RHE)) and overpotential (η10 = 0.320 V) to be lower than Ni2P and CoP. Jiayuan Li et al. came up with a facile synthesis of single-phase ternary Ni2-xCoxP (x ≤ 1.0) rGO hybrids with well-regulated Co doping concentration. [105] It is noted that the presence of rGO increases the number of surface-active sites and enhances the hybrid electrodes’ activity. Co doping controls the active sites’ catalytic activity and accelerates the charge transfer process of the catalyst. Onset potential of 0.251 V (vs. RHE), η of 0.270 V at 10 mA cm−2, and a small Tafel slope of 65.7 mV dec−1 were reported for the NiCoP/rGO hybrids (x = 1) in an electrolyte of 1.0 M KOH. Long-term catalytic stability proved stable OER current density of 50 mA cm−2 at 0.360 V overpotential for 18 h. Hanfeng Liang and coworkers fabricated ternary a NiCoP nanostructure from hydrothermally formed NiCo hydroxides via PH3 plasma-assisted approach, supported on nickel foam, for the first time (NiCoP/NF). [106] The plasma-assisted process promoted low-temperature reaction and fast preparation of the catalyst. From energy-dispersive X-ray spectroscopy (EDS) mapping, they found the ratio of Ni:Co:P to be 1.106:1:1.138, which was close to NiCoP. Electrocatalytic studies measured an η of 0.28 V to obtain a current density of 10 mA cm−2, which was lower than Ni2P/NF (0.34 V) and NiCo-OH/NF (0.404 V). They attributed the enhanced OER activity to the Co addition which lowered the activation barrier, altered the electronic structure, and synergistic effect between Ni and Co. Junyuan Xu et al. worked on tri-metallic equimolar FeCoNiP on carbon nanofiber (CNF) pre-catalyst prepared by chemical reduction followed by phosphorization treatment. [107] Overpotential as small as 0.2 V at 10 mA cm−2 current density and a high TOF of 0.94 s−1 at an η of 0.35 V was measured from the LSV data under alkaline medium. Also, a greater mass activity of 5000 mA mg−1 was obtained at η = 0.330 V. CNFs aided in the improved charge transfer and increased the nucleation sites during wet chemical reduction. The authors interpreted that Co helps in reducing the overpotential in the low potential region and Ni boosts the anodic current in the high potential region. Jingchao Zhang et al. fabricated mesoporous Ni-Co sulfide nanotubes via template-free solvothermal method followed by anion-exchange process. [108] Due to the synergetic effect between Ni and CO, altered electronic structure, and increased surface area, Ni0.13Co0.87S1.097 nanotube exhibited improved performance for OER with lower onset potential (0.262 V at 1 mA cm−2), overpotential (0.316 V at 10 mA cm−2) and Tafel slope (54.72 mV dec−1) in comparison to CoS1.097 (η1 = 0.280 V, η10 = 0.331 V and Tafel slope = 55.54 mV dec−1). A 3-D structure promotes OER activity by contributing to the high specific surface area, more defects as exposed active sites, accelerated H2O adsorption, and easy gas permeability. Chengzhou Zhu et al. designed a 3D bimetallic Ni-Co oxide hollow nanosponges (HNS) by a sodium borohydride reduction strategy (Ni-Co2-O HNS). [109] Due to the hollow structure, synergetic effect between Ni-Co and high specific surface area, the catalyst exhibited a superior OER activity. Ni-Co2-O HNS had a porous and interconnected network and an ultra-low density of around 0.08 g cm−3. LSV curve showed an onset potential of as low as 1.501 V (vs. RHE) and an η of 0.362 V in 0.1 M O2 saturated KOH solution. In a different work, Seok-Hu Bae and co-workers shaped a 3D conductive carbon-shelled Ni-Co nanowire structure (CCS Ni-Co NWs) (Fig. 5 ). [110] The Ni-Co nanowires grown on the carbon fiber woven fabric (hydrothermal method) were coated with conductive carbon shell via glucose carbonization followed by annealing processes. The granular and porous structure of Ni-Co nanowires aids in the rapid release of O2 and provides an enlargement in the number active of sites. Whereas the carbon shell aids in fast electron transmission from the active site to the current collector (carbon fiber fabric) and avoids the dispersion of catalytic particles during active O2 evolution. These properties of the structure helps in enhancing the catalyst’s OER activity. A 0.302 V overpotential @10 mA cm−2 with a Tafel slope of 43.6 mV dec−1 in KOH solution of 1 M concentration was reported. When compared with Ni-Co NWs, the charge transfer resistance of CCS Ni-Co NWs was lower, implying increased current access and enhanced charge transport efficiency. [111] Another notable merit of this catalyst is that it acts as a catalytic electrode, which can be deposited directly on the working electrode without any binders. Cheng Du et al. reported a continuous hydrothermal, oxidation, and phosphidation process using NaH2PO2 to synthesize a 3-D nest-like ternary NiCoP supported on carbon cloth (CC) electrocatalyst. [112] In alkaline medium, η of 0.242 V @ 10 mA cm−2 with a Tafel slope of 64.2 mV dec−1 was reported. They concluded that the addition of urea, carbon cloth, and the coexistence of Ni and Co precursors to be the main reason for the 3-D nest-like structure. Furthermore, the carbon cloth acts as a current collector to improve the conductivity and charge transferability. [105] Linzhou Zhuang and coworkers reported Fe-Co oxide nanosheet (FexCoy-ONS, x/y indicates the molar ratio of Fe/Co) synthesized by a solution reduction process using NaBH4 reducing agent to improve oxygen vacancies and the catalyst’s active sites. [113] The optimized Fe1Co1-ONS had a high specific surface area of about 261 m2 g−1 which resulted in an η of only 0.308 V@10 mA cm−2 and Tafel slope as low as 36.8 mV dec−1 in an alkaline solution (0.1 M KOH). The results obtained were superior to those of commercial RuO2. At η = 0.350 V, Fe1Co1-ONS showed a current density of 54.9 A g−1, which is 5.8 times larger than that of RuO2 available commercially. A detailed characterization using X-ray photoelectron spectroscopy (XPS) and Photoluminescence spectroscopy confirmed that the excellent OER performance of the catalyst was due to abundant oxygen vacancies which result in an easy excitation of the delocalized electrons into the conduction band near the oxygen-deficient sites. Jin-Xian Feng et al. designed a FeOOH sandwiched cobalt hybrid nanotube arrays supported on nickel foam (FeOOH/Co/FeOOH HNTAs-NF). [114] Co and FeOOH were loaded as 0.28 mg cm−2 and 0.22 mg cm−2, respectively. EIS studies confirmed that FeOOH/Co/FeOOH HNTAs-NF has a significantly smaller electronic resistance when compared to FeOOH NTAs-NF. This result confirms that the Co metal layer enhances electron transmission because of its high electrical conductivity and Ni foam acts as a current collector which together overcomes the poor electric conductivity of FeOOH. The optimum thickness of 25 nm for FeOOH showed the highest OER activity.FeOOH/Co/FeOOH HNTAs-NF showed an overpotential of just 0.250 V (Fig. 5) to reach 20 mA cm−2 current density and a Tafel slope as low as 32 mV dec−1 in an alkaline medium. Chronopotentiometric studies performed for 50 h showed no negligible change in overpotential to maintain the current density values of 20, 50, 100, and 200 mA cm−2. Theoretical studies suggest that the energy of OER intermediates can be modulated with the inclusion of metal elements for a given metal oxide. Bo Zhang et al. fabricated a gelled FeCoW (oxy)hydroxides (G-FeCoW) using a sol–gel procedure that would include incorporation of W6+ into FeCo (oxy) hydroxides, which is hydrolyzed at a controlled rate to achieve atomic homogeneity. [115]) Here, tungsten (W) modulated 3d metal oxide (CoOOH), provided excellent adsorption energies for OER intermediates which in turn enhanced catalytic activity. When the G-FeCoW catalyst underwent electrocatalytic studies, it exhibited surprising results compared to the FeCo LDH and NiFe LDH. It presented an η of 0.191 V at 10 mA cm−2 current density when deposited on a gold-plated Ni foam. This is significantly lower than the precious metal-based electrocatalyst used for OER previously. The stability test showed no appreciable increase in potential under 30 mA cm−2 current density for 550 h. When the catalytic measurements were conducted on glassy carbon (GC) electrode, the catalyst showed an η of 0.223 V at 10 mA cm−2 current density, a TOF of 0.46 s−1 and mass activity of 1175 A g−1. Similar work has been reported by Peng Fei Liu et al. where they used molybdenum (Mo6+) to modulate 3d metal (oxy) hydroxides (FeCoMo) to attain better adsorption energy for the OER intermediates and provide rich active sites for OER. [116] FeCoMo displayed an η of 0.277 V at 10 mA cm−2 current density on GC and no evidence of degradation was reported for about 40 h at constant 10 mA cm−2 current density. It exhibited a bulk mass activity of 177.35 A g−1 at 0.3 V overpotential, which is approximately 7 times larger in comparison to IrO2. Harshad A. Bandal and coworkers prepared a composite electrode of high activity for water splitting by placing ordered spinel Fe-Co oxide (50 nm thickness) on the surface of Ni foam (FeCoO-NF). [117] When compared to CoO-NF (η10 = 0.268 V), FeCoO-NF (η10 = 0.244 V) revealed advanced performance for OER activity. Tafel slope of FeCoO-NF (57 mV dec−1) was also lower than the compared CoO-NF (67 mV dec−1), which indicates that the incorporation of iron into Co3O4 has a positive effect on enhancing the OER activity. When compared to conventional RuO2, FeCoO-NF required comparatively less overpotential to reach the 50 and 100 mA cm−2 current density marks. The 3-D white fungus-like structure of FeCoO-NF aided in the effective transport of electrons between the catalyst and electrolyte, easy dissipation of O2, and reduced solution and charge transfer resistance. Wei Liu et al. fabricated an amorphous Co-Fe hydroxide (CoFe-OH) nanosheets (20–30 nm thickness) via facile electrodeposition for 20 min grown homogeneously on a graphite substrate’s surface. [118] Due to their hierarchical network formed by the nanosheets, it resulted in a high electrochemically active surface area which exhibited a low η (0.280 V at 10 mA cm−2) and low Tafel slope (28 mV dec−1when compared to Fe-OH and Co-OH samples in alkaline medium. In a different work; Hui Xu, Jingjing Wei, and coworkers created a 2-D CoFe oxyhydroxide nanosheet doped with phosphorous (2D-CoFeP NS) in alkaline medium (1 M KOH), which delivered 0.305 V overpotential at 10 mA cm−2 current density with a low Tafel slope of 49.6 mV dec−1. [119] 2D-CoFeP NSs electrode maintained long term stability with a negligible decrease in potential at the 10 mA cm−2 current density for 24 h. The doped phosphorus played a critical part in modifying the surface-active sites of the catalyst. To add to the phosphorization, and synergetic effect between Co and Fe, the unique structure provides a great surface area and ample interlinked channels for O2 release and mass transport. In a similar work, Xiao Zhang and team introduced a novel CoFeP multi-void nanocages (CoFeP-NC) that were derived from CoFe-PBA (Prussian blue analog) nanocubes via a self-template phosphorization process with uniform size ranging from 250 nm to 350 nm. [120] From the XPS spectra, it was clear that both Co and Fe acted as active sites in the catalyst, and the synergism induced between them improved the electronic structure as well. In alkaline medium, CoFeP-NC showed an η as low as 0.180 V at 10 mA cm−2 current density, superior stability, and a turnover frequency of 0.93 s−1 at 0.270 V overpotential. This enhanced OER activity could be explained by the porous hollow system with large surface area, high effective active sites, and reduced charge transfer distance. The pyridinic N doped in the CoFeP catalyst provided an added synergistic effect to OER activity. The catalyst exhibited a fast-current density increase within a small change in overpotential (η100 = 0.280 V). Yuan et al. developed a hierarchical hollow nanocube structure that was based on ultrathin CoFe-layered double hydroxide (CoFe-LDH) nanosheets. [121] The group first prepared Cu2O nanocubes as the self-sacrificing template. They employed a template-assisted route for the production of hollow nanocubes based on CoFe-LDH nanosheets through coordinating etching. The performance of the resulting material was demonstrated when it displayed a low overpotential of 0.270 V for 10 mA cm−2 current density for water oxidation. A low Tafel slope value of 58.3 mV dec−1 as well as a long-term stability was also displayed in an aqueous solution of 1 M KOH. DFT study was also performed by the research group and the analysis revealed that Fe addition provided a metallic identity with Co(OH)2, assisting in electron transfer. Qian Zhou et al. reported a facile cation-exchange process for creating iron-doped Co(OH)2 nanosheets with the augmented active site. [122] Iron-doped Co(OH)2 nanosheets showed lower Tafel slope and overpotential (53 mV dec−1, η10 = 0.320 V) when compared to pristine Co(OH)2 nanosheets (69 mV dec−1, η10 = 0.370 V). After the cation exchange process, Fe3+/Co2+, the Fe-doped Co(OH)2 nanosheets had substantial grain boundaries, rougher surface, improved hydrophilicity, and enhanced electronic properties which resulted in an enhanced activity for OER. In recent work, Lei Zhong and coworkers produced Fe doped CoTe (Fe-CoTe) by a one-step solvothermal process, which showed excellent activity and stability without any activation process. [123] A 0.300 V overpotential at 10 mA cm−2 current density and 45 mV dec−1 Tafel slope value were observed from the electrocatalytic measurements. The authors elucidated that Fe-CoTe had the maximum Fe-Co synergy and the catalytic performance was due to intrinsic properties of Fe-CoTe, and not from the Fe impurity adsorbed from the electrolyte. The improved amount of lattice oxygen also aided to the enhanced OER activity. The charge transfer resistance of Fe-CoTe was only 1/6th of the pristine CoTe catalyst, representing the Fe-doping effect. Sheng-Hua Ye et al. fabricated Fe substituted CoOOH porous nanosheets arrays developed on a cloth of carbon fiber (FexCo1-xOOH PNSAs/CFC, 0 ≤ x ≤ 0.33 ) with 3-D structures via in-situ anodic oxidation of α-Co(OH)2 NSAs/CFC. [124] Fe0.33Co.0.67OOH PNSAs/CFC showed enhanced activity towards OER with a low η10 of 0.266 V and a Tafel slope value of 30 mV dec−1. X-ray absorption fine spectra (XAFS) studies indicated a partial substitution of CoO6 octahedral structures in CoOOH by FeO6 octahedral during the conversion from α-Co(OH)2 to FexCo1-xOOH. Detailed DFT calculations indicated that such substitution can reduce the energy levels of the intermediates and products as FeO6 octahedron is a highly active site for OER. Li-Ming Cao and team proposed a concrete pathway for the hierarchical fabrication of a novel self-supporting 3-D porous sulphur-doped NiCoFe LDH nanosheets (S-NiCoFe LDH) on carbon cloth. [125] The EIS measurements revealed that the charge transfer resistance (RCT) value of S-NiCoFe LDH was smaller than those of undoped LDH (NiCoFe LDH and NiFe LDH), which indicates that sulphur doping helped in improving the catalyst’s electrical conductivity. A low η of 0.206 V at 10 mA cm−2 current density as well as a Tafel slope of 46 mV dec−1 was reported from the electrocatalytic measurements. The XPS results supported that the Co-S bonds and Ni-S bonds were altered into Ni/Co oxyhydroxides that further enhanced the OER activity.Jingrui Han et al. developed an amorphous Mn-Co-P layer on MnCo2O4 supported on a titanium mesh (Mn-Co-P@MnCo2O4/Ti) through a cathodic polarization in NaPO2H2 solution. [126] Under alkaline medium, the catalyst demonstrated an η of 0.269 V at a current density of 10 mA cm−2, Tafel slope of 102 mV dec−1 which was lower compared to MnCo2O4/Ti (η10 = 0.362 V, 210 mV dec−1). XPS results revealed that the Mn-Co-P layer was produced on the surface of the MnCo2O4 as a shell that boosts the OER activity. Xijun Liu et al. reported hierarchial ZnxCo3-xO4 nanoarrays which had secondary nanoneedles grown on primary rhombus-shaped pillar arrays supported on titanium (Ti) foil prepared by the co-deposition of zinc and cobalt precursors followed by calcination in air. [127] A 0.320 V overpotential @ 10 mA cm−2 was detected for ZnxCo3-xO4-1:3 (1:3 is the ratio of Zn and Co precursor used) nanoarrays with a low Tafel slope value of 51 mV dec−1. ZnxCo3-xO4 can be directly used as electrodes for OER. [128] The close contact of the 3-D porous structure to Ti foil ensured long term stability and gave way for the conduction of electrons. [127,129] The enhanced performance of ZnxCo3-xO4 is attributed to the unique hierarchical 3-D nanostructure which brings about high porosity, large surface area, more active sites, increased roughness factor, and improved gas permeability. Jianfeng Ping and team prepared a 3-D porous CoAl-layered double hydroxide (LDH) nanosheets (CoAl-NS) onto a 3D graphene network (3DGN) by electrostatic self-assembly (3DGN/CoAl-NS). [130] Here to obtain the CoAl-NSs, the CoAl-LDH (NO3−) crystal with the largest interlayer area was made use for exfoliation. The electrochemical activity was studied in 1 M KOH with a loading mass of CoAl-NSs on 3DGN as about 0.05 ± 0.01 mg cm−2. The results revealed an η at 10 mA cm−2 current density to be 0.252 V and a 36 mV dec−1 Tafel slope value. At η = 0.300 and 0.350 V, the current densities values were 45.37 and 91.74 mA cm−2, respectively. Stability tests confirmed a nearly constant current density for 18 h at η = 0.250 and 0.280 V. The exposed active edge sites of the CoAl-NSs made it easy for the proton paired electron transfer process during OER. [131–132] Also a constant coating of single layer CoAl-NSs on the 3DGN by electrostatic self-assembly is an effective way by which it can expedite the reaction kinetics and accelerate the electron transfer. [133–134] Furthermore, the unique structure of 3DGN aids for the access of ions to the catalysts and prevents the restacking of CoAl-NSs [135–137].Perovskites exhibit an ABO3 type of empirical formula with A generally being a rare-earth or alkaline earth metal while B being commonly a transition metal. It has been stated that doping of A- or B- site cations in the perovskites structures is an effective way to improve OER activity. [138–139].Denis Kuznetsov et al. introduced a high electronegative (Bi3+) element into the A-site of the strontium cobalt perovskites to sustain high Co-O covalency through the inductive effect. [140] An exceptionally low Tafel slope of 25 mV dec−1 was obtained for the bismuth substituted strontium cobalt perovskites. This was attributed to the potentially increased hydroxide kinship on the catalyst’s surface by the introduction of Bi3+ ions. Xi Cheng et al. analyzed the influence of Sr substitution into the A- site of LaCoO3 perovskites. [141] The surface composition, bulk electronic structure, electrochemical activity, and conductivity for the La1-xSrxCoO3 perovskite series ( 0 ≤ x ≤ 1.0 ) were investigated experimentally and theoretically. A phase transition from rhombohedral (LaCoO3) to cubic structure (La1-xSrxCoO3) was observed after the gradual replacement of La by Sr. They found that Sr substitution has the effect of aligning along the Co-O-Co axis, straightening the octahedral cage, and rising the average oxidation state of Co ions. DFT calculations proved that the above merits improve the overlap between the unoccupied Co 3d conduction bands and the occupied O 2p valence bands which further improved the catalyst’s OER activity.To understand the influence of B-site substitution in perovskites, Maria A. Abreu-Sepulveda et al. investigated an organized substitution of Co by Fe in La0.6Ca0.4CoO3 perovskites (La0.6Ca0.4Co1-xFexO3) under alkaline medium via a facile glycine-nitrate synthesis. [142] A rise in the surface concentration of different Co oxidation states by the incorporation of Fe was showcased by the XPS results. Specific activity trend of the substitution followed the trend: Fe0.9 > Fe0.8 > Fe0 > Fe0.1 > Fe0.2 > Fe0.5 > Fe1.0. Iron incorporation decreased the barrier for electron transfer and facilitated the generation of cobalt-hydroxides. A Tafel slope value of 49 mV dec−1 was determined for La0.6Ca0.4Co0.1Fe0.9O3 (x = 0.9). They found that Fe complexes are significant for OER by enabling the interaction of Co-OH bond and CoOOH are responsible for the electronic conductivity. Under alkaline medium layered double hydroxide perovskites PrBaCo2O6- δ (PBC) has been found to be very active. The addition of Fe further enhances the activity. Xiaomin Xu et al. reported BaCo0.9-xFexSn0.1O3- δ (BCFSn) perovskites oxides through doping Fe and Sn in BaCoO3-σ parent oxide via solid-state reaction under alkaline medium. [143] BCSFsn-721 (x = 0.2) displayed a low value of onset potential ( ≈ 1.53 V vs. RHE), overpotential ( ≈ 0.420 V at 10 mA cm−2 current density) and a Tafel slope value of 69 mV dec−1. They established that the catalyst’s OER activity can be tuned by simply altering the concentration of Fe and Sn. The mass activity of the catalyst can be further enriched by reducing their particle size or creating pore structures with a large surface area. [144] Bae-Jung Kim et al. doped iron into the B-site of PBC with different ratios to synthesize PrBaCo2(1-x)Fe2xO6-δ (x is 0.2 or 0.5; designated as PBCF82 and PBCF55, respectively) nanoparticles in the size range of 5–30 nm. [145] PBCF82 and PBCF55 exhibited the same Tafel slope value of 50 mV dec [1] which was lower than PBC (72 mV dec−1) and greater current densities at 1.55 V vs. RHE (17.1 and 19.7 A g−1) (Fig. 6 , A-E). When stability tests were conducted, PBCF55 lost only 32% of its starting current density, while PBC lost approximately 74% of its initial current density. They elucidated from their studies that Fe incorporation stabilizes cobalt in the lower oxidation state by delivering finer distribution of charge, encouraging the emergence of oxygen vacancies, and improving the structural stability of the layered double perovskites catalyst by supporting the formation of oxy(hydroxide) layer. Yinlong Zhu and coworkers fabricated SrNb0.1Co0.7Fe0.2O3- δ (SNCF) under alkaline medium. [146] SNCF was ball milled to increase its surface area, which further increases the OER activity. Advanced OER ability with low onset potential (1.49 V vs. RHE), overpotential (η10 = 0.420 V), and Tafel slope of 76 mV dec−1 was observed due to excellent ionic and charge-transfer capabilities along with optimized eg orbital filling, and high O2 desorption and OH− adsorption abilities. It also exhibited good stability for the long term due to the incorporation of Nb5+ cations on the B-site of the catalyst. The catalytic performance of conventional Ba0.5Sr0.5Co0.8Fe0.2O3− δ perovskites (BSCF) is limited by a low specific surface area (0.5 m2 g−1). Yisu Yang et al. developed porous BSCF perovskites with ordered pore structure (3–10 nm) via a novel in-situ tetraethoxysilane (TEOS) template technique to increase the specific surface area (reaching a value of 32.1 m2 g−1) of the conventional BSCF perovskites. [147] This method increased the specific surface area of the nonporous BSCF by 60 times. An optimum ratio of 3.4 for TEOS to BSCF was found to have the highest performance for OER. Under alkaline conditions, BSCF-3.4 exhibited the highest current density of 35 A g−1 at 1.63 V vs. RHE (η of 0.4 V) (Fig. 6, f-j), which was 5.3 times higher than nonporous BSCF (6.6 A g−1), and a Tafel slope value of 62 mV dec−1. They concluded from their studies that silica-containing impurities reduce the conductivity of the electrodes and the enhanced activity was strictly related to the microstructural properties of the catalyst.Chao Su et al. synthesized perovskites oxides with the composition of SrM0.9Ti0.1O3-δ (M = Co, Fe) via the sol–gel method. [148] SrCo0.9Ti0.1O3-δ (SCT) showed better functioning stability in comparison to SrFe0.9Ti0.1O3-δ (SFT), BSCF, and IrO2. Such OER activity could be accredited to the low average bond energy of Co-O, optimal eg electron filling, and good charge transferability. Xiaoming Ge et al. designed a novel La(Co0.55Mn0.45)0.99O3-δ (LCMO) nanorods (diameter of 45–55 nm and aspect ratio of 3–10) using a hydrothermal method followed by heat treatment. [149] The 1% B-site lattice vacancy offers an added advantage for good OER activity of oxides. [150] The further synergetic covalent coupling that exists between LCMO and reduced graphene oxide doped with nitrogen (NrGO) exhibited exceptional bifunctional activity for OER and ORR. The OER onset potential of LCMO/NrGO was about 0.45 V vs RHE and the potential to reach 10 mA cm−2 was 0.787 V vs. RHE, which was lower compared to Ir/C. The coupling between the NrGO and LCMO, NrGO’s permeating electrical conduction, and the intrinsic activity of LCMO perovskites resulted in the advanced OER activity. Anchu Ashok et al. investigated on lanthanum based electrocatalytically active LaMO3 (M = Cr, Mn, Fe, Co, Ni) perovskites produced through a single-step solution combustion method. [151] Results from the study showed enhanced OER for LaCoO3 that is because of the optimum stabilization of reaction intermediates present in the RDS of OER. The stability test proved LaCoO3 to be the most stable among the perovskites studied in the report. Taking inspiration from a previous work by Mohamed A. Ghanem [152] on the mixed anion perovskites (ABOxXy, X is a non-oxygen anion), Yuto Miyahara and co-workers studied the bi-functionality, for OER and ORR, of layered cobalt perovskite oxychlorides, namely Sr2CoO3Cl and Sr2Co2O5Cl2, synthesized through a solid-state reaction that utilizes Sr2Co2O5 as a precursor. [153] The catalyst was found to be highly active, which was because of the upshift of the O p-band center compared to the Fermi level caused by the incorporation of Cl− into the oxygen sites. The onset potential of the oxychlorides was exhibited to outperform the state-of-the-art BSCF perovskites. Tafel slope also reveals the same outcome about the activity of oxychlorides (60 and 62 mV dec−1, respectively) when compared to BSCF perovskites (72 mV dec−1).Over the past few decades, carbon materials have received substantial attention as a support in various electrocatalyst due to their high thermal stability, environmental friendliness, good conductivity, chemical inertness, high specific surface area, corrosion-resistant, tunable surface function, and higher stability in both acidic and alkaline medium. [154–155] Based on the crystal structure, carbon atoms can be of various allotrope forms with distinct and unique physical and chemical properties. Various carbonaceous nanomaterials such as carbon nanofiber (CNF), carbon nano coil (CNC), nano carbon black (CB), single/multi-walled carbon nanotubes (SWCNT/MWCNTs), carbon mesoporous (CMS) and graphene/graphene oxides (G/GOs) are possibly incorporated with the cobalt-based catalyst in order to enhance the electronic conductivity and electrochemical performance. [156] Here we discuss various cobalt-based catalysts supported on carbon materials towards oxygen evolution reaction.Carbon nanotubes (CNTs) have received significant attention in the area of fuel cells as effective support because of the large surface area, high electronic conductivity, thermal stability, and durability that they offer. CNTs are rolled-up sheets of single (SWCNTs) or multi-layer (MWCNTs) carbon atoms (graphene) in cylindrical form.Lu and Zaho prepared crystalline cobalt oxide nanoparticle of ~6 nm size incorporated with mildly oxidized multiwalled carbon nanotubes (Co3O4/mMWCNT) and used it as an effective catalyst for H2O oxidation. They studied the correlation between various other carbon structures such as single-walled CNTs (SWCNTs), graphene, and multi-walled CNTs (MWCNTs) with different oxidation states in terms of charge transport and surface functionalization towards water oxidation reaction. The results showed that the hybrid mildly oxidized MWCNTs (Co3O4/mMWCNT) with a 0.390 V overpotential value (at 10 mA cm−2) and 1.51 V vs. RHE onset potential that can sustain the electrochemical reaction even under harsh environment with minimum carbon corrosion acting as a promising electrocatalyst towards OER [157]. Zeng et al. reported the bifunctional cobalt (II/III) oxides strongly anchored onto a lightweight,conductive, and crosslinked aerogel film of carbon nanotubes (CNTs) as a free-standing air electrode. The LSV profile showed improved performance of crosslinked aerogel film of carbon nanotubes (CNTs) when compared with pristine CNT aerogel, N‐CNT aerogel, and pure Co3O4 in terms of onset potential (1.45 V) and potential (1.7 V vs RHE at a current density of 10 mA cm−2) (overpotential of 0.47 V) [158]. Shuo and coworkers followed the pyrolysis of metal–organic framework (MOF) encapsulated Co3O4 for the successful generation of Co-embedded N-doped CNTs with porous carbon (PC) that showed prolonged stability and excellent activity in alkaline solution. The polarization curve for Co–CNT/PC exhibited lower onset potential than Co-doped over porous carbon (Co-PC) that could be attributed to the improvement in the electrical conductivity for CNT. [159] Zhang and coworkers reported a superior activity and remarkable stability for cobalt carbonatehydroxidehydrate(CCHH)nanosheets strongly adhered on the mildly oxidized MWCNTs in presence of diethylenetriamine(DETA). They found that the presence of DETA greatly influences the structure and morphology of the CCHH/MWCNT composite and thereby enhanced the resulting OER performance. Thus, prepared hybrid CCHH/MWCNT exhibited lower onset potential (approximately 1.47 V vs RHE) (overpotential of 0.285 V @ 10 mA cm−2) and good kinetics that was clear from the Tafel slope analysis. [160] In 2014, the same group reported a study on Co3O4 nanorod–multi-walled carbon nanotube hybrid (Co3O4@MWCNT) that exhibited 0.309 V overpotential at 10 mA cm−2 current density in an alkaline medium that also possess superior activity and stability. [161] Fang et al. analyzed the synergistic influence of Co(II), organic ligands, and CNTs that offered excellent activity and durability to sustain in a harsh environment without any carbon corrosion. The hierarchical 3-D unique system with a large surface area improved the transportation of electrons and secured the anchoring of the catalyst’s active sites to the CNTs. Co-MOF@CNTs offered impressive durability and activity when compared to 20 wt% Pt/C and RuO2 catalysts. They studied the influence of the OER performance on the amount of CNT in the overall catalyst and found that the overpotential followed an inverted volcano type trend with CNT weight percentage with increasing order of 5% < 1% < 10% <15% of CNTs. [162] In order to improve the surface defect, chemically active sites, and the surface defects of CNTs; hetero atom doping with boron, phosphorous and nitrogen are widely used. Dicobalt phosphides (Co2P) are another category of Co-based catalyst that recently achieved wide attention owing to their catalytic and magnetic properties. Various reports are available on Co2P anchored CNTs delivering excellent electrochemical performance. Hui et al. utilized N, P co-doped CNTs for the anchoring of CoP/CoP2 nanoparticles that exhibited low overpotential, higher current density and excellent stability over 100 h. Moreover, they conducted density functional theory calculations and molecular dynamics simulations that concluded the synergetic effects of CoP and CoP2 improved the electrocatalytic performance; also the heteroatom-doped CNTs readily diffuse out the generated O2 molecule to help in improving the electrocatalytic oxygen evolution reaction. [163] Das and co-workers followed a novel one pot synthesis of phosphine free (PH3) Co2P anchored over N, P dual doped carbon nanotubes without any external carbon additive. The average diameter of prepared Co2P was found to be 55 nm and for NPCNTs the range was between 80 and 250 nm. The hybrid Co2P/NPCNT displayed a small onset potential value of 1.293 V and an η of 0.370 V (at 10 mA cm−2) that was expected for a solar water-splitting device with 10% efficiency. [164] Guo and team utilized a combination of dicobalt phosphide (Co2P)–cobalt nitride (CoN) core–shell nanoparticles synthesized using direct pyrolysis method as double active sites for the incorporation with N doped CNTs that showed excellent trifunctional performance. The interface between CoN and N‐doped CNTs was the active site for OER that has attracted applications in flexible and rechargeable Zn-air batteries. [165] Cobalt sulfides,including CoS2, CoS4, Co3S4, and Co9S8, have been found to be attractive and are novel electrocatalysts for the storage of energy as well as conversion applications because of the unique chemical and physical properties. [166–167] The performance of cobalt sulfideswere improved by further optimization of electrode surface with carbon-based materials. Wang and co-workers prepared an integrated 3-D model of carbon-paper/carbon-tubes/cobalt-sulfide array that displayed impressively high performance towards OER. The unique hybrid structure possibly enhanced the accessibility and availability of active sites, the capacity to transport electron, and improved the release of product gases. They studied the catalyst without any CNT incorporation and found that CP/CTs/Co-S have excellent behavior in terms of potential and current density when compared to CP/Co-SN. [168] Xinwei and his team used atomic layer deposition (ALD)for the successful deposition of the thin layer of Co9S8 (~7 nm) onto the CNT network scaffold with a high surface area that showed remarkably great performance for rechargeable Zn-air batteries [169].The electrical conductivity of oxidized-CNTs is lower than non-oxidized CNTs that limit their catalytic performance, nonetheless, this limitation, in some cases can be overcome by using a mildly oxidized graphene/CNTs. [170] Ting Ma et al. performed an in-situ synthesis of ultra-small Co–Mn–O spinel nanoparticles that have an average nanoparticle size of 4.4 nm reinforced over the non-oxidized CNTs that enabled strong coupling to aid with the transfer of electron and enhanced the activity. CMO@CNTs exhibited a lower onset potential of 2.558 V and a Tafel slope of 81.1 mV dec−1 that showed superior performance when compared with CMO@rGO, and CMO@Vulcan, CMO@oxCNTs, and CMO + CNTs. [171] Fig. 7 shows the synthesis of non-spinel MnCo oxide, Co x Mn1− x O (Mn2+, Co2+)anchored over N-doped CNTs that showed much higher OER activity than commercial IrO2. Co2+ was regarded as the active site for the evolution of oxygen because NCNT/MnO exhibited inferior performance than NCNT/CoO as reported by Liu’s group. [172] Kunpeng and his team followed a two-step gas phase process for the fabrication of hierarchical hybrid Co3O4–MnO2–CNT. The prepared spinel Mn-Co mixed oxide triggered the growth of multi-walled carbon nanotube and the active metal particles remained on the CNT surface were greatly influenced by the growth time. An HNO3 vapor treatment was used to convert the active metals to their higher oxidation state, where MnO was oxidized to MnO2, and Co was converted to Co3O4. Subsequently, a small amount of oxygen functional group was created on the surface of the catalyst that facilitated the release of gases during the reaction. [173] Liu’s group synthesized morphology-controlled La2O3/Co3O4/MnO2–CNTs hybrid nanocomposites which showed excellent durability and activity when compared to the commercial 20% Pt/C catalyst. The oxygen evolution reaction onset potential for La2O3/Co3O4/MnO2–CNTs, CNT, MnO2, La2O3/Co3O4/MnO2, La2O3/Co3O4–CNTs, Co3O4/MnO2–CNTs and 20% Pt/Cwas found to be1.42, 1.69, 1.51, 1.70, 1.52, 1.52 and 1.67 V, respectively. This indicates that the powerful coupling effect that is present between La2O3 nanorod and MnO2 nanotubes, Co3O4 and CNTs produces a synergy for the catalytic performance. [174] Another catalyst, Ni, showed enhanced performance in OER when alloyed with Co to form bimetallic Ni-Co. Many works have been reported on the encapsulation of NiCo with conducting CNTs in order to improve the electronic conductivity. Jie et al. fabricated a 3D network of NiCo encapsulated with nitrogen-doped CNTs (NiCo@NCNTs) as shown in Fig. 7e -hthat showed superior activity than the bimetallic composite (Co@NCNTs and Ni@NCNTs) owing to the synergy between cobalt and nickel. They conducted a detailed study on the effect of coupling on NiCo@NCNTs in comparison with physically mixed NiCo and CNTs. The superiority of encapsulated NiCo@NCNTs caused the transfer of electrons from NiCo alloy to the walls of carbon nanotubes that reduced the local work function on the carbon surface. Also, the wrapping of CNTs over the active NiCo alloy effectively resisted the etching in harsh environment and made the catalyst active and stable for long duration. [175] Yang and co-workers fabricated ultra-small NixCo3−xO4 nanocrystals (~5 nm) decorated over pristine MWCNTs using the solvothermal method without destroying the CNTs. Pristine MWNTs showed an efficient electron transfer network and its incorporation with well-constructed spinel NixCo3−xO4 led to an outstanding electrochemical performance [176]. Numerous cobalt-carbon based catalysts were reported showing extremely high activity and performance due to the hierarchical structure and synergetic effect including Co-N/CNT [177], Co‐NRCNTs [178], Co(OH)x-NCNT [179], Ni foam‐supported N‐CNT@Co3O4 [180], CNTs-Au@Co3O4 [181], CoHCF/CNT [182], Co-CNT/Ti3C2 [183], CoFe/Co8FeS8/CNT. [184] Exceptional properties of graphene, such as great electrical conductivity, large surface area, and fine chemical and mechanical stability has led to extensive research activities, particularly utilizing them as a catalyst substrate. [185] As a result, it has also been implemented in electrical devices such as fuel cells and lithium batteries with enhanced electrochemical operation. [186–190] In terms of the use of graphene as a material for catalyst production, many graphene-based composite catalysts have been synthesized in the recent past. The resulting composites are applied on substrate electrodes, normally, using drop-casting techniques [191–192].A composite of graphene and Co3O4 (G-Co3O4 composite) has been reported by Zhao’s research team, that possesses a unique sandwich-architecture as shown in Fig. 8 a. [193] Analysis of the composite using TEM and FESEM has shown that there is a homogeneous distribution of Co3O4 on the 2 sides of graphene nanosheet (Fig. 8, b–e). A superior catalytic behavior towards OER in an alkaline solution of 1 M KOH has been observed (Fig. 8, f–i). An onset potential of 1.454 V vs. RHE was exhibited by the composite. Furthermore, within the same alkaline solution, the achievement of 10 mA cm−2 current density was observed at an η of 0.313 V that is more superior than that of the mesoporous Co3O4 catalyst (0.525 V) and Co3O4/SWNTs (0.593 V). In terms of stability, the composite is expected to exhibit long-term stability as it demonstrated no clear decay in current density during testing in alkaline solution after 10 h as well as an undisturbed morphology. The extraordinary behavior could be due to the synergistic influence arising from the combination of both Co3O4 and graphene that include swift electron transfer rate, large electroactive surface area, and better chemical and electrical coupling of the composite. In another study performed by Zhao et al., a catalyst of CoO nanoparticles wrapped by porous graphene sheets was synthesized using 1-D silica nanorods as a template to prepare the porous graphene. [194] The catalyst possessed great specific surface area and porosity and showed rapid charge transport kinetics. An improvement in catalytic activity was also seen for OER that includes large current density and a low onset potential. When the performance of the catalyst was studied in a KOH solution of 0.1 M concentration via LSVs, the PGE-CoO hybrid demonstrated a small onset potential of 1.4934 V vs. RHE that is considerably lower than GE-CoO (1.5494 V vs. RHE) and CoO (1.5594 V vs. RHE) itself. Moreover, at 10 mA cm−2 current density, the composite exhibited a low overpotential of 0.348 V. To measure the efficiency of the PGE-CoO catalyst Tafel plots were obtained from the LSVs and a Tafel slope value of 79 mV dec−1 was determined. In comparison to the value obtained for GE-CoO and CoO this value is way smaller (GE-CoO showed 192 mV dec−1 and CoO showed 354 mV dec−1). The improvement in performance could be accredited to the presence of large electroactive surface area, porous structure, and a strong chemical coupling between both CoO NPs and graphene. In addition, the catalyst could maintain fine stability towards OER in an alkaline solution, possibly due to CoO NPs corrosion prevention characteristic introduced by the wrapped structure.Wang’s group produced a series of electrocatalyst in which graphene and cobalt oxide NPs nano-hybrids (Co-N/G) are doped with nitrogen via a one-pot hydrothermal method. [195] A nitrogen precursor is known as 2, 4, 6-Triaminopyrimidine was also utilized to anchor cobalt oxides NPs onto the graphene oxide surface. The composites synthesized consisted of cobalt oxides in the form of Co3O4 and CoO as well as a high content of doped nitrogen (~6 at. %) comprising of pyrrolic, pyridinic, and graphitic types. The resulting synergistic effect generated from the coupling between Co NPs and nitrogen-doped graphene allowed the composite samples to be used as catalysts for both OER and ORR. The as-synthesized Co-N/G 600 (sample carbonized under N2 atmosphere at 600 °C) showed the highest potential for application in reversible electrochemical energy conversion fuel cells and metal-air batteries. This is because Co-N/G 600 demonstrated an excellent bifunctional catalytic activity with high efficiency in which high activities for oxygen evolution reaction and oxygen reduction were observed at a potential of 0.76 V (1.554 V onset potential vs RHE) and −0.2 V (0.855 V onset potential vs RHE), respectively. In addition, the Co-N/G 600 catalyst showed both fine stability and durability for both the type of reactions. Graphene-based materials doped with nitrogen have been used by Hou et al., where nitrogen-doped graphene was combined with a Co-embedded porous carbon polyhedron to form N/Co-doped PCP//NRGO. [196] A simple pyrolysis of graphene oxide (GO) and zeolitic imidazolate-framework (ZIF), ZIF-67, was implemented in the preparation of the new novel hybrid electrocatalyst after which metallic cobalt was partially etched away. The utilization of ZIF-67 was performed to take advantage of the plentiful Co-N moieties and the unique dodecahedral morphologies available with ZIF-67. [197] With these properties, ZIF-67 may be a fitting precursor for the generation of N/Co-doped PCP. The as-synthesized hybrid catalyst showed excellent performance, including great stability, not only for OER but also for ORR and HER. Such enhancement could be associated with the dual-active-site mechanisms that emerge from the synergetic influences between NRGO sheets and PCP doped with N/Co. The hybrid electrocatalyst also demonstrated a four-electron pathway, great durability, and high tolerance towards methanol. Furthermore, during the performance analysis of N/Co-doped PCP//NRGO for oxygen evolution reaction, only a small η value of 1.66 V was noticed at a current density of 10 mA cm−2. Qiao’s group co-doped graphene with both Co and nitrogen and inserted carbon nanospheres into the graphene sheets interlayers. [198] The carbon nanospheres behaved as “spacers” that enlarged the accessible surface area of graphene and provided many electrolyte channels. These two unique properties helped in promoting the diffusion of reaction species to the active sites. Enhanced conductivity could also be guaranteed as the carbon nanospheres could further act as “shortcuts” for interplanar electron transport. The synthesized catalyst possessed bifunctional stability and catalytic activity for both ORR and OER in a basic medium. When compared to Pt/C catalysts the overall oxygen electrode activity parameter (ΔE) of the bifunctional Co-N-GCI electrocatalyst was relatively lower (0.807 V). The overpotential for Co-N-GCI catalyst at 10 mA cm−2 current density was determined to be 0.426 V, that was much lower than those obtained for Co-N-G hybrid (not intercalated with conductive carbon nanospheres, 0.472 V), and commercial Pt/C (0.621 V), at the same current density. Moreover, an excellent intrinsic OER kinetic of Co-N-GCI was confirmed by a relatively lower Tafel slope value of about 69 mV dec−1 in comparison to Co-N-G (~78 mV dec−1), IrO2/C (~83 mV dec−1) and Pt/C (~168 mV dec−1). A strongly coupled hybrid electrocatalyst of CoOx NPs grown on B, N-decorated graphene (CoOx NPs/BNG) was produced by Tong et al. that is suitable for catalyzing both ORR and OER. [199] An abundant presence of oxygen vacancies and strong CoNC bridging bonds were identified in the hybrid using advanced spectroscopic techniques. These qualities promote the enhancement inability to transfer electron, a greater number of active sites, and a strong synergetic coupled effect. Towards OER in a solution of KOH with 0.1 M concentration, the hybrid electrocatalyst functioned with high efficiency by demonstrating a low η of 0.295 V (at 10 mA cm−2 current density) and a Tafel slope of 57 mV dec−1. These values are significantly lower than that for NG (0.500 V overpotential, 110 mV dec−1 Tafel slope) and Co-BG (0.320 V overpotential, 70 mV dec−1 Tafel slope) catalysts. Synthesis of N- and B-doped graphene hollow spheres coated with Co3O4 (Co3O4/NBGHSs) was reported by Jiang’s team for use as a potential catalyst for both ORR and OER. [200] The resulting catalyst had the ability to perform with comparatively higher activities and durability for the two reactions than RuO2/C and Pt/C. The coupling between NBGHSs and Co3O4, high electrical conductivity, the strong interaction with O2 being adsorbed, and the specific hollow design were the contributing factors in the improved performance of the catalyst. Using LSV studies in an alkaline 0.1 M solution of KOH, a value of onset potential of about 1.6 V was recorded, which is more negative than those found for pure Co3O4 hollow microspheres, Co3O4/BGHSs, Co3O4/NGHSs, Co3O4/GHSs, NBGHSs, and Pt/C. However, in comparison to the conventional RuO2/C catalyst, the OER onset potential of Co3O4/NBGHSs is greater. The η for attaining a current density of 10 mA cm−2 was determined to be approximately 0.47 V for Co3O4/NBGHSs, which is less than that of RuO2/C with a potential of about 0.52 V. Lu et al. produced N- doped Co3O4 nanocrystals combined with core–shell structured carbon nanotube-graphene nanoribbon (N-csCNT-GNR) scaffolds. [201] A high loading of Co3O4 was achieved during the synthesis by utilizing a microwave-assisted controlled unzipping of MWCNTs. The high surface area of carbon nanomaterials, as well as excellent electrical conductivity, could both be achieved as the csCNT-GNR structures possess an interlinked unzipped graphene nanoribbon and an intact MWCNT core. [202–204] The composite catalysts were also proven to be incredibly active towards both OER and ORR. OER investigation results obtained from the study in 0.1 M KOH have shown that Co3O4/N-csCNT-GNR could perform very actively by exhibiting an onset potential value of 1.51 V. An η of 0.360 V (iR corrected polarization curve) was also observed to obtain 10 mA cm−2 current density. The remarkable activities demonstrated by the synthesized composite was found to be more superior in comparison to Ir/C catalyst for OER [205–206].Ganesan’s research team prepared a bifunctional hybrid electrocatalyst for ORR and OER in which cobalt sulfide NPs are grown on a nitrogen and sulfur co-doped graphene oxide surface through a solid-state thermolysis technique. [207] During the synthesis process, the size, and phase of the particle could be controlled by altering the treatment temperature. Three different treatment temperatures of 400 °C, 500 °C, and 600 °C were employed in addition to the use of cobalt thiourea and graphene oxide to successfully disperse cobalt sulfide NPs onto graphene oxide. Analysis performed using X-ray diffraction has shown that the hybrids produced at 400 °C and 500 °C consisted of pure CoS2 phase while that synthesized at 600 °C contained Co9S8 phase. A simultaneous co-doping of both nitrogen and sulfur on graphene oxide was confirmed via X-ray photoelectron spectroscopy that acts as sites to strongly anchor CoS2 NPs onto the GO surface. Amongst the catalysts synthesized CoS2(400)/N, S-GO displayed an excellent electrode performance. It exhibited a potential of approximately 0.82 V vs. RHE in basic medium (Fig. 8n), that was far superior compared to Ir/C (0.92 V), Ru/C (1.01 V), and Pt/C (1.16 V).In addition to the graphene-based OER electrocatalysts mentioned earlier, many other catalysts have been reported which have the potential to catalyze OER with finer electrocatalytic activity and stability. Some of these catalysts include Fe3O4@Co9S8/rGO-2 [166], Cu@GDY/Co [208], N-CG-CoO [209], and Co-Bi NS/G. [210] Many of the recent research studies that focused on the replacement of platinum-based catalysts with hybrids consisting of non-precious metal have employed mesoporous carbon in addition to graphene and nanotubes. These carbon materials are generally doped with heteroatoms (e.g., nitrogen) before they are introduced into transition metals such as cobalt, iron, manganese, and their complexes. [211–213] Doping with heteroatoms can help modify the surface electronic structure as well as develop surface hydrophilicity to adsorb O2 species particularly in ORR. [214] Non-precious metal-supported carbon materials doped with nitrogen commonly have excellent performance for both OER and ORR when they contain cobalt oxides. This can be because of their ability of cobalt to change their valence states and maintain a steady activity [215].Liu et al. prepared an efficient OER electrocatalyst in the form of 3-D mesoporous carbon-framework-encapsulated CoTe2 nanocrystals from a metal–organic framework (MOF) precursor. [216] They also implemented tellurization and carbonization processes that aided in yielding nanocomposites of CoTe2 and graphitic carbon doped with nitrogen (CoTe2@N-GC) immediately from ZIF-67. The resultant catalyst demonstrated a much greater performance towards OER by exhibiting an η value of 0.300 V (at 10 mA cm−2 current density) and value of Tafel slope of 90 mV dec−1 in comparison to porous N-doped graphitic carbon powder and pristine CoTe2. The presence of N-doped graphitic carbon matrix support provides an interaction with the confined CoTe2 nanocrystals to enhance OER in addition to offering fully accessible active sites and better electrical conductivity. A mesoporous carbon material doped with nitrogen has previously been utilized by Hu’s group along with cobalt oxide NPs enclosed in graphitic layers as a promising non-noble metal oxygen electrode catalyst. [217] A series of catalysts were produced through a facile one-pot synthesis technique that involved polymerization, centrifugation washing, and pyrolysis. Several bifunctional catalysts were developed that possess incredible performance through the adjustment of the carbonization temperatures. Analysis performed on the optimal and as-produced Co-N/C 800 (800 °C carbonization temperature) catalyst showed that the catalyst presented a small reversible η value of 0.96 V between OER and ORR. This value recorded is even greater than those offered by 20 wt% Pt/C (0.270 V), RuO2 (0.390 V), and IrO2 (0.460 V) catalysts, and indicates that the catalyst can act as a top performance non-noble metal bifunctional catalyst for reactions involving reversible oxygen electrode. A facile soft-template mediated technique was employed by Shen’s research team that assisted in fabricating nanostructured Co-Fe double sulphides that are covalently enclosed within N-doped mesoporous graphitic carbon (Co0.5Fe0.5S@N-MC). [218] Characterization methods such as X-ray absorption spectroscopy, X-ray photoelectron spectroscopy, and X-ray diffraction were conducted during the study to unravel the connection between the structural characteristics and the composite’s catalytic behavior. Based on the analysis, there was a moderate substitution as well as a fine distribution of Fe in bimetallic sulfide composites that were suspected to generate a beneficial influence on both the activation and adsorption of species containing oxygen. As a result, a unique catalyst with enhanced performance towards OER and ORR was produced that is far better than the monometallic counterparts. In addition, a covalent bridge exists between the mesoporous carbon shells and the active sulfide particles that create easy pathways for the transport of mass and electron. The features possessed by the Co0.5Fe0.5S@N-MC catalyst resulted in an early onset potential value of around 1.57 V and 10 mA cm−2 current density at a low η of 0.41 V. A relatively lower Tafel slope value of 159 mV dec−1 than IrO2 (267 mV dec−1) was also observed. Wang’s group also described the use of mesoporous carbon to produce electrocatalysts that have the bifunctional ability in the form of Co/Co3O4/Co(OH)2/N-doped mesoporous carbon (Co-NC) through a one-pot synthesis process. [219] The remarkably superior Co-NC catalyst synthesized was Co-NC 750 (750 °C carbonization temperature) that had the highest quantities of pyridinic nitrogen as well as an optimized ratio of three cobalt species. The advantage that was achieved from the presence of the strong enclosure influence between N-graphitic shell of Co-NC and Co/Co3O4/Co(OH)2 core included a reduced reversible overvoltage value of 1.02 V between both OER and ORR in an alkaline medium. Table 1 presents some of the important parameters related to the carbon-supported cobalt catalysts for OER in a conveniently accessible manner.Various other forms of carbon supports have also been combined with cobalt-based materials to create stable and active electrocatalysts for OER. Carbon supports that include carbon cloth, carbon nanodiamond, and carbon black were some of the unique carbon-based materials that have shown improvement in reactions involving oxygen evolution. The following sections will provide reviews on some of the contemporary research works incorporating the special type of carbon supports for OER.Wang and co-workers made use of carbon cloth as support for cobalt phosphide nanoarrays that could efficiently catalyze OER and hydrogen evolution reaction (HER) in basic media. [220] An overall potential of 1.61–1.63 V (ηoverall = 0.380–0.400 V) was necessary for water splitting in a 2-electrode configuration at a current density of 10 mA cm−2 over 72 h. It was found that a layer of CoOx, the active species, covered the CoP catalyst surface during electrolysis, but the improved activity was mostly due to the presence of CoP core and the nanoarray morphology. 10 mA cm−2 current density was obtained at an η value of 0.281 V when the synthesized catalyst was utilized for OER. The development of a bifunctional electrocatalyst for OER and HER using Ni promoted Co disulfide nanowire array and carbon cloth support (Ni2.3%-CoS2/CC) was performed in the past by Fang et al. [221] A simple hydrothermal method was used for the preparation of Ni2.3%-CoS2/CC that allows carbon cloth to be uniformly coated with Ni2.3%-CoS2 nanowires of 50 to 100 nm in diameter and length of several micrometers. The OER activity of Ni2.3%-CoS2/CC was studied in a basic 1 M KOH solution and 0.270 V overpotential was needed to obtain 10 mA cm−2 current density. This value is far more superior than previously reported catalysts that include CoMn LDH (0.324 V), Co-P (0.345 V), NiCo LDH (0.367 V), and Ni-doped Co3O4 (0.530 V) [131,228–230] In addition, a Tafel slope of 119 mV dec−1 was recorded and Ni2.3%-CoS2/CC retained 91% of its current density after 12 h of fixed overpotential electrolysis. Another example of electrocatalyst previously synthesized that implemented carbon support in the form of carbon cloth was Co(OH)2@Ni(OH)2/CC which was produced by Wang’s research team. [222] In comparison to Ni(OH)2/CC, Co(OH)2/CC, and commercial RuO2 catalyst, the newly synthesized hybrid catalyst demonstrated a relatively better OER performance by showing approximately 0.330 V overpotential at 10 mA cm−2 current density. Furthermore, the catalyst exhibited prolonged durability even after 10 h of operation. These enhanced features could be accredited to the distinctive 3-D hierarchical core–shell system present as well as the synergistic influence between Ni(OH)2 and Co(OH)2. Wang and co-workers supported cobalt carbonate hydroxide (CCH), a cobalt-based mineral salt, on carbon black to form a resultant catalyst, indicated as CCH/C that can catalyze OER, as well as ORR.223 Investigations on phase-dependent electrochemical characteristics performed during the research work, showed that extending the time of hydrothermal reaction can considerably modify the CCH’s crystalline phase in CCH/C. This alteration could further influence the activity of the catalyst towards both ORR and OER. Two types of catalyst, denoted as CCH-2/C and CCH-16/C, were generated by applying thermal treatment at 170 °C for 2 h and 16 h respectively. Excellent activity and stability were observed for CCH/C in an alkaline media for ORR in comparison to a Pt/C catalyst (Vulcan XC-72 supporting 40 wt% platinum) available commercially. With regards to OER, a small η of 0.509 V was recorded for CCH-2/C to obtain 10 mA cm−2 current density. This value is less positive than that of Pt/C and it is 0.065 V less active than that found for Ir/C catalyst. These findings provided an evidence that CCH-2/C could be a promising catalyst when utilized as a cathode material for OER. Fan’s research team produced a composite catalyst of Co-OBA/C (OBA = 4,4′-Oxybis (benzoic acid)) involving carbon black. [224] The synthesis process included an integration of a metal–organic framework of Co-OBA with black carbon through a hydrothermal process. The composite was evaluated for ORR and OER using linear sweep voltammetry (LSV) in an alkaline medium, and results indicate a potential of 0.553 V for Co-OBA/C at 10 mA cm−2, relatively smaller than the value obtained from Co-OBA (0.758 V) and Co-OBA + C (0.691 V). In addition, the Tafel slope of Co-OBA/C was the lowest, with a value of 85.7 mV dec−1, when compared to Co-OBA (110.9 mV dec−1) and carbon black (178.4 mV dec−1).A special form of carbon support in the form of nanodiamond (ND) was used by Wu et al. to synthesize a Co-embedded nitrogen-doped graphitized carbon shell that covers an ND core (CoNC/ND). [231] The final catalyst synthesized had a bifunctional property that can improve both ORR and OER. CoNC/ND showed an onset potential of 1.285 V (vs. RHE) for OER and better durability relative to CoNC catalyst obtained from carbon black. The synergistic effect of the Co-N moieties in the carbon shell is expected to have helped improve the catalytic performance while the ND core plays a critical part in maintaining the high stability of CoNC/ND catalyst. In addition to the electrocatalysts described earlier, various research teams have made use of other forms of carbon supports in their studies, and some of these include CoTPP/C [232], NiCoP/C [225], Co2P@NPC [226], and CS-Co/Cs. [227] Moreover these discussed catalysts, there were more advancement in the Co-based catalyst and the electrochemical parameters were represented in Table 2 .Water splitting is one of the most effective and green way to easily convert sustainable energies (solar, wind, and blue energies) into useful high purity fuels (H2 and O2). Among the two half-reactions of water splitting (HER and OER), water oxidation reaction (OER) is a kinetically sluggish reaction which hinders the easy conversion of water into H2 and O2, and considered a bottleneck for large scale applications. So, it is vital to develop a potent catalyst that can demonstrate both prolonged stability and low overpotential, which can further improve the OER activity and display a better overall faradaic efficiency. However, there are some challenges that we must first overcome to create such an effective electrocatalyst:The atomic rearrangement and reaction mechanism is still not well understood due to rapid changes and multiple possibilities between the steps of the OER process. Without understanding the mechanism, we cannot predict the rate-determining step (RDS) of a reaction, and without knowing the RDS we cannot pinpoint the phenomenon regulating the activity for the catalyst, which makes it difficult to further improve it. In the case of multi-metal compound catalysts, the exact recognition of catalytically active sites is important to improve the OER activity. Because of the rapid transformation in OER process, chemical changes and restructuring of the catalyst is difficult to detect and requires advanced tools e.g. combination of in-situ spectrometric methods, electrochemical techniques, microscopy techniques and theoretical calculations for finding out the critical factors affecting a reaction and precisely determining the catalyst active sites.Most of the precious metal-based catalysts for OER (such as Ir- and Ru-based) work efficiently under acidic medium, whereas transition metal-based catalysts work best under alkaline medium. We have only very few catalysts that show excellent behavior within a broad pH range. The search for a versatile catalyst working in a wide range of industrial electrolysis conditions (strongly acidic to strongly basic conditions) is underway. Perhaps, even a better option is to have a catalyst that operates best in the neutral media to avoid corrosion issues and increase the durability of the catalyst compared to alkaline and acidic condition systems.The use of carbon-based support (e.g. carbon black) has a possibility of carbon corrosion and electrochemical oxidation that produces CO or CO2 at high potentials for long-time use. This effect hinders the performance of the catalyst which is due to a reduction in the reactive surface area as well as the dissolution of the support into the electrolyte. Dilution of the doped material into the electrolyte can also disturb the accurate measurement of the activity.In the case of soluble active catalysts, binders are needed to immobilize the catalyst onto the solid surfaces (e.g. GC). The polymer binders hinder effective charge transfer between the catalyst and electrolyte and disrupt the gas permeation from/to the catalyst. So a need exists to convert such catalysts into electrode materials by efficient and economical methods for grafting them onto a solid surface without the use of any binders.The experimental results obtained and the DFT results vary often because the practical catalysts do not comprise of a perfect single surface as the theoretical model frequently use. DFT helps in finding the active sites of a catalyst. As the catalysts get complicated (multimetal doping or addition of support), it is more difficult to identify the exact active site of the catalyst using computational tools.A good electrocatalyst should possess good electrical conductivity, large number of active sites, and resistance to corrosion under high anodic potential. The concept of using different metals to meet individual functional requirements in extended lattices may be useful in constructing catalysts that can follow and satisfy multiple criteria simultaneously. Integrating these high-performance catalysts on to solar cells for fuel production is admired by many as they offer clean and sustainable solutions for energy requirements. Formulation of a bifunctional or trifunctional catalyst that can work in all media is still a dream to be achieved. Creative modeling and production of exclusive nanostructures to enhance the performance for OER is possible with advanced facilities. A catalyst design approach is required following systemic steps based on descriptors rather than the trial-and-error method to create good catalysts. Effective studies should focus on a comprehensive evaluation of reaction mechanism, online monitoring of chemical changes, analysis of structural transformation by applying operando, and in-situ techniques to decipher the catalyst structure that transitions into active sites during the reaction conditions. This knowledge with advanced synthesis tools could help in designing versatile catalysts that can perform outstandingly under varying industrial conditions.World energy consumption has been increasing at a drastic rate. The energy produced from the burning of the limited fossil fuels is not sustainable and affects the environment adversely. So, the need for an energy transition towards a more sustainable and renewable form is of paramount importance for our future generations. Electrolysis of water using renewable energies has received much attention due to its clean method to produce chemical fuels that have the capability to substitute the existing carbon-emitting fossil fuels. In this review, we discussed various cobalt-based electrocatalysts for the OER. It has the capacity for large scale applications by replacing the precious-metal-based catalysts that are scarce and costly. We have discussed cobalt in presence of noble metal (Ir, Ru, Au, and Ag) as catalysts with high activity towards OER in acidic medium with a low amount of noble metals present in the electrocatalyst. Largely the review focuses on catalysts where cobalt is present with other transition metal (such Fe, Ni, Co, and Mn) in a bimetallic or tri-metallic form, which show outstanding OER activity in alkaline media. Furthermore, we have discussed some of the cobalt-based perovskite oxides with partial doping of A-site and B-site with other elements and anion substitution, which aided in the high activity of the perovskite catalyst. As support, we have included the effect of carbon-based compounds that enhance the OER activity in presence of cobalt catalysts. The review also includes a discussion on the mechanism of OER along with comparing the performance of OER catalysts with the help of measurement standards like the overpotential and Tafel slope.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This study was supported by the NPRP grant (NPRP8-145-2-066) from the Qatar National Research Fund (a member of the Qatar Foundation). The statements made herein are solely the responsibility of the authors. The author(s) would also like to acknowledge the support from Qatar University's internal grant QUCG-CENG-19/20-7.

The future of the world energy lies in clean and renewable energy sources. Many technologies, such as solar cells, wind turbines, etc., have been developed to harness renewable energies in different forms of fuel. Amongst them, electrolysis of water to produce oxygen and hydrogen is one of the paramount developments towards achieving clean energy, which has attained significant attention due to its green and simple method for the production of fuels. In electrolysis of water, the half-reaction containing the oxygen evolution reaction (OER) is a reaction that is kinetically sluggish, which requires higher overpotential to produce O2, when compared to the other half-reaction, i.e. hydrogen evolution reaction (HER). Many electrocatalysts are studied extensively to be used in the OER process to get an economical yield out of it. Noble metal-based catalysts are the state-of-the-art catalyst used for OER currently. But due to their high cost and scarcity, they cannot be applied in a large-scale manner to be used in the future. The non-noble metals (transition metals and perovskites) are gaining interest by exhibiting on par or better OER performance compared to the noble metal used. Due to their low cost, ample resources, and several metals available, they have opened up a variety of areas with a different combination of metals to be used as a catalyst for OER. Amongst these metals, cobalt has received massive appreciation for performing as an excellent OER catalyst. Multi metals, multimetal mixed oxides, multimetal phosphides, perovskites, and carbon-supported catalysts containing cobalt have shown low overpotential with high long-term stability. Therefore, in this review, we go through different cobalt-based electrocatalysts for OER, the general mechanism governing the OER process, the challenges that we are facing today to enhance the catalytic performance, and future aspects to overcome such challenges.

The hydrogenation of aromatic rings has been widely concerned due to its valuable applications in industry [1–3]. The hydrogenation of toluene to methylcyclohexane (MCH) is an important way of hydrogen storage [4–7]. The saturation of toxic aromatic compounds is desired because they are the main source of air pollutions [8–10]. Benzene [11,12] and toluene were usually used as probe molecules to investigate the hydrogenation of aromatic rings, in which toluene is of relatively low toxicity [13]. Supported Pd and Ru catalysts have been widely studied for the hydrogenation of aromatic rings [14–16].It is known that solvents affect the catalytic activities due to different solvent polarity, H2 solubility, hydrogen transfer ability, solvent-reactant interactions and solvent-catalyst interactions. Up to now, quite a few studies were published concerning the effects of hydrogen solubility and solvent polarity on the hydrogenation of aromatic rings [17–19]. However, the interactions among reactants, solvents and catalyst surfaces are complicated and to understand such interactions needs massive efforts of studies. In fact, such studies were relatively few.In this paper, we present a preliminary study on how the pre-adsorbed solvents affected the strengths of adsorption of reactants on the catalysts and thus changed the activity for the hydrogenation of toluene. Specifically, the Pd/SiO2 and Ru/SiO2 (5%wt) were prepared for the hydrogenation of toluene in n-hexane, isopropanol (IPA), tetrahydrofuran (THF) and methanol. The microcalorimetric adsorption was employed to measure the interactions of solvents with the catalysts, as well as their effects on the adsorption of toluene. The hydrogen transfer from IPA to toluene on the surfaces was observed by IR, which accounted for the promotion effect of IPA for the hydrogenation of toluene over the supported Pd and Ru catalysts.The preparation, characterization and catalytic tests of catalysts were described in detail in the Supporting Information (SI).A calculation proved that the reaction rates were not affected by the mass transport limitation (see Tables S2 and S3 in SI).The physical properties of the support and supported catalysts were studied previously [15]. The surface areas of the SiO2, Pd/SiO2 and Ru/SiO2 were 877, 634 and 821 m2/g, with the average pore sizes of 4.5, 3.4 and 4.1 nm, respectively. The averaged metal particle sizes were estimated to be approximately 2.1 and 6.5 nm, respectively, in the Pd/SiO2 and Ru/SiO2, by TEM images. In addition, the H2 uptakes were about 204 and 21 μmol/g for the Pd/SiO2 and Ru/SiO2, respectively, as measured by the microcalorimetric adsorption of H2. The physical properties of the samples were described in detail in SI.The conversion and turnover frequencies (TOF) of toluene at different weight hourly space velocities (WHSV) on the Ru/SiO2 and Pd/SiO2 in n-hexane, IPA, THF and methanol are presented in Fig. 1 . MCH was the only product detected (Fig. S4). The conversion of toluene decreased with the increase of WHSV. The conversion of toluene was different in different solvents over the catalysts. The results indicated that the activities for the reaction in the solvents followed the order of n-hexane> IPA > THF > methanol over the two catalysts, revealing that aromatic rings were easier to be hydrogenated in n-hexane and IPA than in THF and methanol. In addition, the conversions of toluene were higher on the Ru/SiO2 than on the Pd/SiO2 in each of these solvents, demonstrating that Ru was more active than Pd for the hydrogenation of aromatic rings.The turnover frequency (TOF) of toluene was calculated according to the number of converted toluene molecules per second divided by the number of surface active metal sites determined by the microcalorimetric adsorption of H2 (considering that the low H2 pressures might minimize the extents of absorption of H2 in Pd [20]). The results are given in Fig. 1. The conversion of toluene decreased, while the TOF of toluene increased with WHSV, until the constant values were reached, which represented the intrinsic activity of catalysts for the reaction [21–23]. It was meaningless to compare the TOF values for the Pd/SiO2 and Ru/SiO2 owing to the absorption of H2 in Pd. However, it was meaningful to compare the TOF values of the same catalyst for the hydrogenation of toluene in different solvents. The maximum TOF values of toluene were about 0.062, 0.029, 0.010 and 0.002 s−1 on the Pd/SiO2 in n-hexane, IPA, THF and methanol, respectively, while they were about 1.9, 1.6, 0.28 and 0.037 s−1, respectively, on the Ru/SiO2. The intrinsic activity with the solvents followed the order of n-hexane> IPA > THF > methanol on the Pd/SiO2, while the order was n-hexane≈ IPA> > THF > methanol on the Ru/SiO2.The dielectric constants are 1.88, 19.9, 7.58 and 32.7 (Table S4) for n-hexene, IPA, THF and methanol, respectively. No clear relationship was found for the catalytic activity and dielectric constants of solvents, consistent with the results reported [17,24].IPA, THF and methanol are hydrophilic while n-hexane is hydrophobic. SiO2 is also hydrophilic so that it adsorbed n-hexane weakly, which might be a reason why it was a good solvent for the hydrogenation of toluene over the Pd/SiO2 and Ru/SiO2 catalysts.Previously, we found that the hydrogenation of diisopropylimine was significantly promoted by IPA on the Ni catalysts due to the HTR (hydrogenation transfer reaction) between diisopropylimine and IPA [25]. A similar effect must occur here with IPA as a solvent. To confirm the effect, toluene and IPA were co-fed onto the Pd/SiO2 and Ru/SiO2 under N2 atmosphere. The results are given in Table S5. The data showed that some IPA was dehydrogenated to acetone while some toluene was simultaneously hydrogenated to MCH. The equivalent molar ratio for the dehydrogenation of IPA and the hydrogenation of toluene was 3/1. The ratio was found to be 3.3 on the Pd/SiO2 and 5.0 on the Ru/SiO2, indicating that not all the H atoms dehydrogenated from IPA were used to hydrogenate toluene (they might be also combined to H2). With such effect, IPA promoted the activity for the hydrogenation of toluene, as compared to the solvents THF and methanol. In addition, the HTR effect was stronger on the Ru/SiO2 than on the Pd/SiO2, owing to the higher activity for the dehydrogenation of IPA to acetone on the Ru/SiO2 than on the Pd/SiO2.Fig. S5 (A) shows the results for the microcalorimetric adsorption of solvents on the SiO2. The initial heats for the adsorption of n-hexane, IPA, THF and methanol on the SiO2 were 46, 70, 79 and 59 kJ/mol, respectively, with the coverages of about 585, 2379, 2412 and 2635 μmol/g. Thus, the interaction was weak between SiO2 and n-hexane, while that was quite strong between SiO2 and IPA or THF. The initial heat for the adsorption of methanol was significantly higher than that of n-hexane, but lower than those of IPA and THF. In addition, the heats for the adsorption of methanol decreased significantly slower with coverage than those of IPA and THF, probably owing to that the methanol molecule is smaller than those of IPA and THF so that more molecules of methanol could be adsorbed on the same surface area of SiO2. Fig. 2 shows the adsorption of solvents on the Ru/SiO2. The initial heats for the adsorption of n-hexane, IPA, THF and methanol on the Ru/SiO2 were 52, 74, 82 and 66 kJ/mol, respectively, with the coverages of about 637, 2933, 2441 and 2591 μmol/g, while those on the Pd/SiO2 were 80, 90, 95 and 76 kJ/mol, respectively, with the coverages of 598, 2578, 2170 and 2702 μmol/g (see Fig. S5 (B)). Thus, the adsorption heats of these solvents were higher on the metals than on the SiO2, indicating the stronger interactions of solvents with the metals than with the support. In addition, the adsorption heats of these solvents were significantly higher on Pd than on Ru, revealing the weaker interactions of solvents with Ru than with Pd, which might be one of the reasons why the Ru/SiO2 was less affected than Pd/SiO2 by the solvents and why the Ru/SiO2 was more active than Pd/SiO2 for the hydrogenation of toluene in these solvents.The coverages of each of the following solvents n-hexane, THF and methanol on the catalysts were close to those on the support, while the coverages of IPA on the catalysts (2933 and 2578 μmol/g) were significantly higher than that on the support (2379 μmol/g), indicating the dehydrogenation of IPA on the catalysts.The adsorption heats of solvents could be correlated with the conversion of toluene. Specifically, the weaker adsorption of solvents favored the conversion of toluene. For example, n-hexane was weakly adsorbed and the conversion of toluene was high on the catalysts with n-hexane as a solvent, while THF and methanol were quite strongly adsorbed on the catalysts and inhibited the conversion of toluene. However, this rule was not applicable to IPA as a solvent, owing to the HTR effect which promoted the conversion of toluene in another way, although IPA was more strongly adsorbed than methanol on the catalysts. Fig. 3 shows the adsorption heats and coverages of toluene over the Ru/SiO2 before and after the pre-adsorption of solvents. The initial heats of toluene on the clean SiO2, Pd/SiO2 and Ru/SiO2 were 72, 77 and 85 kJ/mol, respectively, with the coverages of 2521, 2502 and 2569 μmol/g [15] (Table S6), indicating that toluene adsorbed more strongly on metals than on SiO2 and more weakly on Pd than on Ru which could explain the higher conversion of toluene on Ru than on Pd.After the pre-adsorption of n-hexane, IPA, THF and methanol, the initial heats of toluene on the Ru/SiO2 decreased by 8, 13, 21 and 22%, respectively, with the decreases of coverages by 23, 18, 24 and 29%. Similar results were obtained on the Pd/SiO2 (Fig. S6), indicating that the pre-adsorbed solvents inhibited the adsorption of toluene on Pd and Ru. After the pre-adsorption of solvents, the adsorption heats of toluene on the Pd/SiO2 and Ru/SiO2 catalysts followed the order of n-hexane> IPA > THF > methanol (Table S6), in consistence with their activity orders for the hydrogenation of toluene in these solvents. The results demonstrated that the strength of adsorption of toluene with the solvents on the catalysts was a key factor affecting the activity for the hydrogenation of toluene. In addition, toluene was more strongly adsorbed on the Ru/SiO2 than on the Pd/SiO2 with and without the solvents, which might be an important reason why the Ru/SiO2 was significantly more active than the Pd/SiO2 for the hydrogenation of toluene.Although the adsorption heat of THF was higher than that of methanol, the coverage of methanol was significantly higher than that of THF, on the catalysts, leading to the more decreased coverage of toluene and thus the more decreased activity in methanol than in THF on the catalysts for the hydrogenation of toluene (Scheme S1). Fig. 4 shows the IR spectra of adsorbed IPA on the support and catalysts. For the adsorbed IPA on SiO2, the bands at 2978, 2937 and 2893 cm−1 were assigned to the vibrations of ν(-CH3) in IPA, while the bands at 1466 and 1387 cm−1 were attributed to the vibrations of δas(-CH3) and δs(-CH3) [26–29]. The band at 1345 cm−1 belonged to the vibration of δ(α-C-H) [27,30]. Since O atoms in IPA may interact strongly with surface cations, we suggested the surface structures of adsorbed IPA on SiO2 as in Scheme S2 (a) and (b).The IR spectra of IPA adsorbed on the Pd/SiO2 and Ru/SiO2 were mostly similar to that on SiO2, except for the two newly appeared bands at 1692 and 1292 cm−1. Thus, the same surface structures might be formed for the adsorption of IPA on the metals as on SiO2 (Scheme S2 (c) and (d)). The new bands at 1692 and 1292 cm−1 could be assigned to the vibrations of ν(C=O) [29,31] and ν(C-C-C) in acetone [32,33], respectively, indicating the formation of surface acetone (Scheme S2 (e)) resulted from the dehydrogenation of IPA on the metals. It should be mentioned that the intensity of the band at 1692 cm−1 was significantly stronger on the Ru/SiO2 than on the Pd/SiO2, indicating that the dehydrogenation of IPA was easier on Ru than on Pd. Fig. 5 shows the FTIR spectra of co-adsorbed IPA and toluene on the Ru/SiO2, as well as those of adsorbed toluene and IPA only on the Ru/SiO2 for comparison. The IR spectra for the co-adsorbed IPA and toluene on the Pd/SiO2 were not shown since they were similar to those on the Ru/SiO2.It is seen that the bands of adsorbed toluene did not change with the pre-adsorbed IPA, but the intensities of bands of the skeletal vibrations of aromatic rings (1603 and 1496 cm−1) [15,34,35] were weakened, indicating that the pre-adsorbed IPA did not change the surface structure of adsorbed toluene, but reduced the amount of toluene adsorbed on Ru.The intensities of the characteristic bands at 1603 and 1496 cm−1 in the spectra of IPA/toluene almost disappeared. Meanwhile, the bands belonging to ν(-CH3) in IPA at 2978, 2937 and 2893 cm−1 were weakened while the band for ν(-C=O) at 1692 cm−1 (surface acetone) was slightly enhanced. These results suggested the surface reactions occurred between adsorbed toluene and IPA which might be schematically expressed in Scheme 1 .Adsorption heats are the good measures of interactions of solvents with catalysts. The weaker adsorption of a solvent resulted in the stronger interaction of toluene with the catalysts. The strength of adsorption of toluene with the solvents on the catalysts was a key factor affecting the activity for the hydrogenation of toluene. Normal hexane was a good solvent for the hydrogenation of toluene on the Pd/SiO2 and Ru/SiO2 since it adsorbed weakly on these catalysts.The solvents studied adsorbed more weakly on Ru than on Pd so that toluene was more strongly adsorbed on Ru than on Pd with pre-adsorbed solvents, which might be a reason why Ru was more active than Pd here.Although IPA adsorbed strongly on the metals and inhibited the adsorption of toluene significantly, it promoted the activity for the hydrogenation of toluene by the HTR effect. Such effect was more profound on Ru than on Pd, leading IPA to be a good solvent (as good as n-hexane) for the hydrogenation of toluene on Ru.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.Financial supports from the NSFC (21773108), NSFC-DFG (21761132006) and fundamental research funds for central universities are acknowledged. Supplementary material Image 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.catcom.2021.106330.

The effects of solvents (n-hexane, isopropanol (IPA), tetrahydrofuran (THF) and methanol) on the hydrogenation of toluene over the Pd/SiO2 and Ru/SiO2 catalysts were studied. Microcalorimetric adsorption and IR spectroscopy were employed to understand the effects. It was found that n-hexane adsorbed weakly on the catalysts and thus affected less the hydrogenation of toluene, while THF and methanol adsorbed strongly on the catalysts and inhibited the activity of hydrogenation of toluene significantly. IPA also adsorbed strongly on the catalysts, but it exhibited a hydrogen transfer effect on the surfaces that promoted the conversion of toluene.

The need to diminish fossil fuel dependence in the chemical industry has led to the intensification of bio-based alternative products. Bio-based chemicals, products that are derived from biomass and other biological materials, are pivotal to the development of a more innovative and low-emissions economy while ensuring biodiversity and environmental protection [1]. The use of bio-based chemicals have raised as an alternative to standard chemicals since their environmental footprint is limited compared to their traditional counterparts. From a technical point of view, almost all the fuels and chemicals normally produced from fossil feedstocks can be produced from biomass in the so-called biorefinery [2]. Thus, bio-chemicals could drive a new economy model based on more thoughtful uses of natural resources. Hence, the European Commission aims to accelerate the market uptake of bio-chemicals, and estimates an annual growth rate of around 3.6% per year from 2018 to 2025, what represents an important expanding sector [1]. At present, researchers and entrepreneurs are involved in the development and commercialization not only of biofuels, but also of bio-derived replacements for basic chemicals. The path to a more sustainable industry will inevitably depend on the development of new high value-added chemicals, via catalytic process, from renewable sources [3].A very attractive platform chemical from biomass is glycerol, which was identified as a top 12 bio-based chemical by the US Department of Energy [4]. Production of chemicals from raw glycerol is a good alternative to absorb the glycerol surplus generated from biodiesel industries. Several products can be obtained from glycerol, either gaseous products (i.e. hydrogen, syngas, alkanes, etc.), as well as high value-added liquid products (i.e. hydroxyacetone, pyruvaldehyde, acrolein, propanal, lactic acid, glyceric acid, acetaldehyde, etc.) [5–10].Another valuable chemical that can be produced from hydrogenolysis of glycerol is 1,2-propanediol. It is widely used in the synthesis of unsaturated polyester resins, as coolant and also in pharmaceutical, food, and cosmetics formulations [11]. Currently, 1,2-propanediol is produced through the hydration of propylene oxide, which in turn comes from the selective oxidation of propylene, a fossil resource [12].Conversion of glycerol to 1,2-propanediol involves the removal of an oxygen atom by the addition of hydrogen (C-O bond cleavage with simultaneous hydrogen addition), a catalytic reaction known as hydrodeoxygenation (HDO). The conventional way to carry out this process requires an external H2 supply, usually derived from fossil fuels, which increases the cost and can generate safety issues due to the high diffusivity and flammability of molecular hydrogen [13]. In addition, hydrogen can block metal centers causing an inhibitory effect on catalytic activity [14]. Other explored alternative is the catalytic transfer hydrogenation (CTH) by using a hydrogen donor molecule such as methanol, 2-propanol, or 2-butanol [15]. Nevertheless, the formation of by-products and the cost of acquisition and handling of the hydrogen donor can significantly increase process costs [13]. A suitable alternative to circumvent the use of external and non-renewable sources of hydrogen is to carry out glycerol HDO with in-situ produced H2 via aqueous-phase reforming (APR) of the substrate, in a one-pot reaction.The present investigation aims to develop a catalytic system that could drive the in-situ generation of hydrogen and its immediate consumption, in the reaction medium, in the HDO reaction. Therefore, a metal-acid bifunctional catalyst is required. In a simple way, the acid function is pivotal for the dehydration reaction whereas the metal function catalyses the hydrogenation reaction [16]. Scheme 1 shows the widely accepted reaction path for the glycerol HDO in liquid phase.Non-noble metals are less expensive than noble metals and present a greater selectivity towards C-O bond cleavage, and therefore, they are more efficient and economical for 1,2-propanediol production [17]. Transition metals and, more specifically, cobalt-based catalysts, have proved effective in both the production of hydrogen by aqueous-phase reforming (APR) and for glycerol HDO reaction depending on metal–acid/base characteristics [18–20].Promoters can increase the metal-support interaction, the dispersion of the active metal and can also tune the acid/base characteristics of the catalyst [21]. The addition of a promoter is an economical and time-efficient strategy that can enhance activity [22]. Cerium has been used by several researchers to improve catalyst activity and stability in liquid phase reactions. Its positive effect has been mainly attributed to the existence of oxygen vacancies in the cerium oxide (CeO2) lattice, which may promote WGS reaction by the activation of the H2O molecule [23–26]. In addition, its high oxygen mobility can improve hydrothermal stability and avoid the formation of coke precursors [27,28].In our previous work [29], cobalt-based catalysts were synthesized from cobalt aluminate spinels with different Co/Al molar ratio. These catalysts showed remarkable glycerol conversion and considerable efficiency for C-O cleavage where catalyst with CoAl = 0.625 molar ratio exhibited the most promising performance. Based on that result and known the notable catalytic properties of cerium, in the present work, Ce-modified cobalt aluminate (Co/Al molar ratio = 0.625) catalysts are investigated for the glycerol HDO. In this way, we seek to modify the physico-chemical properties of the pristine catalyst to produce value-added liquid products from glycerol HDO. As far as we know, this type of one-pot synthesized catalysts have not yet been studied in glycerol HDO. The catalytic experiments were carried out in a continuous reactor, where hydrogen was in-situ generated via APR. The synthesized catalysts were thoroughly characterized by several techniques, either in the fresh and reduced form. The various physicochemical properties were correlated with the catalytic performance. In addition, exhausted catalysts were characterized to identify the main deactivation causes.A series of catalysts based on cobalt aluminate (nominal Co/Al molar ratio = 0.625) doped with cerium were one-pot synthesized by coprecipitation method. In a typical synthesis, an aqueous solution containing Co(NO3)2.6 H2O (99.999% trace metal basis, Sigma Aldrich), Al(NO3)3.9 H2O (98% trace metal basis, Fluka) and Ce(NO3)3.6 H2O (99.999% trace metal basis, Sigma Aldrich) was dropwise added into a beaker containing an aqueous solution of Na2CO3 (99.8%, Fluka), under stirring. The synthesis was carried out at room temperature at pH 10, adjusted with NaOH solution (2 M). The resulting suspension was aged at room temperature for 24 h, filtered, washed several times with de-ionized water, dried overnight at 110 °C in an oven and calcined in a muffle furnace at 500 °C for 5 h (heating at 5 °C/min), in static air atmosphere. The obtained solids were abbreviated xCeCoAl (x: 0, 0.3, 2.1), where x denotes the weight percentage of Ce. As a reference, bare CeO2 and Co3O4 were also synthetized by precipitation, following the same protocol.Bulk chemical composition of the solids and metal leaching in catalytic runs were analysed by inductively coupled plasma optical emission spectroscopy (ICP-OES). Textural properties of the solids were obtained from nitrogen adsorption-desorption isotherms at 77 K in a Micromeritics TRISTAR II 3020 equipment. Prior to the adsorption, the samples were outgassed at 300 ºC for 10 h for removing moisture and adsorbed gases. The specific surface area and the main pore size were determined with the BET and BJH (desorption-branch) methods.XRD spectra were obtained in a PANalytical Xpert PRO diffractometer with CuKα radiation (λ = 1.5418 Å). X-ray diffracted radiation was recorded from (2θ values) 20–80º for the powder samples. The identification of the crystal phases was carried by comparison with International Centre of Diffraction Data (ICDD) database. Crystallite size was calculated from the X-ray line broadening analysis using Debye-Scherrer formalism. The lattice parameter (a) of cubic crystal structure was calculated by Eq. 1: (1) a = λ 2 sin θ h 2 + k 2 + l 2 were; θ is the diffraction angle; and h, k and l are the Miller indices.STEM images were obtained in a FEI Titan Cubed G2 60–300 electron microscope with a high-brightness and a Super-X EDX system under HAADF detector for Z contrast imaging (camera length of 185 mm). The samples were dispersed in ethanol and kept in an ultrasonic bath for 15 min. Afterwards, a drop of suspension was spread onto a TEM copper grid (300 Mesh) covered by a holey carbon film and dryed under vacuum. Particle size distribution was obtained from the statistical analysis of at least 300 particles. The average size of the nanoparticles was calculated from volume to surface ratio (Eq. 2), using ImageJ software: (2) < d > = Σ n i d i 3 Σ n i d i 2 being di the diameter of ni particles.The oxidation state of the surface elements was analyzed by X-ray photoelectron spectroscopy (XPS) on a SPECS spectrometer with Phoibos 150 1DDLD device, using monochromatic Al Kα (1486.7 eV) X-ray source with 30 eV pass energy at 0.05 eV steps. Samples, previously degassed, were introduced to the ultra-high vacuum analysis chamber (10-6 Pa) where the detailed analyses of the elements were performed (time 0.1 s and step energy 30 eV) with an exit angle of 90°. Samples were reduced in-situ, when required. The spectrometer was previously calibrated with Ag (Ag 3d5/2, 368.26 eV). The BE were calibrated by taking C 1 s peak (284.6 eV) of adventitious carbon as reference. Peaks were deconvoluted after Shirley background subtraction, using a mixed Gaussian Lorentzian function (CASA XPS software).The UV–vis–NIR DRS spectra were recorded in a Cary 5000 equipment coupled to Diffuse Reflectance Internal (Varian). The reflectance data were converted into absorption by the Kubelka-Munk transformation. The Tauc plots were used to evaluate the difference in the energy of inner electron transitions of the solids. For the calculation, ( α h ν ) 2 v s h ν plot was used, where α is the absorption coefficient, ν is the frequency of light and h is the Planck’s constant. 27Al Solid State NMR measurements were performed on a 9.4 T Bruker AVANCE III 400 spectrometer operating at resonance frequencies of 104.26 MHz for 27Al. AlCl3 aqueous solution was used as a reference. The spectra were acquired at a spinning frequency of 60 kHz employing a PH MASDVT400W BL 1.3 mm ultrafast probe head. A single pulse of 0.3 μs duration was applied (recycle delay 0.2 s, 36,000 scans).The reducibility of the calcined samples was analysed by H2-TPR, carried out in a Micromeritics AutoChem 2920 apparatus. The solid was initially heated in He stream at 550 °C for 1 h to desorb impurities, and then cooled down to room temperature. Then, 5% H2-Ar flow was passed through the bed containing the sample while temperature was increased up to 950 ºC (heating rate 10 ºC/min) and hold for 1 h. A cold trap was used to prevent water generated by reduction, and reactor exhaust was analysed by Thermal Conductivity Detector (TCD).Both the hydrogenation and dehydrogenation reactions require metallic function. The number of accessible metallic cobalt atoms was measured by H2 pulse chemisorption, carried out in a Micromeritics AutoChem 2920 apparatus, at 35 ºC. Samples were previously reduced at 600 °C for 30 min. A chemisorption stoichiometry HCo = 1/1 [30] and a cross-sectional area of 0.0662 nm2/atCo were assumed [31]. Further information on the metallic function of the catalysts was obtained by measuring their activity in cyclohexane dehydrogenation, which will mainly depend on the metal accessibility [32]. The cyclohexane dehydrogenation was performed over 20 mg of reduced catalysts, in a fixed-bed reactor at 250 ºC and atmospheric pressure, feeding a mixture of anhydrous cyclohexane and hydrogen (1:3000 mol ratio). The gas product (benzene and cyclohexane) were online analysed by GC (column Al2O3-KCl, HP) coupled to a flame ionization detector (FID).The amount and strength of surface acid sites of the reduced catalysts were measured by means of NH3-TPD and isomerization of 3,3-dimethyl-but-1-ene (33DM1B) model reaction. For ammonia chemisorption/desorption experiments (Micromeritics AutoChem 2920 equipment) a series of 10% NH3-He pulses were introduced at 90 ºC, until saturation. Subsequently, the sample was exposed to He flow for 60 min to remove reversibly bound NH3. Finally, the temperature was raised to 950 °C (heating rate 5 °C/min) with continuous ammonia monitoring. The total acidity was calculated from the integration of the pulses, and the strength of the acid sites was evaluated from the corresponding TPD profile. The model reaction of skeletal isomerization of 33DM1B was used to characterize the Brønsted acid sites, since Lewis acid centers are not involved in this reaction [33]. The catalyst (100 mg) was in-situ reduced, and cooled down to the reaction temperature (300 °C) under inert flow. The 33DM1B partial pressure and flow rate were set at 20 kPa and 15.2 mmol/h, respectively. The obtained products were online analysed by GC-FID on a RTx-1 (Restek) column.The characterization of the carbonaceous deposits in the spent catalysts was carried out by Raman spectroscopy (Renishaw InVia Raman spectrometer, Leica DMLM microscope) using 514 nm laser. The power density of the laser beam was reduced in order to avoid the photo-decomposition of the samples. In order to improve the signal to noise ratio, 40 s were used for each spectrum and 10 scans were accumulated at 10% of the maximum power of the 514 nm laser, in the 1000–2000 cm−1 spectral window.Carbonaceous deposits on spent catalysts were quantified by Temperature Programmed Hydrogenation (TPH) in a Setaram Setsys Evolution thermobalance coupled to a Mass Spectrometer (Pffeifer Vacuum OmniStar) following the evolution of m/z = 15 (CH4) signal. First, sample was cleaned under a He flow, at 550 °C for 1 h in order to remove absorbed organics. After cooled down to 40 °C, 5%H2/Ar flow was passed through the sample heated at 10 °C/min up to 900 °C.Catalytic performance was evaluated in a bench-scale fixed-bed up-flow reactor (Microactivity Effi, PID Eng&TEch) with synthetic aqueous solution of glycerol (10 wt% glycerol) at 260 ºC/50 bar, operating at WHSV= 24.5 h-1. About 0.5 g of catalyst, particle size of 0.04–0.16 mm, were mixed with deactivated quartz wool and in-situ reduced under 10% H2/He flow at 600 ºC and atmospheric pressure, for 1 h. The reactor was pressurised with He and when the desired pressure was achieved, the He flow was switched to bypass, and the glycerol solution (0.2 mL/min) was fed to the reactor while the temperature was progressively raised up to the reaction temperature, at 5 ºC/min. Catalytic performance was measured at 3 h TOS (time on stream). Zero time was taken when reactants reached the catalyst bed, once the reaction temperature was reached. Gaseous and liquid phases were separated at 5 ºC in a Peltier device. The gaseous products were swept away with a He flow (40 mL/min) applied immediately after the backpressure regulator. Product distribution was online analysed by a μGC (Agilent) equipped with four columns (Al2O3-KCl, PPQ and MS5A columns that used He as a carrier, and MS5A column which used Ar as a carrier). The liquid product was collected every hour in 2 mL glass vials and off-line analysed by GC-FID (Agilent, 6890N) and HPLC-RI (Waters, Hi-Plex H column). The quantification of liquid compounds was performed by external calibration. Total organic carbon (TOC) in the liquid phase was measured on a Shimadzu TOC L apparatus. The carbon balance was 96% ± 5 for all the experiments. After catalytic test, spent catalyst was recovered and characterized by a sort of techniques.The catalytic performance was calculated according to the following indices. The total glycerol conversion (XGly) was calculated as: (3) X G l y ( % ) = 100 × F G l y i n − F G l y o u t F G l y i n where F G l y i n and F G l y o u t are the glycerol molar flow at the reactor inlet and outlet, respectively.The carbon yield to liquid (Xliq) is the ratio of the total moles of carbon in the liquid products to the moles of carbon fed: (4) X l i q ( % ) = 100 × ∑ i = m n F C atoms , liq o u t 3 F Gly i n yield (Yi) and selectivity (Si) of liquid product i were calculated on the basis of carbon molar flow of i product in liquid phase, as follows: (5) Y i ( % ) = 100 × F C atoms , i o u t 3 F Gly i n (6) S i ( % ) = 100 × F C atoms , i o u t 3 F Gly i n · X G l y Selectivity to alkanes (Salk) was calculated on the carbon basis, as follows: (7) S a l k ( % ) = 100 × ∑ n = 1 3 n · F C n H 2 n + 2 o u t 3 F Gly i n · X G l y While the overall selectivity to liquid products accounts the total C atoms in liquid phase, excluding unreacted glycerol. (8) S l i q ( % ) = 100 × ∑ i = m n F C atoms , i o u t 3 F Gly i n · X G l y Hydrogen yield (YH2) is the ratio between the molar flow of hydrogen produced and the theoretical one according to the glycerol fed to the reactor. (9) Y H 2 ( % ) = 100 × F H 2 o u t 7 · F Gly i n The degree of oxygen removal (DOR) was defined as the ratio of the removed oxygen atoms in the liquid phase to the initial molar flow of oxygen. (10) D O R ( % ) = 100 × F Oatoms i n − ∑ i = m n F O atoms , i o u t F Oatoms i n Table 1 lists the bulk chemical composition and textural properties of the synthesized solids. Actual metal content measured by ICP-OES revealed a good agreement with the nominal Ce content (variation less than 3%) and Co/Al ratio (slightly higher than nominal).The isotherms of all the calcined solids ( Fig. 1A) were IV type, which were characteristic of mesoporous materials, with the P/P0 position of the inflection point (around 0.5) corresponding to a mesoporous range diameter. The hysteresis loop shape suggested randomly distributed mesopores formed by nanoparticle assembles. Regarding the BJH pore size distribution of the solids, it varied with Ce content (Fig. 1B). While pristine 0CeCoAl and 0.3CeCoAl presented a unimodal PSD with maximum at 6.8 nm sized pore, sample 2.1CeCoAl presented a bimodal distribution with pores at 7.8 nm and 13 nm, which would anticipate a certain ceria segregation in the later assay. The surface area, pore volume and average pore size of the solids are shown in Table 1. Pristine 0CeCoAl showed SBET = 125.3 m2/g, which was slightly lower than stoichiometric CoAl2O4 (about 150 m2/g) [29]. Recall that 0CeCoAl solid had more Co than stoichiometric (CoAl = 0.625 vs. 0.5), in form of Co3O4, characterized by its low surface area [34]. Coprecipitation of Ce together with Co and Al leaded to an increase by 16–18% in the SBET as compared to 0CeCoAl counterpart. Likewise, an increase in pore volume was observed (between 22% and 33% increase) for Ce-doped solids.After reduction (at 600 °C, 2 h), the isotherms were also of IV type. The specific surface area of Ce-doped solids slightly dropped (by 7–10%) as compared to calcined counterparts, with a concomitant increase in the mean pore size values (increase by 14–18%). It should be noted that this increase was much less pronounced than that revealed by 0CeCoAl sample (41%). Decrease in the SBET in the reduced solids was attributed to the dilution effect by metallic cobalt in the solid surface. Compared to 0CeCoAl assay, SBET of xCeCoAl reduced samples was 30–36% higher. Therefore, up to this point, we can conclude that the addition of Ce notably improved the textural properties of the solid in both calcined and reduced conditions, acting as structural promoter.XRD diffractograms of Ce-doped and pristine cobalt aluminate solids, in their calcined and reduced forms, are displayed in Fig. 2. All the calcined samples exhibited the characteristic peaks for spinel phase. It should be noted that the Co/Al ratio in our catalyst was 0.64 which is around 30% larger than the stoichiometric ratio in CoAl2O3 (i.e. CoAl = 0.5). Thus, the segregation of part of Co atoms to form Co3O4 was expected to occur. The formation of CoAl2O4 spinel phase occurs due to the counter-diffusion of Coδ+ and Al3+ ions at the interface between Al2O3 and Co3O4 oxides, both isotopic crystal structures, formed at the early stages of the thermal treatment [35]. Unfortunately, it was difficult to differentiate between Co3O4 and CoAl2O4 by XRD analyses due to the similarities in the crystal structure and d-spacing (PDF 00–042–1467 and PDF 00–044–0160, respectively). Diffraction patterns of Ce-doped solids were almost the same as that of pristine 0CeCoCAl, indicating that the solid retained the same crystal structure upon Ce doping. Moreover, the spinel peaks became narrower upon Ce addition (inset in Fig. 2A), indicating increase in the crystallite size. Only the catalysts with the highest Ce content showed diffraction peaks at 28.7º and 48.3° (2θ) attributed to CeO2 in its cerianite-phase structure (PDF 00–034–0394), indicating segregation of Ce. The broadness of its characteristic XRD peaks suggested small size of the ceria crystallites. This peak was not observed for sample 0.3CeCoAl, which confirmed that for that solid, Ce ions were substituted in the spinel lattice. According to literature, the cerium solubility into spinel phase is CeCo = 0.03 molar ratio [36]. In our 2.1CeCoAl solid, the actual Ce/Co ratio was 0.025, close to the solubility limit. A careful inspection of (311) peak of the spinel phase showed a slight downshift in its position (inset of the Fig. 2), indicating that the addition of Ce caused a distortion in the spinel structure. Accordingly, Ce-doping generated a growth in the spinel lattice parameter ( Table 2), which was attributed to the larger ionic radius of Ce4+ ions (101 pm) as compared to Co2+ (79 pm), Co3+ (69 pm) and Al3+ (67.5 pm) host ions. Similar features were obtained by others [37].The intensity ratio of the (220) to (440) diffraction planes, indicative of the cobalt ions in tetrahedral to octahedral sites [38], increased as Ce-content augmented (Table 2), indicating a preferential arrangement of cobalt ions in tetrahedral coordination [39]. The mean crystallite size of spinel phase (without distinction between CoAl2O4 and Co3O4) increased between 28% and 40% with Ce doping. The average crystallite sizes of segregated ceria phase could not be determined due to its weak signal. Undoubtedly, lattice changes would affect spinel structure.After reduction, diffraction peaks of spinel phase remained (Fig. 2B), indicating that the reduction temperature was not enough for complete reduction of all cobalt species. Upon reduction, diffraction peaks assignable to metallic Co emerged in all samples, whereas in sample 2.1CeCoAl, the formation of an alloy-like Ce5Co19 rhombohedral phase (PDF 026–1084) more stable structure was observed. The formation of this stable phase would explain the shift at higher temperatures of peaks III and IV observed in the H2-TPR study (see below). The mean crystallite size of metallic Co of Ce-doped catalysts (Table 2) were notably smaller than that of the pristine 0CeCoAl assay. It could be concluded that, Ce insertion in the cobalt aluminate spinel structure, strengthens the metal-support interaction in the reduced catalyst and contributes to a decrease of Co0 particles size.The morphology of the reduced materials was analysed by STEM micrographs. The obtained micrographs and the resulting particle size distribution histograms are depicted in Fig. 3. The chemical composition of the nanoparticles was also confirmed by EDX analysis. Cobalt nanoparticles (mixture of cube and cuboid shapes) in Ce-doped catalysts were homogenously distributed, as were the tiny cerium nanoparticles that were found completely scattered throughout the samples. Ce-rich domains were observed for 2.1CeCoAl assay, indicating some Ce-rich phase. Co0 particle size was measured by imageJ software, and a similar mean diameter (ca. 20 nm) was measured for all catalysts, suggesting they are composed by agglomerates of smaller crystallites. Regarding the particle size distribution, particle size ranged between 10 and 26 nm while 75% of the nanoparticles were smaller than 20 nm. Ce-modified solids showed a broader distribution curve, in line with the Co0 crystallite sizes evaluated by XRD (0.3CeCoAl: 9.8 nm; 2.1CeCoAl: 7.2 nm).XPS spectroscopy was employed to examine the oxidation state and the surface composition of xCeCoAl solids. The BE of the photoelectron peaks for both calcined and reduced samples are summarised in Table 3. The high-resolution XPS spectra for Co, Ce and Al in the calcined solids are shown in Fig. S1 (Supporting Information). It can be seen that the spin-orbit splitting (ΔBE) between Co 2p3/2 and Co 2p1/2 peaks was 15.21 eV for 0CeCoAl, which increased to 15.67 eV at the lowest Ce loading (0.3CeCoAl). Further increase in Ce doping decreased the ΔBE (i.e. 15.35 eV for sample 2.1CeCoAl). The satellite peak of Co 2p3/2 is located at approximately 4–5 eV higher BE than the main Co 2p3/2 peak, which is characteristic of Co2+ in CoAl2O4 framework [40,41]. In the Al 2p region, both the calcined and reduced samples show peaks at around 74 eV which can be assigned to Al3+ cations bonded with oxygen. The peaks could be deconvoluted into octahedral and tetrahedral Al3+ contributions, being the former at higher BE [42]. All the solids had mainly octahedral Al3+. In addition, in the fresh catalysts Co 2p3/2 band appeared shifted to lower binding energies as Ce loading increased, what suggested a stronger interaction between elements in Ce-doped solids, in agreement with H2-TPR results.Ce 3d peaks of the calcined solids (Fig. S1C, Supporting information) showed a 3d3/2 and 3d5/2 separation of around 18.2 eV, in good agreement with literature [43–45]. Unambiguously, no characteristic peaks of tetravalent Ce4+ were detected. Instead, characteristic peaks of Ce3+ were observed at around 902.7 ± 0.5 eV and 885.2 ± 0.7 eV. The latter result of Ce 3d94f1 O 2p6 final-state [43].The Co 2p XPS spectra of the reduced solids are displayed in Fig. 4. The most striking difference was the signal from metallic cobalt, absent for calcined solids. It appeared at 778.0 eV for 0CeCoAl catalyst and shifted to higher BE for samples xCeCoAl (ΔE = 0.7–1.1 eV), in agreement with their lower band gap energies. In addition, for the reduced xCeCoAl, deconvolved Co0 peak were less intense than for 0CeCoAl catalyst, in agreement with H2-TPR results.Two typical bands could be observed in the O 1s XPS spectrum of the reduced samples (Fig. S1D, Supporting Information). The peak at 530.9–531.6 eV corresponded to adsorbed oxygen (Oads) species and the other, located at lower BE coincided with surface lattice oxygen (Olatt) [46]. Oads species has greater mobility than Olatt, and are considered responsible for maintaining the charge balance in the structure [47,48]. The ratio Oads/Olatt significantly decreased after Ce loading (1.07–1.51 vs. 3.34) what implied a decrease in oxygen vacancies.A quantitative evaluation of the chemical composition of Co and Al at solid surface was done for both the calcined and reduced samples (Table 3). It is interesting to note that surface Co/Al ratio was lower than the bulk ratio measured by ICP-OES which was around 0.63 for all catalyst. This enrichment in Al could be attributed to its lower surface energy in comparison to Co [49], which generated cobalt-deficient aluminium-rich phase on the surface. Also, note that upon reduction, the Co/Al ratio decreased by half for the non-doped reference catalyst, whereas it remained constant for Ce-doped samples. The smaller metallic Co particle sizes of the later could be involved in preserving the Co/Al ratio at the surface. Fig. 5 shows the UV–vis DRS spectra of the calcined catalysts. Bare CeO2 sample, used as reference, showed two absorbance bands below 400 nm. Bands at 257–278 nm can be attributed to O2--Ce4+ charge transfer transitions involving Ce3+ (≈255 nm) and Ce4+ ions (≈278 nm) with different coordination numbers, whilst a characteristic vibration of interband transitions at 328 nm was observed [50]. The spectrum of 0CeCoAl sample was very similar to bare Co3O4, with a small blueshift of the bands corresponding to tetrahedrally coordinated Co2+ ions, indicative of Co-Al interaction [29,51]. Thereby, 0CeCoAl shows features from tetrahedral Co2+ (d–d transition bands at 1210, 1330 and 1500 nm; d–d absorption bands at 558, 585 and 621 nm) [52]. On the other hand, Ce-doped solids exhibited two broad bands in the UV region centred at around 420 and 635–644 nm and characteristic cobalt aluminate bands in the NIR region (1200–1500 nm). These bands were found to be less intense than those shown by pristine 0CeCoAl assay with a slight band shift from 1420 to 1460 nm, indicative of Ce incorporation to spinel.The direct band gap (Eg) for all samples was estimated by Tauc plot (Fig. S2, Supporting Information). Typically, Eg corresponds to a ligand-to-metal charge transfer excitation energy, and shows the tendency of adjacent transition-metal centers to gain electron density [53]. A band gap of 1.38 eV was estimated for pristine 0CeCoAl solid and decreased to 0.81 and 0.48 eV for samples with 0.3% and 2.1% of cerium, respectively. The lowering of Eg for Ce-doped samples could be attributed to the incorporation of metal cations into the framework of spinel crystal [54]. The distortion caused in the cell could act as trapping-centers to capture electrons [55]. Fig. S3 (Supporting information) shows the 27Al NMR spectra obtained from the reduced solids. Catalyst 0CeCoAl shows an asymmetric single peak, at 2 ppm, corresponding to octahedral aluminium. A subtle signal at around 60 ppm, corresponding to tetrahedral aluminium, could be also deduced [56]. Ce-doping caused a slight shift of main peak to the right, and the appearance of a bulge in the 25–40 ppm range, which can be ascribed to five-coordinated aluminium. Both features indicated an increase in the disorder degree in the Ce-containing solids [57], in agreement with XRD results. Fig. 6 displays the H2-TPR profiles of xCeCoAl solids. Those of bulk Co3O4 and CeO2 (5 times magnified) are also displayed as a reference. The reduction profile of bulk Co3O4 comprised two well defined reduction bands with maximum at 300 ºC and 425 ºC, and ascribed to Co3+ → Co2+ and Co2+ → Co0 reduction steps, respectively [58]. Bare CeO2 also showed two main reduction bands: the low temperature peak, centred at 496 °C, can be ascribed to the reduction of surface caps of ceria whereas the high temperature peak, centred at 845 °C, can be ascribed to the reduction of the innermost layers (bulk) of the ceria [59].The reduction of the pristine 0CeCoAl [29] started at around 150 ºC. Four reduction peaks could be identified. Peak I (at 292 ºC) was ascribed to Co3+ → Co2+ reduction of the surface cobalt cations without any interaction with the alumina or cobalt aluminate phases. Peak II (at 413 ºC) was also attributed to the reduction of Co3+ species though in close interaction with alumina or cobalt aluminate [29,60]. The reduction peak at 594 ºC (peak III) was assigned to Co2+ → Co0. Finally, peak IV (at 783 ºC) was assigned to the reduction of cobalt ions in the stoichiometric cobalt aluminate (CoAl2O4) phase [29,61]. The XRD signals from Co3O4 and CoAl2O4 are consistent with the peak assignment in H2-TPR measurement. It is interesting to note that the strong interaction between the cobalt ions and the support (mixture of alumina and cobalt aluminate) notably hindered both Co3+ → Co2+ and Co2+ → Co0 reduction stages. Therefore, peak III from 0CeCoCe assay significantly upshifted (by 159 ºC) as compared with bare Co3O4. This behaviour reflects the lower reducibility of Coδ+ species in the Ce doped assays and is in line with XPS data where BE of Co 2p3/2 decreased with Ce loading in the calcined series.In the Ce-doped samples, due to the low amount of CeO2 added, its reduction peaks were overlapped by the most intense cobalt reduction peaks. As for catalyst 0CeCoAl, Ce-doped samples showed four reduction peaks. After addition of a small amount of Ce (0.3 at%), peaks I and II, ascribed to Co3+ → Co2+, shifted to lower temperatures indicating a promotional effect of ceria on the partial reduction of both kinds of Co3+. However, such promotional effect was not observed in catalyst 2.1CeCoAl (see Table 4). Peak III (reduction of Co2+ to Co0) appeared at higher temperatures than those exhibited by 0CeCoAl, probably due to the hardening of Co-O-Al bonds because of the presence of electron trapping sites. Finally, the reduction of cobalt species in CoAl2O4 phase (peak IV), was again favoured at small Ce doping whereas in 2.1CeCoAl it took place at the highest temperature (813 ºC, 30 ºC higher than that of 0CeCoAl). This higher temperature requirement led to a lower degree of reduction of this sample after the activation process carried out before the catalytic experiments (reduction at 600 ºC). Overall, Ce-modified solids would achieve a smaller reduction than pristine 0CeCoAl, especially 2.1CeCoAl, worsened by the formation of Ce5Co19. Table 4 shows the results of H2-TPR analysis. Strictly speaking, H2 consumption due to ceria reduction should be subtracted. However, the maximum theoretical H2 consumption expected from ceria reduction in sample 2.1CeCoAl was 0.076 mmolH2·g-1, what represented 1% of the total H2 uptake. Thus, the total H2 consumption is given.The amount of exposed metal, calculated by isothermal H2 pulse chemisorption, is gathered in Table 5. Consistent with the reducibility results discussed above, Ce addition remarkably decreased the available cobalt. This depletion was around 36% at the lowest Ce loading, and reached up to 44% at 2.1% Ce.The activity of catalysts in the cyclohexane dehydrogenation model reaction was measured in terms of TOF values (Table 5). Measured values clearly indicate a higher activity of surface atoms in Ce-doped samples. The activity in cyclohexane dehydrogenation, expressed as TOFdehyd, increased between 42% and 52% for 0.3CeCoAl and 2.1CeCoAl catalysts respect to non-doped catalyst. UV–vis DRS analysis revealed differences in the inner electron transitions of the solids. Also, Ce-doped catalysts had smaller Co0 crystallites than pristine 0CeCoAl. Both features modified the metal-surroundings interaction, which could affect the intrinsic activity of cobalt in cyclohexane dehydrogenation. Moreover, the higher intrinsic activity of Ce-doped catalysts is also a desirable characteristic for activating H2 molecule in hydrogenolysis reactions.The surface acid site density of the reduced solids is summarised in Table 5, and the strength distribution, according to the desorbed NH3 profiles, is displayed in Fig. S4 (Supporting Information). The acid site strength were categorized according to their desorption temperatures, as follows: weak (desorption temperature range 90–300 °C), medium (300–650 °C) and strong (> 650 °C), while the percentage contribution of each strength site was estimated from the area under ammonia desorption profile. Ce-doping led to an increase in the density of surface acid sites (increase by 55–72%) and a change in the distribution of the acid strength. Ce-doped solids exhibited a higher density of medium and strong acid centers compared to their counterpart 0CeCoAl assay. In percentage terms, strong acid sites increased from 8.1% for sample 0CeCoAl to 26% for both Ce-containing samples. Morterra et al. [62] found that Ce-doped alumina had more acidity than parent γ-alumina, and ascribed this to the modified environment of Al3+ cations by cerium. In fact, 27Al NMR and UV–vis DRS results pointed to this.Cobalt aluminate is characterized by having Lewis acidity rather than Brønsted acidity [63]. Thus, it is expected the Lewis acid centers to prevail in Ce-doped solids. In glycerol APR conditions, dehydration of primary or secondary hydroxyl from glycerol depends on the abundance of Lewis or Brønsted acid sites [64]. Therefore, it was considered useful to evaluate the Brønsted acidity of the catalysts. Evaluation of Brønsted acidity was done in terms of the activity in the skeletal isomerization of 33DM1B. Table 5 showed large differences between 0CeCoAl and Ce-doped catalysts. Both Ce-containing catalysts were 70–90 times more active for skeletal isomerization of 33DM1B, indicating their acid sites had more acid Brønsted characteristics than 0CeCoAl, probably linked to their higher penta-coordinated Al content. In conclusion, the total acid sites density increased with Ce, which is a key feature to protonate the hydroxyl groups before oxygen loss as water.The Weisz-Prater and Mears criteria (Table S1, Supporting Information) was used to confirm the absence of mass-transfer limitations in catalytic experiments. Blank tests carried out with the reactor bed filled solely with quartz wool (used to support the catalyst inside the reactor) showed no glycerol conversion, which suggested that homogeneous APR had no contribution to the catalytic conversion of glycerol. Similarly, HDO reaction with calcined 0CeCoAl catalyst resulted in almost null conversion, which indicated that metallic function was required for the HDO reaction. Furthermore, catalytic experiments with reduced cobalt oxide (Co3O4) resulted in low liquid yield [29], indicating the acid sites are involved in the production of liquids.The catalytic performance in the glycerol HDO without external addition of hydrogen was studied in a continuous fixed bed reactor feeding a solution of 10 wt% glycerol/water, at WHSV of 24.5 h-1, during 3 h TOS. The results expressed as glycerol conversion, carbon yield and selectivity to liquids are shown in Fig. 7.Catalyst 0CeCoAl [29] exhibited higher glycerol conversion (95.7%) than Ce-doped catalysts. Among Ce-modified catalysts, that with the lowest cerium load showed slightly higher glycerol conversion (Xgly= 60.0% for 0.3CeCoAl vs 55.0% for 2.1CeCoAl). The low glycerol conversions obtained with Ce-containing catalysts might be related to the stabilizing effect of cerium on the Co2+ ion, delaying its reduction to Co0, as seen by H2-TPR analyses. Furthermore, and related to the above, available metallic area was lowered for Ce-doped catalyst.Carbon yield to liquid (Xliq) was also markedly lowered for Ce-containing catalysts, and decreased with increasing dopant loading. This trend is intrinsically linked with the glycerol conversion presented for these samples. Nevertheless, selectivity towards liquid products improved by 34.5% with the addition of 0.3% Ce and by 27.1% on 2.1% Ce added, evidencing that the selectivity to liquid phase products was undoubtedly favoured by doping with Ce. Fig. 8 A displays the yield of the main liquid products obtained. For all the catalysts, 1,2-propanediol was the foremost product generated, which indicated that H2 was indeed produced by APR. Hydroxyacetone, the major intermediate product in 1,2-propanediol formation, was the second leading liquid product. The high yields to both compounds, in addition to acetone (dehydration product of 1,2-propanediol), demonstrated high activity of the synthesized catalysts for the glycerol hydrodeoxygenation (HDO). Catalyst 0.3CeCoAl presented the maximum yield to 1,2-propanediol (34%) while 0CeCoAl and 2.1CeCoAl exhibited 28% and 25%, respectively. 0.3CeCoAl catalyst also manifested the highest yield to hydroxyacetone (22.6%). However, the utmost yield towards ethanol was achieved with pristine 0CeCoAl catalyst. The density and distribution of the acid sites is of important consideration in the production of by-products [65]. Ce-doped catalysts presented a higher density of medium and strong acid centers than pristine 0CeCoAl, which makes them better catalysts for dehydration reactions compared to bare cobalt aluminate catalyst that is conducive to dehydrogenation reactions.Besides the liquid products displayed in Fig. 8A, 1-propanol was detected for all the catalysts. Concerning 0CeCoAl catalyst, a bunch of products was detected: propionic acid, acetic acid and very low amounts of 2-propanol, acetaldehyde and methanol. Obtaining this great variety of liquid products could be explained by the highest metal availability, and therefore a high reactivity, of the 0CeCoAl catalyst, which could facilitate the concatenation of successive dehydration/hydrogenation reactions. These results showed the relevance of the Ce-doped catalysts for 1,2-propanediol production from glycerol HDO with in-situ generated H2. No liquids produced by Brønsted acid sites (acrolein, or 1,3-propanediol) were detected, indicating prevalence of the Lewis acidity.Catalyst activity for bond cleavage was also analysed. From data in Fig. 8B it is shown that carbon selectivity to primary products (products from single C–C or C–O bond scission, such as ethylene glycol, hydroxyacetone and propylene glycol) was remarkable high for Ce-doped catalyst, 0.3CeCoAl and 2.1CeCoAl (82.5% and 79.6%, respectively). In turn, catalyst 0CeCoAl achieved 44.1% selectivity towards primary products while its selectivity to secondary ones (products obtained after cleavage of additional C-C or C-O bonds) was 24%, which is three-fold higher than that of Ce-doped catalysts. Pristine cobalt aluminate catalyst also attained a considerable selectivity to gas products (24.6%) in accordance with its high metallic surface area. Among the Ce-doped samples, catalyst 2.1CeCoAl displayed slightly lower selectivity to primary and secondary products than catalyst with 0.3% of cerium.A summary of the reaction products is shown for the third hour TOS ( Table 6). Carbon selectivity to ethylene glycol (EG) was around 5% for both Ce-doped catalysts and 3.3% for pristine 0CeCoAl. Carbon selectivity to 1,2-propanediol (1,2-PD) step up in the following order: 0CeCoAl (28.9%) < 2.1CeCoAl (42.4) < 0.3CeCoAl (46.5). Ce-containing catalysts also presented a higher selectivity towards hydroxyacetone (HA) (31–32%) compared to 0CeCoAl (11.9%). According to these results, Ce-doped catalysts can balance between acid centers (necessary for obtaining hydroxyacetone via glycerol dehydration) and metal availability for the further hydrogenation of hydroxyacetone into 1,2-propanediol. Both Ce-doped catalysts have less available Co0, but of higher intrinsic activity as deduced from the cyclohexane dehydrogenation activities (Table 5). Ce-doped catalysts still have scope for improvement to decrease hydroxyacetone production in favour of 1,2-propanediol.The low selectivity to hydroxyacetone and 1,2-propanediol presented by 0CeCoAl catalyst was due to the formation of other liquid compounds and its proficiency to break C-C bonds. In this sense, the formation of light alcohol and short chain alkanes is an inconvenience to be solved for the commercial development of this process for primary products [13].The oxygen removal degree (DOR) reflects the ability of the catalysts to generate deoxygenated liquid compounds. Ce-doping enhanced DOR, as it passed from 60.7% for 0CeCoAl to 63.5–68.3% for 0.3 and 2.1CeCoAl. Liquid products were clustered into two main group. Products obtained by C–C bond cleavage only (ethylene glycol, methanol) and those classified as C–O bond scission products. The selectivity to both types of cleavage are given in Table 6. For all the catalysts, selectivity to C-O cleavage products (SC-O) was considerable high (i.e. up to 86.9% for catalyst 0.3CeCoAl). In the same vein, 0CeCoAl catalyst presented the lowest value (SC-O = 64.5%). In contrast, selectivity to SC-C barely reached 5% (0.3CeCoAl). We can affirm that these catalysts, and particularly Ce-containing samples, exhibited excellent C–O bond scission function. Thus, the presence of cerium favoured the dehydration pathway, by slight increase of surface acid sites density, even under conditions in which cobalt aluminate based-catalysts have shown to favour dehydrogenation route [29]. Table 6 also gathered hydrogen yield (YH2) and selectivity to alkanes (Salk). Regarding YH2, it follows the same trend as the liquid yield. It indicated that the obtained H2 readily reacts to yield liquid products. Pristine 0CeCoAl showed the highest YH2 and the lowest value was obtained for catalyst 2.1CeCoAl (10.2%). This catalyst, instead, displayed the maximum selectivity to alkanes (54%, mainly methane). In brief, selectivity towards alkanes increased with cerium content, which was evidenced by the decrease in YH2. It should be stressed that YH2 reflects the hydrogen consumed in side reactions for the formation of both liquid and gaseous compounds while Salk was calculated taking into account the flow of carbon atoms in alkanes with respect to the total carbon atoms only in gaseous products. These data lead us to conclude that, for catalyst 0CeCoAl, hydrogen produced in-situ is consumed mainly in the formation of diverse liquid products, while Ce-containing catalyst were more prone to the formation of alkanes.A slight decrease in XGly and Xliq was observed in the course of reaction (Fig. S5, Supplementary Information), more pronounced for Ce-doped catalysts. From 2 h to 3 h TOS, XGly decreased by 9–14% for Ce-containing catalysts vs 2% for 0CeCoAl assay. In the same period, conversion to liquid decreased by 4–9% for the doped catalysts, while for the non-doped assay increased by 4%, which was due to the decrease in the production of gaseous compounds while the conversion of glycerol remains more stable. In order to investigate the potential changes in the physico-chemical properties underwent by catalysts in the reaction, the characterization of spent catalysts was carried out, and the results obtained are shown in Table 7.N2 adsorption-desorption isotherms and BJH pore size distribution of the spent catalysts are shown in Fig. 1. The form of the isotherms of the spent catalysts resembled those of fresh reduced solids, indicating the pore structure was preserved. All the spent catalysts adsorbed more nitrogen than the parent fresh reduced solid, which, in turn, was reflected in the increase in SBET. Spent 0CeCoAl catalysts showed 75% increment in SBET, while Ce-doped catalysts incremented by 26–57%. Reynoso el at. observed that hydrated alumina species were involved in the specific surface area gain [29]. Under hydrothermal conditions γ-alumina could be hydrated to gibbsite or boehmite, which in turn, could be leached-off and re-deposited on the catalyst surface, generating additional porosity in the solid [66]. Differences between reduced and spent catalysts were appreciated in the BJH pore size distribution. Spent catalysts had bimodal distributions, with maxima at approximately 3–4 nm and 8–10 nm pores. The peak from smaller pores suggested the presence of new phases not observed in the fresh reduced form (e.g. gibbsite or cerium hydroxycarbonate). The former narrow peak was more pronounced for solids with higher amount of cerium. Many phenomena could explain this evolution of the textural properties, such as hydration of the alumina [67], deposits of carbonaceous materials [68] or the formation of new phases [69].Crystalline phases in the exhausted catalysts were identified by XRD ( Fig. 9A). The obtained diffractograms still showed the well-known characteristic peaks of spinel (either cobalt oxide and cobalt aluminate). XRD reflexions from metallic Co were still visible for all catalysts. The poor resolution of the spectra from those peaks prevented us to calculate the crystallite size. We could hypothesize that coalescense of metallic cobalt was not significant, despite of the high temperature used. A new peak at around 34º (2θ) emerged for all spent catalysts, which corresponded to cubic CoO phase (PDF 048-1719). This fact indicated that cobalt could be oxidized in the aqueous media of the reaction [70,71]. It could be concluded that the oxidation of cobalt was limited to the outmost caps of the particles, since XRD signal from bulk Co0 remained. None of the spent catalysts showed diffraction peaks from hydrated alumina (boehmite or gibbsite), which could be due to its amorphous form. For Ce-doped catalysts, new XRD peaks arose and became significant for 2.1CeCoAl catalyst. Those peaks centred at 24.6º and 30.6º were attributable to the hexagonal phase of cerium hydroxycarbonate (Ce(CO3)OH) (PDF 32-0189). This new phase was formed by carbonation of ceria by the CO formed, promoted by the harsh hydrothermal conditions [72]. Cerium hydroxycarbonate has been also detected in the spent Ni/CeO2 catalysts after usage in the aqueous-phase reforming of methanol [73]. The sharpness of these Ce-containing new phase peaks indicated that the product was well crystallized. All the Ce-containing spent catalysts showed peaks from Ce5Co19 phase structure, though weakened with respect the reduced form.Leaching of catalyst components contribute to the deactivation of heterogeneous catalysts in the liquid-phase reactions [22]. Accordingly, the catalysts metal constituents Al, Ce and Co in the reaction liquid product were analysed by ICP-MS. All spent catalysts experienced leaching of the three metals, to a different extent. At first glance, Ce-containing catalysts underwent greater leaching of all metals. Cobalt was the most leached metal and the percentage of leaching in the Ce-containing samples was 2–3 times higher than for catalysts 0CeCoAl. This behaviour could be due to the smaller size of metallic Co in the Ce-doped catalysts, which are more prone to oxidation and subsequent leaching [74]. Co was thermodynamically prone to be oxidized (preliminary stage of leaching), also the phase transition during alumina hydration could promote metal particle detachment. Due to the re-deposition of aluminium hydroxyde on the catalyst surface, leaching of aluminium was 1–2 less intense than cobalt. Meanwhile, the amount of Ce leached out from the catalysts increased in proportion to the amount of Ce in the samples.The spent catalysts were also subjected to H2-TPR analysis (Fig. S56, Supplementary Information) and the hydrogen uptake values are reported in Table 7. The hydrogen consumption was higher than the required for the reduction of solely cobalt species (deduced through the reducibility degree). Indeed, all samples showed hydrogen consumption below 600 °C, which indicated the catalysts were oxidized in the course of the reaction. It is also noteworthy the intense H2 consumption at temperatures above 600 °C. Actually, the used catalysts showed negligible consumption below 500 °C. Other authors [75] have investigated reduction temperature range after and before reactions, concluding that spent catalysts show a shift towards higher temperatures compared to fresh one due to the aggregation or sintering of metal clusters.Overall, the metallic availability of all catalysts was seriously compromised, as was also confirmed by H2 chemisorption that revealed a decrease of about 64% for catalyst 2.1CeCoAl and up to 96% for monometallic 0CeCoAl. One plausible explanation is the re-deposition of aluminium on the catalyst surface, covering the Co0 sites by hydrated alumina [76,77]. Two other phenomena that cannot be ignored are the oxidation of cobalt and the preferential leaching of the smallest particles, thus remaining the largest particles, consequently reducing accessible metal atoms [78,79].The results obtained by TPH, displayed in Table 7, present a very low deposition of carbonaceous material, which decreased with the Ce-content. The recorded Raman spectra of the used catalysts (Fig. 9B) exhibited the typical peaks from carbon, named D and G bands, at ca. 1350 and 1590–1600 cm-1, respectively. The shape of the spectra was similar for all catalysts with a broad and less intense D-band and a more intense G-band. The shoulder observed at 1710 cm-1 in the sample 2.1CeCoAl could be attributed to an overtone of an M point phonon [80]. The degree of graphitization of carbonaceous materials was estimated by the relative intensity ratio of D and G bands (ID/IG inset in Fig. 9B). The ID/IG values obtained were in the 0.22–0.38 range and did not show correlation with the composition of the samples. Raman and TPH agree with the absence of any peak from coke in the XRD pattern. It seems that the principal deactivation issues are those related to the loss of metal surface area, due to the encapsulation by alumina, and the oxidation of cobalt species. Cobalt oxidation could lead to metal leaching, another phenomenon responsible for the decrease in the metal surface.The effect of Ce-coprecipitation to improve the activity and selectivity of cobalt aluminate-based catalysts for glycerol HDO was discussed. Properties of catalysts based on cobalt aluminate spinels were tuned through Ce doping, a simple, fast and cost-effective synthesis method. Ce-modification of the 0CeCoAl catalyst increased SBET (by around 16–30%) and also the spinel crystallite size. For the reduced sample, in addition to cobalt aluminate and metallic cobalt species, formation of Ce5Co19 phase was identified at the highest Ce-content. Also, Ce-doping increased total acidity through a prominent increment of medium-strong acid centers.Ce-doped catalysts exhibited lower glycerol conversion than their counterpart, 0CeCoAl. Nevertheless, their higher acid sites density have influenced the activity of these catalysts, presenting a higher selectivity towards deoxygenated liquid products. Catalyst testing in a fixed-bed reactor at 260 °C and 50 bar without external hydrogen source revealed that: (i) H2 is produced; and (ii) it is in part consumed hydrogenating liquid intermediates, which makes these catalysts to have advantageous catalytic properties for glycerol HDO in a sustainable way. The main products of glycerol conversion were 1,2-propanediol and hydroxyacetone. Selectivity to liquid products of catalyst 0.3CeCoAl was 92% and selectivity to 1,2-propanediol was over 46%, indicating that Ce-doped catalysts are compelling HDO catalysts.Future research should focus on stability issues since post-reaction characterization verified a decrease in cobalt availability that can be attributed to re-deposition of hydrated alumina on the catalyst surface. This phenomenon, in conjunction with the deposition of carbonaceous material, could be responsible for the increment in SBET and the change in the pore size distribution. Further confirmed fact was that despite the addition of cerium, the catalysts continued to exhibit metal oxidation and leaching phenomena. A.J. Reynoso: Investigation, Writing – original draft. U. Iriarte-Velasco: Formal analysis, Writing – review & editing. M.A. Gutiérrez-Ortiz: Resources, Funding acquisition, Supervision. J.L. Ayastuy: Funding acquisition, Conceptualization, Writing – review & editing.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This research was supported by grant PID2019-106692EB-I00 funded by MCIN/AEI/10.13039/501100011033. The authors thank for technical support provided by SGIker of UPV/EHU and European funding (ERDF and ESF). Cyclohexane dehydrogenation and isomerization of 33DM1B model reaction were conducted at the Institut de Chimie des Milieux et des Matériaux, Université de Poitiers, (CNRS UMR 7285 IC2MP). The authors acknowledge the kind assistance of Dr. Catherine Especel and Dr. Laurence Vivier.Supplementary data associated with this article can be found in the online version at doi:10.1016/j.jece.2022.107612. Supplementary material .

In this study, the hydrodeoxygenation (HDO) of glycerol with in-situ produced H2 , via aqueous-phase reforming, was investigated over Ce-doped CoAl2O4 catalysts synthesized by coprecipitation. The catalytic runs were performed at 50 bar/260 ºC for 3 h TOS in a fixed bed reactor. The synthesized catalysts were extensively characterized to better understand the effect of physicochemical and surface characteristics on the catalytic performance. The results revealed that doping with Ce increases the population and strength of acid centers, which results in a higher selectivity towards deoxygenated liquid products. Ce-doped catalysts exhibited higher yield to hydroxyacetone and 1,2-propanediol. Post-reaction characterization revealed a decrease of the metallic surface area, mainly due to alumina coating, and to a lesser extent, due to the oxidation and leaching of the cobalt.

With the increasing demand for energy and the need to decrease the use of conventional energy sources, the development and utilization of renewable energy becomes extremely important. Another aspect is the environmental pollution due to the use of traditional energy sources, so it is urgent to develop clean and efficient energy systems. Hydrogen is one kind of efficient and clean energy carriers, which has attracted extensive worldwide attention in recent years [1–3]. One of the key technologies for hydrogen energy development and utilization is how to storage hydrogen safely. Magnesium hydride (MgH2) is one of the most suitable candidates for hydrogen storage materials because of its high gravimetric energy density. Unfortunately, the relatively high operating temperature and the slow de/hydrogenation rate hinder its extensive industrial application. Therefore, the hydrogen storage properties of MgH2 need to be further improved to overcome these drawbacks.One of the effective ways for reducing the de/hydrogenation temperature and improving the de/hydrogenation rate of MgH2 is the introduction of catalysts, such as transition metals [4–9], intermetallic compounds [10–12], transition metals oxides [13–18], and carbon materials [19–26]. Zhang et al. [4] introduced Fe into MgH2 by wet-chemical ball milling and found that the onset temperature of desorption and the apparent activation energy of dehydrogenation of MgH2 + 5 wt.% Fe were 182.1 °C and 40.7 ± 1.0 kJ/mol, respectively, thus being far lower than the values for pure MgH2. Korablov et al. [6] reported that the activation energy of Ti-doped MgH2 is 53.6 kJ/mol and the investigated 0.75Mg-0.25Ti composite can absorb hydrogen at room temperature. Yang et al. [10] used FeCo nanosheets to enhance the hydrogen storage properties of MgH2 and observed that FeCo-containing MgH2 can rapidly uptake 6.7 wt.% of H2 within one minute at 300 °C. The hydrogenation and dehydrogenation activation energies of FeCo-containing MgH2 were reduced to 65.3 ± 4.7 kJ/mol and 53.4 ± 1.0 kJ/mol, respectively. Transition metal oxides have also been proved to be promising alternatives for improving the hydrogen storage properties of MgH2. Wang et al. [14] doped 9.0 wt.% of V2O3@C in MgH2 by mechanical milling and demonstrated that the presence of V weakens the strength of Mg–H bonds in MgH2; thus, the hydrogenation and dehydrogenation temperatures of V-catalyzed MgH2 are strongly reduced. Valentoni et al. [17] reported that VNbO5-doped MgH2 can adsorb more than 5 wt.% of H2 within five minutes at 160 °C and its hydrogen storage capacity does not decline even after 70 hydrogen absorption-desorption cycles. It has also been shown that carbon materials can significantly improve the thermodynamic and kinetic properties of MgH2. For example, Liu et al. [21] introduced one-dimensional bamboo-shaped carbon nanotubes (BCNTs) with a high specific surface area to improve the integrated hydrogen storage properties of MgH2 and found that the dehydrogenation activation energy (97.97  kJ/mol) and enthalpy (68.92 kJ/mol) of MgH2@BCNTs were reduced by 111.24 kJ/mol and 6.07 kJ/mol as compared to those of pristine MgH2, respectively. Zhang et al. [25] noticed that the hydrogen absorption/desorption rates of TiO2@C-containing MgH2 are significantly faster than those of pristine MgH2 and the Mg-H bond strength weakens under the catalytic action of TiO2@C. The above results imply that although the operating temperature and hydrogen storage kinetics of MgH2 can be significantly improved by adding catalysts, this approach still cannot meet the requirements of practical applications.In recent years, two-dimensional (2D) materials were used as catalysts to improve the hydrogen storage properties of MgH2. Liu et al. [27] used two-dimensional Nb4C3T x to enhance the hydrogen storage properties of MgH2 and asserted that the hydrogen storage kinetics and thermodynamics of MgH2 are improved by the unique layered structure of in-situ-formed NbH x . Wang et al. [28] prepared NbTi nanocrystals using a NbTiC solid-solution as precursor and revealed that NbTi-containing MgH2 starts to release H2 at 195 °C and desorbs about 5.8 wt.% of H2 within 30 min at 250 °C. Liu et al. [29] found that 5 wt.%Ti3C2-containing MgH2 has excellent hydrogen storage kinetics and can uptake 6.1 wt.% of H2 within 30 s at 150 °C. Furthermore, the hydrogenated samples can absorb hydrogen at room temperature. In the periodic table, V and Ti are neighboring elements; thus, they have similar electronic structures and also manifest some similar catalytic effects on the hydrogen storage performance of MgH2. Theoretical calculations have revealed that the heat of formation for dehydrogenated V-containing MgH2 is −43.42 kJ/mol, which is 7.85 kJ/mol lower than that of Ti-containing MgH2 [30]. da Conceição et.al. [31] studied results indicated that VC could enhance the hydrogen absorption and desorption kinetics properties of Mg, and a desorption rate of 1.0  ×  10−2 wt.% s−1 at 300 °C was obtained for VC-catalyzed Mg system. In our previous study [32], it was found that the hydrogen storage properties of V-doped Mg-Al alloys were better than those of the Ti-doped samples. It was also reported that the catalytic effect of V-based compounds on the hydrogen storage performance of Mg-based alloys is superior to that of single V [33]. Therefore, it can be speculated that V-based 2D materials might have some unexpected catalytic effects on the hydrogen storage properties of MgH2. Hence, in this research, VC was synthesized using V4AlC3 as a precursor and employed to improve the hydrogen storage performance of MgH2.V4AlC3 (1.5 g) (purity≥98%, Beijing Beke New Material Technology Co., LTD) was slowly poured into 40 ml of 40% HF (purity≥99%, Aladdin) under magnetic stirring for 96 h at 55 °C. The resulting product was washed with deionized water more than three times until the pH value became ≥6. Finally, VC was obtained by drying the washed sample for 24 h in a frozen drying oven. Subsequently, 10 wt.% of VC was incorporated into commercial MgH2 (purity≥98%, Langfang Bede Trading Co., LTD) by milling a mixture of MgH2 and 10 wt.% of VC for five hours at a milling speed of 500 rpm (ball-to-power weight ratio = 40:1) in an Ar atmosphere; the product was named MgH2-VC.The phase compositions of the samples were characterized by X-ray diffraction (XRD; Miniflex 600, Rigaku) under Cu-Kα radiation at 40 kV and 200 mA with a scanning step size of 5°/min. The micromorphologies of the samples were examined by field-emission scanning electron microscopy (FE-SEM; SU8020, HITACHI) and transmission electron microscopy (TEM; FEI Tecnai G2, f20 s-twin 200 kV). The distributions of V and C in the VC-doped samples were characterized by energy-dispersive X-ray spectrometry (EDS) coupled with SEM and TEM. The surface chemical bonding structures of the samples were characterized by an X-ray photoelectron spectroscope (XPS; ESCALAB 250Xi Microprobe), and binding energy spectra were fitted via XPSPEAK41 software. The de/hydrogenation performances of the samples were analyzed by a Sievert-type device. The activated samples were heated from room temperature to 380 °C at a rate of 1 °C·min−1 under a hydrogen pressure of 6 MPa during hydrogenation and at a rate of 0.5 °C·min−1 under a hydrogen pressure of 0.01 MPa during dehydrogenation. The dehydrogenation and hydrogenation kinetics of the samples were determined at different temperatures under a hydrogen pressure of 6 MPa and a vacuum pressure of 0.01 bar, respectively.Dehydrogenation simulations were performed with the Vienna Ab Initio Simulation Package (VASP) [34–36]. The interaction between electrons and ions and the exchange-correlation effect were analyzed using the projector augmented wave (PAW) method of Blöchl [37] and the Perdew-Burke-Ernzerhof (PBE) functional [38] under generalized gradient approximation (GGA) [39], respectively. The cutoff energy of the plane wave basis set was set to 500 eV. The convergence criteria for the Hellmann–Feynman force and the total energy were 1  × 10−2eV/Å and 1  × 10−4 eV/atom, respectively. The dipole correction along the surface normal was also considered. The Monkhorst-Pack method with 5  ×  5  ×  1 k-point meshes was employed for the dehydrogenation of Mg4H8 clusters on the VC (100) surface. Fig. 1 displays a XRD pattern and a SEM image of VC, and SEM-EDS images of VC-doped MgH2. It is noticeable that Al in V4AlC3 was corroded in 40% HF (Fig. 1a), implying the successful synthesis of VC. The sharp diffraction peaks at 2θ = 37.4°, 43.4°, 63.1°, 75.6°, and 79.7° (Fig. 1a) correspond to the (111), (200), (220), (311), and (222) lattice planes of VC (JCPDS card no. 74–1220). An additional diffraction peak appears at low-angles stemming from the (002) lattice plane of the graphite nitrate (JCSDS, card No. 742330). The high-resolution V 2p and C 1 s XPS spectra of VC are presented in Fig. 1b and 1c, respectively. The binding energies of 520.7 eV (V 2p3/2) and 513.3 eV (V 2p1/2) with an energy gap of 7.4 eV can be assigned to V-C bonds, and the peak at 522.4 eV (V 2p1/2) and 514.9 eV (V 2p3/2) correspond to the V-T x bond (T x = –O, OH, and –F) [40,41]. The peaks at 282.3 eV, 284.8 eV, 286.2 eV, and 288.6 eV are due to C-V, C-C, C-O and O-C=O bonds (Fig. 1d), respectively [41,42]. The microtopography of VC is presented in Fig. 1(d–f). Apparently, VC possesses an accordion-like layered structure. These results suggest that VC was successfully synthesized using V4AlC3 as a precursor. After VC was introduced into MgH2 sample, numerous small particles attached to the large particles in VC-doped MgH2 composite (Fig. 1g), and EDS mapping image indicates that VC had been dispersed in MgH2 after ball milling (Fig. 1h).The microstructure of 2D VC was further examined by TEM, HRTEM, SAED, and EDS. The TEM image in Fig. 2 a confirms the synthesis of nanoscale VC. The distance between crystal faces in VC was calculated as 0.208 nm (Fig. 2b), which corresponds to the (200) lattice plane of VC. The SAED pattern in Fig. 2c reveals diffraction rings from the (111), (200), (220), (311), and (222) lattice planes of VC. The EDS mapping image indicates that V and C are homogeneously distributed in VC. The above results suggest that VC was successfully prepared by the etching method using V4AlC3 as a precursor. Moreover, Fig. 2g reveals that many small particles clustered around the large particles in VC-doped MgH2 composites, which is consistent with the observation from Fig. 1g. In Fig. 1g, the size of the large particles is in the range of 1.0um∼10.0um, the small ones are at the submicron level. In Fig. 2g, the grains are about 200 nm in size, and the small ones are about 50 nm. The distances between crystal faces of 0.219 nm and 0.210 nm can be assigned to the (111) plane of MgH2 and the (200) plane of VC, respectively. In addition, diffraction rings associated with the (002), (101), (212) planes of MgH2 and the (200) plane of VC are visible in Fig. 2i. Therefore, the SEM-EDS, HRTEM and SAED results are in good agreement with the XRD observations, suggesting that VC and MgH2 were well mixed during the preparation process. Fig. 3 (a, b) display the isothermal and non-isothermal hydrogenation and dehydrogenation curves of MgH2 and VC-doped MgH2. MgH2 and VC-doped MgH2 can absorb approximately 6.98 wt.% and 6.42 wt.% of hydrogen, respectively, when heated from room temperature to 350 °C. Undoped MgH2 can hardly absorb hydrogen until the temperature reaches 125 °C. However, dehydrogenated VC-doped MgH2 can absorb hydrogen even at room temperature, and its hydrogen absorption capacity is about 5.0 wt.% when the temperature reaches 150 °C. It was found that undoped MgH2 and VC-doped MgH2 release about 6.95 wt.% and 5.75 wt.% of hydrogen, respectively, when heated from room temperature to 400 °C. However, the initial dehydrogenation temperatures of these two samples are different. Undoped MgH2 cannot release hydrogen until the temperature reaches 320 °C. The onset dehydrogenation temperature of VC-doped MgH2 significantly decreases after the incorporation of VC, and it starts to release hydrogen at 170 °C, which is 150 °C lower than for undoped MgH2. The operating temperature of VC-doped MgH2 is also found to be lower than those of TiO2@C-doped MgH2 [43], FeB@CNTs-doped MgH2 [44], SrTiO3-doped MgH2 [45], and other materials [46,47]. To investigate the catalytic effect of VC on the hydrogen absorption/desorption kinetics of MgH2, the isothermal hydrogen absorption/desorption kinetics of undoped and VC-doped MgH2 were analyzed at various temperatures (Fig. 3(c–f)). Undoped MgH2 can absorb approximately 4.0 wt.% of H2 within 180 min at 175 °C and 6.0 wt.% of H2 within 180 min at 200 °C, respectively, and its hydrogen absorption capacity increases to about 6.2 wt.% for temperatures beyond 225 °C. In contrast, VC-doped MgH2 exhibits excellent hydrogenation properties and can absorb about 3.0 wt.% of H2 at 25 °C within 180 min, and its hydrogen absorption capacity increases as the temperature increases. It can absorb approximately 4.0 wt.% of H2 within 180 min at 50 °C, and the hydrogen absorption capacity increases to about 5.5 wt.% when the temperature increases above 100 °C. In particular, VC-doped MgH2 can absorb 5.0 wt.% of H2 within 9.8 min at 100 °C. Hence, the catalytic effect of VC on the hydrogen absorption performance of MgH2 is superior to that of Ni-V [48], NiS [49], Ni@rGO [5], Fe3S4 [50], and Co@C [51].For example, Mg-Ni-V [48], Mg-5 wt.%NiS [49], and MgH2 Co@C [51] can only absorb 1.0 wt.%, 3.5 wt.%, and 2.71 wt.% of hydrogen within 10 min at 100 °C, respectively. During dehydrogenation, undoped MgH2 can desorb approximately 7.0 wt.% of hydrogen at 375 °C at a hydrogen desorption rate of 1.06 ± 0.02 wt.%/min, and the hydrogen desorption rate decreases to 0.42 ± 0.04 wt.%/min and 0.13 ± 0.01 wt.%/min at 350 °C and 325 °C, respectively. VC-doped MgH2 can release approximately 6.0 wt.% of hydrogen at 325 °C at a hydrogen desorption rate of 2.18± 0.03 wt.%/min, which is 16.8 times faster than what was found for undoped MgH2 under the same conditions. The dehydrogenation rates of VC-doped MgH2 at 300 °C, 275 °C, and 250 °C are 1.32 ± 0.02 wt.%/min, 0.54 ± 0.02 wt.%/min, and 0.26 ± 0.04 wt.%/min, respectively; thus, desorption proceeds even faster than for undoped MgH2 at 325 °C (0.13 ± 0.01 wt.%/min). At 300 °C, VC-doped MgH2 can desorb 5.0 wt% of H2 within only 3.2 min. The dehydrogenation kinetics of VC-doped MgH2 are also better than those of Mg-Ni-V [48], Mg-5 wt.%NiS [49], and MgH2-Co@C [51]. Therefore, both the hydrogenation/dehydrogenation properties and hydrogen absorption/desorption kinetics of MgH2 are significantly improved after the addition of VC.The apparent activation energy is an important parameter to evaluate the hydrogen adsorption and desorption kinetics of materials. It can be calculated by the Johnson-Mehl-Avrami (JMA) equation (ln [−ln(1−α)] = nlnk + nlnt, where α, k, t and n are the phase transformation fraction, a temperature-dependent kinetic parameter, the reaction time and the order of the reaction, respectively [52]). According to the slopes of the ln k ∼ 1000 / T plots shown in Fig. 3(g, h), the hydrogenation apparent activation energy (E abs) values of VC-doped MgH2 and undoped MgH2 were calculated as 42.4 ± 1.4 kJ/mol and 77.3 ± 3.0 kJ/mol, respectively. In addition, the dehydrogenation apparent activation energy (E des) values of VC-doped MgH2 (89.3 ± 2.8 kJ/mol) and undoped MgH2 (138.5 ± 2.4 kJ/mol) were also obtained from the linear relationship of the lnk versus 1000/T plots, revealing that the apparent activation energy for hydrogenation and dehydrogenation of MgH2 is significantly reduced after the addition of VC. Thus, VC is an efficient catalyst to improve the hydrogen absorption/desorption properties of MgH2. Table 1 represents the empirical dehydrogenation apparent activation energies of catalyzed common MgH2 systems. It can be seen from the table that the dehydrogenation activation energy of oxides or Nb-based compounds as catalysts was higher than that of VC as catalysts. For example, the dehydrogenation activation energy of VNbO5-catalyzed (99.0 kJ/mol) [17], KNbO3-catalyzed (93.6 kJ/mol) [54], NbN-catalyzed (113.9 kJ/mol) [55], MnFe2O4-catalyzed (108.4 kJ/mol) [59] MgH2 was 9.7 kJ/mol, 4.3 kJ/mol, 24.6 kJ/mol and 19.1 kJ/mol lower than that of VC-catalyzed MgH2. In addition, the dehydrogenation activation energy of VC-catalyzed MgH2 system also lower than those of K2SiF6-catalyzed, Ni@pCNF-catalyzed, HfCl4- catalyzed and some others catalyzed MgH2 systems listing in Table 1. Obviously, the introduction of VC into MgH2 remarkably lowered the hydrogen desorption energy barrier.To further analyze the hydrogen storage properties of MgH2 and VC-doped MgH2, pressure-composition-isotherm (PCI) measurements were performed at various temperatures (Fig. 4 a and c). The hydrogen absorption and desorption plateau increases with the increasing temperature. Undoped MgH2 causes completely reversible hydrogen absorption/desorption at 350, 375, and 400 °C; however, it cannot undergo a reversible hydrogen storage process at 325 °C. The reversible hydrogen storage capacity of undoped MgH2 is approximately 6.7 wt%. After doping with 10 wt.% of VC, although the reversible hydrogen storage capacity of VC-doped MgH2 is reduced to about 5.8 wt.%, completely reversible hydrogenation/dehydrogenation is observed at 300, 325, 350, and 375 °C. Moreover, the reversible hydrogenation/dehydrogenation temperatures for MgH2 remarkably decrease after the addition of VC. The hydrogenation/dehydrogenation enthalpies of undoped and VC-doped MgH2 samples were estimated by the Van't Hoff equation ( ln P = Δ H R T − Δ S R , where P is the hydrogenation/dehydrogenation plateau pressure, R is the universal gas constant, ΔH is the reaction enthalpy, ΔS is the reaction entropy, and T is the hydrogenation/dehydrogenation temperature). The linear relationships between lnP and 1/T for the hydrogenation and dehydrogenation processes are plotted in Fig. 4b and 4d. The hydrogenation and dehydrogenation enthalpies of VC-doped MgH2 were calculated as 71.6 ± 2.8 kJ/mol H2 and 74.7 ± 0.8 kJ/mol H2, respectively. These values are about 2.1 kJ/mol and 2.4 kJ/mol lower than those of undoped MgH2 (73.7 ± 2.7 kJ/mol and 77.1 ± 5.3 kJ/mol), respectively. The decrease of reaction enthalpy is responsible for the decrease of the hydrogenation and dehydrogenation temperature of MgH2, indicating that the addition of VC dramatically improves the hydrogen storage properties of MgH2.To investigate the role of VC in the improvement of the hydrogen storage performance of MgH2, XRD characterization of hydrogenated/dehydrogenated VC-doped MgH2 specimens was performed. For comparison, the XRD characterization results for VC-doped MgH2 and undoped MgH2 are also presented in Fig. 5 . The diffraction peaks of rehydrogenated MgH2 can be assigned to MgH2 (Fig. 5a). The crystal structure of VC does not change during milling and hydrogen absorption and desorption processes, implying that VC remains stable and only acts as a catalyst during these processes (Fig. 5(b–d)). In addition, the diffraction peaks of VC-doped MgH2 became sharp after dehydrogenation, indicating an increment in the crystallization.It has been proved that VC acts as a catalyst during the dehydrogenation of MgH2. Moreover, the reflections of the VC (200) lattice planes can be clearly detected from its XRD Fig. 1a) and SAED (Fig. 2b) patterns. Hence, in the present study, the VC (100) surface consisting of 80 atoms with five atomic layers was constructed and a vacuum layer of 15 Å was used between the slabs. An adsorption model of an Mg4H8 cluster on the VC (100) surface was built to simulate the role of VC in the dehydrogenation process of MgH2 (Fig. 6 ). The geometries of the adsorbate and VC layers were completely relaxed, except for the bottom three layers, during DFT calculations. To reveal the dehydrogenation behavior of VC-doped MgH2, the dehydrogenation enthalpy and electronic structure of MgH2 on the VC (100) surface were studied. It should be noted that dehydrogenation energy affects the dehydrogenation rate and dehydrogenation temperature of hydrogen storage materials. To analyze the catalytic effect of VC on the dehydrogenation properties of MgH2, the dehydrogenation energy of the pure Mg4H8 cluster and a Mg4H8 cluster on the VC (100) surface were calculated by Eqs. (1) and ((2), respectively. In the Mg4H8 cluster, Mg atoms were arranged in a tetrahedron structure, that is, two kinds of H atoms bonded with one Mg atom and three Mg atoms each, which are labeled as Htop and Hface (Fig. 1), respectively: (1) Δ H d e ( M g 4 H 8 ) = E ( M g 4 H 8 − x ) + x 2 E ( H 2 ) − E ( M g 4 H 8 ) (2) Δ H d e ( VC + M g 4 H 8 ) = E ( VC + M g 4 H 8 − x ) + x 2 E ( H 2 ) − E ( VC + M g 4 H 8 ) Here, E(Mg4H8) and E(VC + Mg4H8) are the total energies of the Mg4H8 cluster and of the Mg4H8 cluster on the VC (100) surface, respectively, E(Mg4H8 − x ) and E(VC + Mg4H8 − x ) are the total energies of the Mg4H8 cluster and of the Mg4H8 cluster on the VC (100) surface with the desorption of x number (x = 1, 2, 8) of H atoms, respectively, and E(H2) is the total energy of gaseous H2. Fig. 7 presents the dehydrogenation energy for the desorption of one H atom (Htop or Hface), two H atoms, and eight H atoms in the Mg4H8 cluster and the Mg4H8 cluster on the VC (100) surface. It was found that in all cases, the dehydrogenation energy of Mg4H8 is improved by VC, causing a decrease in the dehydrogenation temperature of VC-doped MgH2. The catalytic effects of the Mg4H8 cluster on the VC (100) surface were also revealed. For the desorption of one H atom, Htop atoms require a lower dehydrogenation energy than Hface atoms. For the desorption of two H atoms, Hface and Htop atoms require the lowest dehydrogenation energy. This finding confirmed that the addition of VC is beneficial to weaken interaction between Mg and H. Further, to reveal the interactions between MgH2 and the VC (100) surface, the total density of state (DOS), the partial density of states (PDOS), and the electron density difference of Mg4H8 on the VC (100) surface were calculated (Fig. 7). It is evident from the DOS shown in Fig. 7a that the s orbitals of Mg are located at −4.0 eV below the Fermi level and are hybridized with the d orbitals of V atoms and the p orbitals of C, indicating a strong interaction between Mg and VC. The yellow (blue) areas in Fig. 8 (b) indicate the increase (decrease) of electron density. It is discernible from Fig. 8(b) that mass charge depletion areas appear around Mg atoms, whereas charge accumulation areas are found between Mg4H8 and the VC (100) surface, indicating that electrons of the s and p orbitals of Mg are transferred to the VC (100) surface, weakening the bonding between Mg and H atoms. The lengths of Mg-Htop and Mg-Hface bonds in Mg4H8 on the VC (100) surface are longer than those in Mg4H8. In the pure Mg4H8 cluster, the lengths of Mg-Htop and Mg-Hface bonds are 1.70 Å and 1.99 Å, respectively, and the corresponding values for Mg4H8 on the VC (100) surface increase to 1.82 Å and 2.07 Å, respectively. These findings indicate that the elongation of Mg-H bonds promotes the dehydrogenation of MgH2.VC was successfully synthesized by an etching method and employed to improve the de/rehydrogenation of MgH2. VC imparts superior catalytic effects on the hydrogen storage thermodynamics and kinetics of MgH2. VC-doped MgH2 can absorb hydrogen at room temperature and release hydrogen at 170 °C. Non-isothermal hydrogenation tests revealed that undoped MgH2 can hardly absorb hydrogen until the temperature reaches 125 °C, which is far higher than for the VC-doped samples. Isothermal re/hydrogenation measurements indicate that VC-doped MgH2 can absorb 5.0 wt.% of H2 within 9.8 min at 100 °C and desorb 5.0 wt.% of H2 within 3.2 min at 300 °C. At 325 °C, VC-doped MgH2 can release approximately 6.0 wt.% of H2 with a hydrogen desorption rate of 2.18 ± 0.03 wt.%/min, which is 16.8 times faster than for undoped MgH2 under the same conditions. The Ea of VC-doped MgH2 is 89.3 ± 2.8 kJ/mol, which is about 49.2 kJ/mol lower than that of undoped MgH2. Mass charge depletion areas were detected around Mg atoms, whereas charge accumulation areas were found between Mg4H8 and the VC (100) surface, weakening the bonding between Mg and H atoms. The Mg-H bond length on the VC (100) surface is significantly longer than that in MgH2. The elongation of Mg-H bonds promotes the dehydrogenation of MgH2; thus, the hydrogen storage properties of MgH2 are remarkable improved through addition of VC.This work was supported by the National Natural Science Foundation of China (Grant Nos. 52261038 and 51861002), the Natural Science Foundation of Guangxi Province (Grant No. 2018GXNSFAA294125), and the Innovation-driven Development Foundation of Guangxi Province (Grant No. AA17204063). J.E. acknowledges additional support by the Ministry of Science and Higher Education of the Russian Federation in the framework of the Increase Competitiveness Program of NUST "MISiS" (grant number K2-2020-046). We thank LetPub (www.letpub.com) for its linguistic assistance during the preparation of this manuscript.

Hydrogen is considered one of the most ideal future energy carriers. The safe storage and convenient transportation of hydrogen are key factors for the utilization of hydrogen energy. In the current investigation, two-dimensional vanadium carbide (VC) was prepared by an etching method using V4AlC3 as a precursor and then employed to enhance the hydrogen storage properties of MgH2. The studied results indicate that VC-doped MgH2 can absorb hydrogen at room temperature and release hydrogen at 170 °C. Moreover, it absorbs 5.0 wt.% of H2 within 9.8 min at 100 °C and desorbs 5.0 wt.% of H2 within 3.2 min at 300 °C. The dehydrogenation apparent activation energy of VC-doped MgH2 is 89.3 ± 2.8 kJ/mol, which is far lower than that of additive-free MgH2 (138.5 ± 2.4 kJ/mol), respectively. Ab-initio simulations showed that VC can stretch Mg-H bonds and make the Mg-H bonds easier to break, which is responsible for the decrease of dehydrogenation temperature and conducive to accelerating the diffusion rate of hydrogen atoms, thus, the hydrogen storage properties of MgH2 are remarkable improved through addition of VC.

Until date, the majority of automobile transportation uses traditional oil fuels and releases a significant amount of CO2 polluting the environment significantly. Thus, the depletion of fossil fuels (including petroleum, coal, and natural gas) and increasing energy demand are the essential challenging problems that cause for the efficient designing of energy storage devices and discovering of alternative earth abundant energies [1–12], while several renewable energy production technologies being developed as alternative energy resources, such as solar, wind, hydrothermal and other renewable energy sources [13]. However, there is a contrast between renewable energy power production and storage technology. The European Union has installed renewable energy storage capacity of 50 GW, which corresponds to only about 5% of our daily generation [14]. The US department of energy launched the ‘‘EV Everywhere Grand Challenge’’ (EV: Electric vehicle) as an initiative for improved batteries with dramatically reduced cost and weight, aimed at producing EVs that are as affordable as today’s gasoline-powered vehicles [15]. Therefore, the need to store energy is increasing, and energy storage devices are one way to do this. Different types of storage devices have been used based on the storage requirements, where long-term usage needs batteries, while short time delivery requires supercapacitors. For the last couple of decades, with pioneering work on lithium-ion batteries (LIB) [16–19], there has been a dramatic increase in portable electronic devices. The emergence of electric and hybrid vehicles demands the high-energy storage systems due to the limitations of state-of-the-art LIB. Even the development of high-energy storage systems is essential to save extra power produced from the renewable energy storage and deliver to where it is demanding. Thus, there is a big interest in developing a new type of batteries beyond Li-ion battery [7–9,13,14,20–22].Among the electrochemical batteries, the metal-air batteries have been given considerable re-attention due to their outstanding higher energy density. In contrast to the LIB, the cathode breathing oxygen from the atmosphere serves as a fuel during the cycling process in the metal-air battery system. Further, it has an infinite source of the reactant on both anode (metallic sheet) and cathode (atmospheric oxygen), which results in high theoretical energy [23–25]. Table 1 shows the theoretical energy density of various metal-air batteries [26]. Among the various metal-air batteries, metals such as Ca, Al, Fe, and Zn are suitable for the aqueous system with open-air friendly, whereas Li and Mg are mainly suitable for the non-aqueous system, where their stability is limited at atmospheric condition [24]. Within this, Li-oxygen and Zn-air batteries (ZAB) have been identified as next-generation energy storage devices. Although Li-oxygen batteries have a much higher energy density, the ZAB system could reach commercialization sooner [27–30]. ZAB has various benefits such as low cost, abundance, low equilibrium potential, environmental benignity, a flat discharge voltage, and a long shelf life. The most important merit of ZAB is that the assembly and working of batteries can be performed under ambient conditions (Fig. 1 ) [26,31,32]. During the last decades, the primary ZAB have been commercialized for various application such as hearing aids in medical applications, telecommunication, and electronic devices with long operation time in remote places [32].For rechargeable ZAB, a number of ways have been proposed, such as hydraulic, mechanical, and electrical recharging. Each method of recharging has its own merits and demerits. Working of hydraulic recharging in ZAB is like fuel cell; there will be a continuous supply of reactants to the electrodes. Secondly, the mechanical recharging is similar to the hydraulic method, whereas it needs a continuous replacement of electrolyte and metallic anode in the system. However, these two methods require special requirement and special setup for regeneration. On the other hand, electrical recharging is like other rechargeable batteries, where the ZAB recharge by running a current through it without changing the system [23]. A rechargeable zinc-air battery is composed of three compartments such as zinc metal as an anode, an air-breathing electrode as cathode, and an alkaline solution as an electrolyte. The cathode compartments were named as gas-diffusion electrode (GDE), which further divided into a gas diffusion layer (GDL) and a catalytic active layer (CL). The GDL is a porous structure made by weaving carbon fibers into a carbon cloth or by pressing carbon fibers together into a carbon paper. The active catalyst plays critical roles in a typical ZAB application, and it is able to catalyze both oxygen reduction reaction (ORR) and oxygen evolution reaction (OER) of the oxygen during the cycling. The GDE was prepared by a uniform coating of active catalyst into the GDL by brush coating or spray coating technique. For a rechargeable ZAB system, during the cycling process, the Zn/Zn2+ redox reaction occurs at the anode side and the ORR/OER occurs at the cathode side, simultaneously. In a discharge conditions, the oxygen from the atmosphere diffuses into the GDE due to the oxygen pressure difference between the outside and inside cell, while the catalyst facilitates the reduction of oxygen to hydroxyl ions (OH−) in alkaline electrolyte with the help of electrons generated from the oxidation of zinc metal as the anode reaction. The OH− ions then moved into anode compartment through the electrolyte, and then it react with Zn2+ to form a zincate (Zn(OH)4 2−) ion (Fig. 1). During the continuous cycling, the soluble zincate ions reaches its saturation limit, and then it is converted into solid ZnO, where the ZnO precipitates on the surface of the anode as thistle-like structure [26,33]. In view of thermodynamics, charging is a reversible chemical reaction of discharging, but this is not easy due to the precipitation of solid ZnO on the anode. On the other hand, OER is also a reversal of ORR reaction, where necessarily more O–O bonds are built by breaking O–H bonds [34,35]. Based on these reactions, GDE is a vital component of ZAB, and it is called a three-phase reaction spot, where the solid electrode interfaces with liquid electrolyte, while the reactant is in the gas form. In recent years, ZAB is advancing, whereas a number of well-defined challenges with the rechargeable ZAB exist in the laboratory level, which need to be addressed to provide superior performance in practical applications. Moreover, the precipitation of insoluble white ZnO with thistle-like structure on the surface of the Zn anode, and shape change and dendrite growth of the Zn anode during repeated charge–discharge processes could be detrimental to the stability of ZAB. To avoid structural deformation of metallic zinc, researchers developed high surface area anode with different morphology such as zinc particles, spheres, flakes, ribbons, fibers, and foams for enhancing the reaction mechanism [26,36].There are several potential problems associated with the cathode compartment such as (i) low efficiency due to sluggish oxygen reduction/-evolution reactions kinetics; (ii) atmospheric CO2 possibly react with hydroxyl ions to form carbonates; (iii) formation of insoluble ZnO discharge product, which deposit on the surface of GDE; (iv) inadequate understanding of catalysts effect, and (v) mechanical breakdown generated on the catalyst from the surface of the GDL due to the oxygen evolution during charging (Fig. 1) [26,37]. Further, the use of electrocatalysts in the GDE needs to be investigated for their ability to lower the over-potential for charge and discharge reactions and enhance the cycle life. To date, several kinds of precious and non-precious catalysts have been explored for ZAB, including metal oxides, perovskites, chalcogenide, allotropes of carbon-based materials, noble metals, and so on (Fig. 2 b) [34,38–45 46–80]. So far noble metals (Pt, Ru, Ir) and their alloys (Pt-Ru/C) are used in commercial rechargeable ZAB, whereas they are too expensive to be viable for large-scale commercial applications [81]. Despite these challenges, recent experimental and computational work has provided much-desired information for understanding new catalyst design and synthesis. These cathode materials essentially improve the cycling stability of ZAB to some extent; in addition, the cost of the material should not be compromised. Current challenges in electrocatalyst research not restricted to materials processing by programmable design and assembly, by understanding and predicting defects across time and length scales as well as functionalizing defects for unprecedented properties, and by the discovery of multilateral systems of extreme environments. Thus, it is anticipated that these challenges can only be overcome by enhancing basic understanding of electrocatalyst, and that will ultimately enable advancement in ZAB system (Fig. 2 a) [82].With the Nobel Prize-winning work “for groundbreaking experiments regarding the two-dimensional material graphene” in 2010, the development of innovative graphene-based materials is a key element to design sustainable and resource efficient industries of modern society [83]. The fundamental understanding of graphene materials with regard to significant structure–property correlations on different spatial and time scales, the future of direct manufacturing and production processes and efficient modelling and simulation methods are derisive prerequisites innovation in graphene-based engineering. From the socio-ecological, technical and economic point of view, graphene-based materials today have the extremely important function of an “enabling material”. The high-tech system in energy and environmental technology, automotive and aerospace engineering, medical sciences, information and communication technology, electrical engineering and process engineering are impossible without key enabling materials and components based on graphene [84–86]. Graphene, an allotrope of carbon having a monolayer of sp2 hybridized carbon atoms tightly packed into a two dimensional (2D) honeycomb lattice, and is a basic building block for graphitic materials of all other dimensionalities. Graphene has attracted great interest in energy application due to its band overlap between the valences bands to the conduction band with almost zero bandgaps [83]. Graphene is better described as a mother of all graphitic forms and it can be wrapped up into 0D fullerenes, rolled into 1D nanotubes or stacked into 3D graphite. Graphene is one of the versatile materials for various applications due to its peculiar properties such as specific surface area (2630 m2g−1), intrinsic mobility (200,000 cm2 v−1s−1), Young’s modulus (1.0 TPa), thermal conductivity (5000 Wm−1K−1), optical transmittance (97.7%), and good electrical conductivity (103 Sm−1). In order to produce high quality and quantity of graphene without losing the properties, various techniques have been established. They were broadly classified into two kinds: top-down and bottom-up approaches. The method such as chemical exfoliation, mechanical exfoliation, and chemical synthesis fit into the top-down approach, whereas the method such as pyrolysis, epitaxial growth, and thermal chemical vapor deposition (CVD) method fit into the bottom-up approach (Fig. 3 ). For the preparation of graphene based-air catalysts, the usage of pure graphene is almost restricted due to the fact that graphene is almost not soluble, while it cannot be dispersed in water or any organic solvent [87,88].Different from graphene, graphene oxide (GO) almost contains a variety of oxygen functional groups, like hydroxyl and epoxy group on its basal plane, and carboxyl at its edge. The GO with its outstanding water solubility, control over the functionalization and ease in preparation make them the most popular precursor of graphene based-composites. From the literature, GO is synthesized frequently via chemical oxidation of natural graphite using Hummers and Offeman method in which NaNO3 and KMnO4 dissolved in concentrated H2SO4 was used to oxidize graphite into graphite oxide [87,89,90]. In Hummers’ method, first reduces the interlayer van der Waals forces of the graphitic layer to increase the interlayer spacing. Then it exfoliates graphene with a single to few layers by rapid heating or sonication. To solve the problem associated with Hummers’ method, various modified Hummers’ method has been developed for better usage. Various research work have been reported to improve the electrochemical performance of the catalyst such as non-metals, noble metals, non-noble metals, metal oxides, perovskites, nitrides, sulfides, and carbides, where the synergistic coupling between catalysts with graphene-based materials is a promising approach to create more active sites, and that can improve the electrical conductivity, chemical stability due to an interaction between graphene structures with the electrocatalyst. The graphene structure possesses the strongest in-plane C–C bond, while π bond in the out-of-plane contributes to a delocalized network of electrons, which is responsible for electron conduction of graphene, and that affords weak interaction among graphene layers or between graphene and catalyst. Further, the electrical and chemical properties of graphene-based materials are easily tuned by substituting with heteroatoms, such as N, P, B, and S, which tailor their electron-donor properties of graphene [91]. The effect of the dopants on the electrocatalytic properties of the graphene-based materials is mainly associated with three features of the dopant element: the number of electrons in the external shell, the electronegativity, and the size [92–95]. These materials generally facilitate the charge transfer, which could be mostly due to the difference in electronegativity between the graphene and the dopant. If the electronegativity of the dopant is larger, then the charge difference on the adjacent carbon atoms can be higher, and hence electrocatalytic activity can be greater. Recently, a large number of the heteroatom-doped (N, S, B, P, and halogens) graphene-based materials are focused for metal-air battery application due to their high activity and long-term stability. To further enhance the activity of heteroatom-doped graphene, research is also conducted on various catalysts such as integration of graphene with non-metals, noble metals, non-noble metals, metal oxides, nitrides, sulfides, carbides, and other carbon composites, where an overall enhancement in the bi-functional activity in metal-air battery can be achieved due to the synergistic effect exerted by the dopants and catalyst [84,85]. In this respect, the potential graphene-based air catalysts such as graphene with heteroatoms, non-metals, noble metals, non-noble metals, metal oxides, perovskites, nitrides, sulfides, carbides, and other carbon composites have been reviewed in the present paper in-light-of-their high oxygen reduction reaction/ oxygen evolution reaction activity and zinc-air battery performance for the development of zinc-air batteries. Moreover, this review further extend the recent progress on the zinc-air batteries including the strategies used to improve the high cycling-performance (stable even up-to 394 cycles), capacity (even up-to 873 mA h g−1), power density (even up-to 350 mW cm−2), and energy density (even up-to 904 W h kg−1). Table 2 depicts the ORR/ OER activity and zinc-air battery performance of various kinds of reported graphene with non-metals based air catalysts [96–106]. Creating defects on the graphene can enhance the ORR/OER activity and ZAB performance. Jia et. al. [101] have observed that DG exhibits high ORR/OER activity with high ZAB performance. It was prepared by the following steps: At first, NG was obtained by annealing the graphene with melamine for 2 h at 700 °C under N2 atmosphere; Finally, DG was prepared by annealing the NG for 2 h at 1150 °C under N2 atmosphere. NG contains pyridinic, pyrrolic, and graphitic N whereas DG contains absence of N, and DG is composed of defect graphene with I D/I G ratio of 1.13, and it contains holes, while it possesses various structural defects (Fig. 4 c) such as pentagons, heptagons, and octagons at the edge of holes, and it exhibits high wettability (contact angle: 44.3°), and that can possibly enhance its ZAB performance with ORR/OER activity. They observed that DG exhibits higher ORR/OER activity than NG. It exhibits high OER (EJ=10: 1.6 V) and ORR activity (E1/2: 0.76 V; n: 3.87) with low ΔE of 0.84 V. ZAB with DG affords 100 mA mg−1 at 1 V and high power density of 154 mW mg−1 at 195 mA mg−1. ZAB with DG (Fig. 4 a) exhibits initial polarization voltage of ~0.76 V with negligible increased polarization voltage of ~0.03 V (Fig. 4 b) after 95 cycles at 10 mA mg−1 with Zn foil as anode and 6 M KOH with 0.2 M Zn(O2CCH3)2 as electrolyte, which indicates its much high performance.Doping N with graphene can enhance the ORR activity and ZAB performance. Tian et. al. [97] have observed that NG exhibits high ORR activity with high ZAB performance. It was prepared by the following steps: At first, a powder was obtained by dissolving dicyandiamide and monohydrate glucose in water followed by drying; Finally, NG was prepared by heating the powder at 580 °C for 4 h followed by 850 °C for 6 h under Ar atmosphere. It is composed of N-doped graphene, and it possesses wrinkled sheet structure, and it exhibits some disordered areas in the graphene layer, and it contains O–C, OC, and O–CO bonds, and it contains pyridinic N at 398.2 eV, pyrrolic N at 400.2 eV, and graphitic N at 400.9 eV, where the N-doped graphene can provide spontaneous adsorption and fast solid-state diffusion of oxygen on ultra-large graphene surface, and that can possibly enhance its ZAB performance with ORR activity. It exhibits high ORR activity (E1/2: −0.18 V (vs Ag/AgCl)). ZAB with NG affords high capacity of 793 mA h g−1 at 100 mA cm−2 and power density of 218 mW cm−2.N-doped graphene obtained through carbonization of natural rice can create edge effects and topological defects, and that can enhance the ORR/OER activity. Tang et. al. [100] have observed that NG exhibits high ORR/OER activity. It was prepared by direct carbonization (at 950 °C for 1.5 h under Ar atmosphere) of sticky rice as carbon precursor, Mg(OH)2 as a template, and melamine as a nitrogen source. It is composed of N-doped graphene, and it contains pyridinic N, pyrrolic N, and quaternary N, and it exhibits I D/I G ratio of 1.24, which suggests the existence of holes and edge defects, and it exhibits high surface area (1100 m2 g−1), and it possesses mesoporous structure, and it contains sp2 hybridized N–C bonds, and it contains nanosized holes all over the plane, and it contains C and N, which are uniformly distributed, and that can enhance its ORR/OER activity. It exhibits high OER (EJ=10: 1.67 V) and ORR activity (E1/2: 0.77 V; n: ~3.8) with low ΔE of 0.9 V. ZAB with NG exhibits OCP of 1.42 V and power density of 3 mW cm−2.Preparing N-doped graphene with electron-withdrawing pyridinic N and electron-donating quaternary N can serve as active sites for ORR and OER, respectively, and that can enhance the ORR/OER activity and ZAB performance. Yang et. al. [102] have observed that NG exhibits high ORR/OER activity with high ZAB performance. It was prepared by the following steps: At first, fine powder mixture was obtained by grinding the melamine and L-cysteine at a mass ratio of 4:1; Finally, NG was prepared by pyrolysis of the mixture at 600 °C for 2 h followed by carbonization at 1000 °C for 2 h under Ar atmosphere. It is composed of N-doped graphene, and it possesses 3D structured nano-ribbon networks with entangled and crumpled wrinkle-like structures, and it exhibits high surface area (~530 m2 g−1), and it contains mesopores and, it exhibits a pore volume of ~2.9 cm3 g−1, and it contains pyridinic N (1.45 atomic %), pyrrolic N (0.95 atomic %) and quaternary N (2.8 atomic %), where electron-withdrawing pyridinic N moieties (p-type domain, Fig. 4 d) can act as active sites for OER, while electron-donating quaternary N (n-type domain, Fig. 4 d) can serve as active sites for ORR, and it exhibits the I D/I G ratio of 3.34, which suggests its high disordered carbon structure, and it possesses small mean average crystallite size of the sp2 domains (1.3 nm), while it exhibits low sp2/sp3 ratio (0.36), which indicates its high edge sites, which can enhance electron transfer rate and electrocatalytic activity, and that can possibly enhance its ORR/OER activity and ZAB performance. It exhibits high OER (EJ=10: 1.66 V) and ORR activity (E1/2: 0.84 V; n: ~3.95) with low ΔE of 0.82 V. ZAB with NG exhibits OCP of 1.46 V, 20 mA cm−2 at 1.09 V, capacity of 873 mA h g−1 and power density of 65 mW cm−2. ZAB with NG affords initial polarization voltage of ~0.8 V with an increased polarization voltage of ~0.2 V after > 150 cycles at 2 mA cm−2 with Zn foil as an anode and 6 M KOH with 0.2 M ZnCl2 as the electrolyte, which indicates its high performance.Preparing N-doped exfoliated graphene can enhance the ZAB performance. Lee et. al. [106] have observed that N-ex-G exhibits high ZAB performance. It was obtained by one-step thermal reduction and NH3 treatment at 1100 °C under an extreme heating rate (>150 °C sec−1). It is composed of N-doped exfoliated graphene, and it possesses wrinkled surfaces and wavy edges, and it exhibits hexagonal symmetry, which suggests the existence of symmetrical three-fold sp2 bonding of carbon atoms, and it exhibits I D/I G ratio of 1.14, and it contains pyridinic N, pyrrolic N, and quaternary N, and that can possibly enhance its ZAB performance. ZAB with N-ex-G affords 47.6 mA cm−2 at 0.8 V and power density of 42.4 mW cm−2.S-doped graphene foam prepared from food (idly) can enhance ORR/OER activity and ZAB performance. Patra et. al. [103] have observed that S-doped graphene foam exhibits high ORR/OER activity with high ZAB performance. It was prepared by the following steps: At first, S-doped graphene-idli was obtained using S-doped graphene and rice flour through microwave treatment; Finally, S-doped graphene foam was prepared by calcination of the S-doped graphene-idli for 1 h at 300 °C under air atmosphere, where the furnace had been pre-heated at 80 °C prior to calcination. It is composed of S-doped graphene foam, and it contains–C–S–C–at 163.7 and 164.3 eV, and–C–SOx–C–at 169.8 eV (Fig. 4 e), and it possesses porous, rough, crumpled, and sponge-like structure, and it exhibits high surface area (499 m2 g−1) with a pore volume of 0.522 cm3 g−1, and that can possibly enhance its ORR/OER activity and ZAB performance. It exhibits high ORR and OER activity with low ΔE of 0.65 V. ZAB with S-doped graphene foam exhibits high OCP of ~1.40 V, ~ 250 mA cm−2 at 0.6 V, and power density of ~300 mW cm−2. ZAB with S-doped graphene foam affords negligible increased polarization voltage after cycled for 125 h at 1 mA cm−2 with carbon tape coated Zn powder as an anode and O2-saturated 6 M KOH with 0.2 M ZnCl2 as the electrolyte, which suggests its high performance.Preparing S, N co-doped graphene-like electrocatalyst can enhance the ORR activity and ZAB performance. Zhang et. al. [105] have observed that SN-G exhibits high ORR activity with high ZAB performance. It was prepared from keratin as precursor along with potassium hydroxide activation, followed by high-temperature graphitization and NH3 treatment. It is composed of S, N co-doped graphene-like nanobubble and nanosheet hybrids, and it exhibits much high surface area (1799 m2 g−1) with a pore volume of 1.01 cm3 g−1. Further it exhibits I D/I G ratio of 1.04, and characteristic peak of graphene (Raman spectra: ~2700 cm−1), and it contains predominant C-sp2 (73%) with C-sp3, C–N, C–O and CO. The presence of pyridinic-N, pyrrolic-N, graphitic-N, thiophene-S, and oxidized-S, and that can possibly enhance its ORR activity and ZAB performance. It exhibits high ORR activity (E1/2: 0.842 V; n: 3.98). ZAB with SN-G exhibits high capacity of 767 mA h g−1 and power density of 201 mW cm−2.Preparing defect enriched S, N co-doped graphene-like carbon can enhance the ORR activity and ZAB performance. Zhang et. al. [98] have observed that Def-SN-GLC exhibits high ORR activity with high ZAB performance. It was obtained from cystine as a precursor through KOH activation and NH3 injection at high temperature. It is composed of defect enriched S, N co-doped graphene-like carbon, and it contains carbon nanosheets with sub-transparent and wrinkled lamellar properties, which possesses similar morphology of graphene. The high surface area (1309 m2 g−1) with pore volume of 0.86 cm3 g−1, and it exhibits I D /I G ratio of 1.09, and it exhibits characteristic peak of graphene (Raman spectra: ~2700 cm−1). Further, it contains pyridinic-N, pyrrolic-N, graphitic-N, thiophene-S, and oxidized-S, and that can possibly enhance its ORR activity and ZAB performance. It exhibits high ORR activity (E1/2: 0.849 V; n: 3.96). ZAB with Def-SN-GLC affords high OCP of 1.5 V, 20 mA cm−2 at 1.24 V and power density of 252 mW cm−2.B, N co-doped graphene with graphitic N and BC3 can facilitate the ORR activity. Qin et. al. [96] have observed that B, N-pG-O catalyst exhibits high activity for ORR. It was prepared by the following steps: At first, N-pG-O was obtained by hydrothermal treatment of N-pG with 8 M HNO3 for 12 h at 60 °C; Finally, B, N-pG-O was obtained by annealing N-pG-O with boric acid for 15 min at 1000 °C. It is composed of B, N co-doped porous graphene, and it possesses honeycomb-like porous structure, and it exhibits high surface area (1303 m2 g−1), and it contains graphitic N and BC3, and it shows I D/I G ratios of 1.05, and that can possibly enhance its ORR activity. They observed through density functional theory (DFT) that the synergistic effect between BC3 and graphitic N can enhance the reduction of oxygen. It exhibits high ORR activity (E1/2: 0.86 V; n: 3.84 to 3.95). ZAB with B, N-pG-O affords open circuit potential (OCP) of ~1.39 V and power density of 30.43 mW cm−2. Moreover, ZAB with B, N-pG-O air–cathode exhibits increased polarization voltage of ~0.55 V after 30 cycles at 1 mA cm−2 with Zn plate as an anode and 6 M KOH as the electrolyte, which indicates its reasonable performance.Integrating the holey graphene framework with CNTs can enhance the ORR activity and ZAB performance. Cheng et. al. [104] have observed that N-CNTs-HGF (HGF: Holey graphene framework) exhibits high ORR activity with high ZAB performance. It was prepared by the following steps: At first, GO was obtained by oxidation of natural graphite powder through a modified Hummers' method; Then, Fe2O3/HGF was obtained by hydrothermal treatment of GO with FeCl3 and KOH/H2O/ethylene glycol at 180 °C for 6 h followed by heat treatment at 850 °C for 2 h under Ar atmosphere; Finally, N-CNTs-HGF was obtained by annealing of Fe2O3/HGF with melamine at 800 °C for 2 h under Ar/H2 atmosphere followed by removal of Fe with concentrated HCl. It is composed of crooked bamboo-like N-doped CNTs, which are anchored on the holey graphene framework. The structure contains pyridinic-N, pyrrolic-N, and graphitic-N, while it shows C–N–C at 285.2 eV, and it contains C–C at 284.8 eV, C–O at 286.6 eV, O–CO at 289 eV, and CO at 288.3 eV, which are attributed to the graphene and CNT based materials, and it contains mesopores, and that can possibly enhance its ORR activity and ZAB performance. It exhibits high ORR activity (E1/2: 0.85 V; n: 3.9). All-solid ZAB with N-CNTs-HGF exhibits high OCP of 1.43 V, 10 mA cm−2 at ~0.55 V, the capacity of 625 mA h g−1, an energy density of 614 W h kg−1, and power density of ~8.5 mW cm−2. All-solid ZAB with N-CNTs-HGF is bendable, which indicates its flexibility, and it affords increased polarization of ~0.17 V after > 60 cycles at 5 mA cm−2 with Zn foil as an anode and solid electrolyte gel as the electrolyte, which indicates its high performance.Preparing N, P co-doped carbon framework through pyrolysis of a supermolecular aggregate of self-assembled phytic acid, melamine, and graphene oxide can enhance the ORR activity and ZAB performance. Zhang et. al. [99] have observed that MPSA-GO (MPSA: Melamine–phytic acid supermolecular aggregate) exhibits high ORR activity with high ZAB performance. It was prepared by self-assembling melamine and phytic acid into MPSA in the presence of graphene oxide, followed by pyrolysis at 1000 °C for 1 h under argon atmosphere. It possesses 3D porous carbon networks, which are co-doped with nitrogen and phosphorus, and it contains graphitic sp2 carbon at 284.6 eV, and it exhibits C–N and/or CN at 285.6 eV, P–C bond at ~131.6 eV and P–O bond at ~133.2 eV, and it contains pyridinic N at 398.3 eV, pyrrolic N at 400.1 eV, graphitic N at 401.2 eV, and oxidized N at 403.7 eV, and that can possibly enhance its ORR activity and ZAB performance. It exhibits high ORR activity. ZAB with MPSA-GO exhibits high power density of 310 W g−1.Thus, various strategies including creating defects on the graphene [101], doping N with graphene [97], N-doped graphene obtained through carbonization of natural rice [100], preparing N-doped graphene with electron-withdrawing pyridinic N and electron-donating quaternary N [102], preparing N-doped exfoliated graphene [106], S-doped graphene foam prepared from food (idly) [103], preparing S, N co-doped graphene-like electrocatalyst [105], preparing defect enriched S, N co-doped graphene-like carbon [98], B, N co-doped graphene with graphitic N and BC3 [96], integrating holey graphene framework with CNTs [104], and preparing N, P co-doped carbon framework [99] enhanced the ORR/OER activity and ZAB performance. Table 3 depicts the ORR/OER activity and zinc-air battery performance of various kinds of reported graphene with non-noble metals based air catalysts [107–116]. Integrating Fe/Fe3C@C nanoparticles with graphene framework can facilitate the electron/charge transport and that can enhance the ORR/OER activity and ZAB performance. Wang et. al. [107] have observed that Fe/Fe3C@C-NG/NCNTs exhibits high ORR/OER activity with high ZAB performance. It was prepared by the following steps: At first, GO was obtained by modified Hummer’s method; Then, a powder was obtained by sonicating the SBA-15 with GO suspension for 2 h followed by freeze-drying for 48 h; Later, black powder was obtained by heating the powder with FeCl3·6H2O and dicyandiamide at 550 °C for 4 h followed by heating at 800 °C for 1 h under N2 atmosphere; Finally, Fe/Fe3C@C-NG/NCNTs was obtained by washing the black powder with 10% HF for 24 h followed by drying. It is composed of α–Fe/Fe3C@C nanoparticles, which are enwrapped in 3D N-doped graphene and bamboo-like CNTs. It exhibits I D/I G ratio of 0.7, which indicates the existence of such high degree of graphitization, which can be ascribed to the co-existence of graphene and bamboo-like CNTs. It possesses mesopores and macropores, while it exhibits high surface area (117.6 m2 g−1), which can enhance the mass transport and afford abundant active sites, and it contains pyrrolic-N, pyridinic-N/Fe-N, graphitic-N, oxidized-N, Fe0, Fe2+, and Fe3+, where Fe–Nx can be formed when the pyridinic-N coordinate with Fe, which could facilitate its ORR activity, and it exhibits high turnover frequency of 3.06 e site−1 s−1 on the basis of Fe–Nx sites, and it contains CO, C–OH, and C–O–C. The existence of these hydrophilic oxygen-containing groups can enhance the three-phase contact among the electrode, electrolyte, and reactants, and that can possibly enhance its ORR/OER activity and ZAB performance. It exhibits high OER (EJ=10: 1.68 V) and ORR activity (E1/2: 0.84 V; n: 3.90 to 4.00) with low ΔE of 0.84 V. ZAB with Fe/Fe3C@C-NG/NCNTs exhibits high capacity of 682.6 mA h g−1, energy density of 764.5 W h kg−1, power density of 101.2 mW cm−2, OCP of 1.37, and 10 mA cm−2 at 1.12 V for 40 h. ZAB with Fe/Fe3C@C-NG/NCNTs affords initial polarization voltage of 0.89 V with an increased polarization voltage of 0.13 V after 297 cycles at 10 mA cm−2 with Zn plate as anode and 6 M KOH with 0.2 M Zn(O2CCH3)2 as electrolyte, which indicates its high performance.Preparing Fe–N active sites, and integrating CNTs and graphene with MOF can enhance its activity and electronic conductivity, and that can enhance the ORR/OER activity and ZAB performance. Yang et. al. [110] have observed that Fe-MOF@CNTs-G exhibits high ORR/OER activity with high ZAB performance. It was prepared by the following steps: At first, MIL-53(Fe) was obtained by hydrothermal treatment at 150 °C for 6 h; Then, a mixture was obtained by ultrasonic treatment of MIL-53, (NH4)2S2O8, and melamine, while the mixture was frozen overnight; Finally, Fe-MOF@CNTs-G was obtained by heating the above product at 240 °C for 2 h and 900 °C for 1 h under N2 atmosphere. It contains graphite carbon and Fe3C phase, and it is composed of graphene-like structure, which is formed around the MOFs, while short CNTs also emerges on the surface of MOFs, where homogenous Fe3C nanoparticles are enwrapped at the end of the CNTs. The structure contains C, N, O, and S, which are uniformly distributed, and it possesses a high surface area (90 m2 g−1), and it contains CC, C–C, C–S, C–N–C/CO, pyridinic N (24.8%), pyrrolic N (24.8%), Fe–N (24.8%), and graphitic N (25.6%), and it exhibits the I D/I G ratio of 0.93, which indicate the existence of defects, and it exhibits high electrochemically active surface area (8.03 mF cm−2). The high surface area within this material can afford abundant Fe–N active sites and facilitate the transfer of the reactants, while CNTs and graphene can enhance the electronic conductivity of the MOF substrate, and that can possibly enhance its ORR/OER activity and ZAB performance. It exhibits high OER (EJ=10: 1.646 V) and ORR activity (E1/2: 0.873 V; n: 3.92 to 3.99) with low ΔE of 0.773 V. ZAB with Fe-MOF@CNTs-G exhibits high capacity of 637.4 mA h g−1, energy density of 734.1 W h kg−1, power density of 95.3 mW cm−2, and OCP of 1.414 V. ZAB with Fe-MOF@CNTs-G affords initial polarization voltage of 1.01 V with negligible increased polarization voltage of 0.04 V after 100 cycles at 10 mA cm−2 with Zn plate as anode and 6 M KOH with 0.2 M Zn(O2CCH3)2 as electrolyte, which indicates its much high performance.Integrating NiFe nanoparticles with graphene can alter the electronic modulation and that can enhance the ORR/OER activity and ZAB performance. Zhu et. al. [115] have observed that NiFe@NCX exhibits high ORR/OER activity with high ZAB performance. It was prepared by the following steps (Fig. 5 a): At first, NiFe-MIL was obtained by hydrothermal treatment at 100 °C for 15 h; Then, a mixture was obtained by stirring the suspension containing NiFe-MIL and melamine in ethanol followed by evaporation; Later, the mixture was pyrolyzed at 600 °C for 1 h and 800 °C for 1 h under N2 atmosphere; Finally, NiFe@NCX was prepared by pickling the pyrolysis product with 1 M HCl for 8 h at 80 °C to remove the unstable metal species. It possesses 3D flake-like structure, and it is composed of ultra-fine NiFe nanoparticles (cubic NiFe2 phase), which are encapsulated by N-doped thin graphene nanosheets; It contains Fe, Ni, C and N, which are uniformly distributed, and it exhibits high surface area (350 m2 g−1), and it is mesoporous, and it exhibits I D/I G ratio of < 1, which indicates the existence of highly ordered graphitic structure, which can enhance its electrical conductivity, and it contains pyridinic N at 398.9 eV and quaternary N at 400.9 eV, and that can possibly enhance its ORR/OER activity and ZAB performance. It exhibits high OER (EJ=10: 1.555 V) and ORR activity (E1/2: 0.86 V; n: 4.1) with low ΔE of 0.695 V. ZAB with NiFe@NCX exhibits high capacity of 583.7 mA h g−1, and energy density of 732.3 W h kg−1 (Fig. 5). ZAB with NiFe@NCX affords initial polarization voltage of 0.39 V with increased polarization voltage of 0.29 V after 205 cycles (Fig. 5 e) at 10 mA cm−2 with Zn plate as anode and 6 M KOH as electrolyte, which indicates its high performance.Integrating Vulcan carbon with CoFe-N-rGO can spatially separate the CoFe-N-rGO layers and can improve its conductivity, and that can enhance the ORR activity and ZAB performance. Kashyap et. al. [116] have observed that CoFe-N-rGO-Vulcan exhibits high ORR activity with high ZAB performance. It was prepared by the following steps: At first, CoFe-N-rGO was obtained by hydrothermal treatment at 130 °C for 12 h; Then, CoFe-N-rGO was annealed at 150 °C for 12 h; Finally, CoFe-N-rGO-Vulcan was obtained by mixing the 2:8 ratio of CoFe-N-rGO and Vulcan carbon. It is composed of cobalt ferrite nanoparticles, which are homogeneously distributed on the N-rGO, while CoFe-N-rGO layers are spatially separated by Vulcan carbon (Fig. 6 a), and it exhibits high surface area (187 m2 g−1), and it contains micro/mesopores, and it contains pyridinic N and pyrrolic N, and it contains Co2+, Co3+, Fe2+, and Fe3+, and that can possibly enhance its ORR activity and ZAB performance. It exhibits high ORR activity (E1/2: −0.133 V (vs Hg/HgO); n: 3.7). ZAB with CoFe-N-rGO-Vulcan exhibits high capacity of ~630 mA h g−1, the power density of 155 mW cm−2, and 30 mA cm−2 at 1.0 V.Preparing Co nanoclusters distributed on N-doped carbon can enhance the ORR activity and ZAB performance. Gao et. al. [108] have observed that Co-N-rGO exhibits high ORR activity with high ZAB performance. It was prepared by the following steps: At first, Co-MOF/GO was obtained by hydrothermal treatment at 120 °C for 36 h; Then, the above product was pyrolyzed at 850 °C for 2 h under N2 atmosphere; Finally, Co-N-rGO was obtained by pickling the pyrolysis product with 3.0 M HCl for 12 h. It is composed of Co nanoclusters (Diameter: ≈ 1–2 nm), which are uniformly distributed on N-doped carbon, and it contains Pyridinic-N, Pyrrole-N, Graphitic-N, an N-oxide, and it contains graphitic carbon, C–O, and CO bonds, and it exhibits I D/I G ratio of 0.95, which suggests the existence of a high degree of graphitized carbon, and it exhibits high surface area (179 m2 g−1), and it contains mesopores, and that can possibly enhance its ORR activity and ZAB performance. It exhibits high ORR activity (E1/2: 0.831 V; n: 3.75). ZAB with Co-N-rGO exhibits high capacity of 795 mA h g−1, a power density of 175 mW cm−2, and 300 mA cm−2 at 0.65 V.Preparing Co/N/O tri-doped graphene with intrinsic structural defects and atomically dispersed Co–N x –C active sites can enhance the ORR/OER activity and ZAB performance. Tang et. al. [111] have observed that Co/N/O-G exhibits high ORR/OER activity with high ZAB performance. It was prepared by the carbonization of a powdery mixture at 950 °C for 1.5 h under Ar atmosphere, where the mixture was composed of gelatinized amylopectin, in situ generated Mg(OH)2 nanoflakes, melamine, and cobalt nitrate. It is composed of Co/N/O tri-doped graphene (Fig. 6 b), and it exhibits much high surface area (541.5 m2 g−1), and it contains micro/mesopores. It exhibits I D/I G ratio of 1.32, which suggests the existence of much high defects, and it contains pyridinic N, Co–Nx, pyrrolic N, quaternary N, oxidized N and chemisorbed N, where the pyridinic N bound to Co is up-shifted with ≈ 1 eV from pristine pyridinic N (≈ 398.4 eV), while C–N shoulder noticeably shifts to higher binding energy, which can be ascribed to the strong electron-withdrawing effect of cobalt coordinated with the nitrogen in Co–N x –C moieties. The decreased electron density in adjacent C atoms can enhance the adsorption of ORR/OER intermediates, that can facilitate the electron transfer, and that can possibly enhance its ORR/OER activity and ZAB performance. It exhibits high ORR/OER activity with low ΔE of 0.95 V. ZAB with Co/N/O-G exhibits high capacity of 750 mA h g−1, energy density of 840 W h kg−1, power density of 152 mW cm−2, and OCP of 1.44 V. ZAB with Co/N/O-G affords initial polarization voltage of ~1.00 V with increased polarization voltage of ~0.12 V after 180 cycles at 2 mA cm−2 with Zn foil as anode and 6 M KOH with 0.2 M ZnCl2 as electrolyte. Moreover, solid ZAB is constructed (Fig. 6 c), where Co/N/O-G, Zn foil, and poly(vinyl alcohol) gel are used as cathode, anode, and electrolyte, respectively, and it is flexible, while it delivers stable potential during charge/discharge cycling even at different bent state.Preparing cobalt nanoparticles encapsulated in N-enriched graphene shells with hollow graphene spheres can afford Co-N-C active sites, and that can enhance the ORR activity and ZAB performance. Zeng et. al. [113] have observed that Co@NG-acid exhibits high ORR activity with high ZAB performance. It was prepared by the following steps: At first, cobalt analogue of Prussian blue (PB-Co) was obtained by co-precipitation method; Then, Co@NG was prepared by annealing the PB-Co at 600 °C for 1 h under Ar atmosphere; Finally, Co@NG-acid was obtained by treating the Co@NG with 1 M HCl for overnight. It is composed of metallic cobalt nanoparticles (Fig. 6 d), which are enwrapped in N-enriched graphene shells, while it possesses hollow graphene spheres due to the leaching of Co by acid treatment. It contains ≈ 10 wt% of Co in both Co2+, Co3+, while it contains pyridinic N, pyrrolic N, and quaternary N, where Co-N-C active sites could be formed, and that can possibly enhance its ORR activity and ZAB performance. It exhibits high ORR activity (E1/2: 0.83 V; n: >3.8). ZAB with Co@NG-acid exhibits a high power density of 350 mW cm−2 and 255 mA cm−2 at 1 V.Preparing Co, N-co doped CNT/graphene heterostructure can generate pyridinic N-C and Co-N active sites, and that can enhance the ORR activity and ZAB performance. Yang et. al. [109] have observed that Co, N-CNT-NG exhibits high ORR activity with high ZAB performance. It was prepared by the following steps: At first, dried precursor was obtained by stirring the solution containing 0.2 g of g-C3N4 and 0.17 M CoCl2·6H2O for 24 h followed by evaporation; Finally, Co, N-CNT-NG was prepared by pyrolyzing the above precursor at 550 °C for 2 h and then at 800 °C for 2 h under N2 atmosphere. It is composed of Co, N-co doped CNT/graphene heterostructure, where Co nanoparticles are enwrapped by CNT/graphene heterostructure, and it contains C, N, and Co, which are uniformly distributed. It contains pyridinic N, pyrrolic N/Co-N, quaternary N, oxidised pyridinic N, where pyridinic N-C and Co-N species can act as highly electroactive sites, and that can possibly enhance its ORR activity and ZAB performance. It exhibits high ORR activity (E1/2: 0.85 V; n: 3.96), while ZAB with Co, N-CNT-NG exhibits a power density of 88 mW cm−2.Preparing the graphene matrix with Cu(I)-N active sites can enhance the ORR activity and ZAB performance. Wu et. al. [114] have observed that Cu–N@G exhibits high ORR activity with high ZAB performance. It was prepared by pyrolysis of solid-phase precursors containing copper phthalocyanine and dicyandiamide at 800 °C for 2 h under Ar atmosphere, followed by acid treatment (0.5 M H2SO4) for 12 h at 70 °C to eliminate unstable Cu species. It is composed of Cu atoms, which are embedded in the graphene matrix, it contains Cu, N and C, which are uniformly distributed, and it contains pyridinic N, pyrrolic N, graphitic N, and oxidized N. It possesses much high surface area (333.877 m2 g−1), and it contains mesopores, while their electrochemical and theoretical studies reveal that Cu(I)-N is considered as the active site for catalyzing the ORR, where O atoms prefer to adsorb on Cu atoms in Cu-N2, and that can possibly enhance its ORR activity and ZAB performance. It exhibits high ORR activity (E1/2: 0.80 V; n: 3.96). ZAB with Cu–N@G exhibits high power density of ~210 mW cm−2 and 142 mA cm−2 at 1 V, which indicates its high performance.Integrating Ag NW with graphene aerogel can inhibit restacking of graphene sheets and facilitate electronic conductivity, and that can enhance the ORR activity and ZAB performance. Hu et. al. [112] have observed that Ag NW-GA exhibits high ORR activity with high ZAB performance. It was prepared by mixing the Ag NW and GO suspensions followed by hydrothermal self-assembly for 3 h at 90 °C. It is composed of Ag NW, which are tightly attached to the graphene sheets, while 0D Ag nanocrystals are uniformly distributed on the graphene sheets (Fig. 6 e). Ag NW-graphene aerogel possesses an open sponge structure, where Ag NW inhibits the restacking of graphene sheets, and facilitate electronic conductivity, and act as an Ag source for the ultrasmall Ag nanocrystals deposition, and that can possibly enhance its ORR activity and ZAB performance. It exhibits high ORR activity. ZAB with Ag NW-GA exhibits high capacity of 637.3 mA h g−1, the energy density of 794.5 W h kg−1, the power density of 331 mW cm−2, OCP of 1.48 V, and 206 mA cm−2 at 1.0 V, which indicates its high performance.Thus, several strategies including integrating Fe/Fe3C@C nanoparticles with graphene framework [107], integrating NiFe nanoparticles with graphene [115], integrating Vulcan carbon with CoFe-N-rGO [116], preparing Co nanoclusters on N-doped carbon [108], preparing Co/N/O tri-doped graphene [111], preparing cobalt nanoparticles encapsulated in N-enriched graphene shells with hollow graphene spheres [113], preparing Co, N-co-doped CNT/graphene heterostructure [109], preparing graphene matrix with Cu(I)-N active sites [114], integrating Ag NW with graphene aerogel [112], and preparing Fe–N active sites and integrating CNTs and graphene with MOF [110] improved the ORR/OER activity and ZAB performance. Table 4 depicts the ORR/OER activity and zinc-air battery performance of various kinds of reported graphene with metal oxides based air catalysts [37,117–126]. Preparing MnO2 nanofilm on N-doped hollow graphene can enhance the ORR activity and ZAB performance. Yu et. al. [120] have observed that MnO2-NG exhibits high ORR activity with high ZAB performance. It was prepared by template method and mild oxidation process through the following steps: At first, SiO2@rGO was obtained by drying the silica/graphene oxide suspension for overnight at 80 °C followed by heating at 850 °C for 2 h under N2 atmosphere; Then, hollow graphene spheres was obtained by removal of SiO2 templates by etching with 10 wt% of HF for 12 h at room temperature; Later, NG was prepared by hydrothermal treatment of the above hollow graphene with NH4OH at 180 °C for 8 h followed by annealing at 800 °C for 2 h under N2 atmosphere; Finally, MnO2-NG was obtained by treating NG with 0.012 mol L−1 of KMnO4 at 60 °C for 2 h followed by annealing at 800 °C for 2 h under N2 atmosphere. It is composed of MnO2 nanofilms, which are uniformly anchored on the transparent N-doped hollow graphene spheres, and it exhibits high surface area (302 m2 g−1) with pore volume of 1.8 cm3 g−1. It contains C, N, O and Mn, which are homogenously distributed, while it contains 84.9 at % of C, 2.3 at % of N, 8.7 at % of O and 4.1 at % of Mn; It contains graphitic N, pyridinic N, pyrrolic N and N-oxides, where graphitic N and pyridinic N can enhance much ORR activity, and it contains C–O and Mn4+, and that can possibly enhance its ORR activity and ZAB performance. It exhibits high ORR activity (E1/2: 0.84 V; n: 3.65 to 3.85). ZAB with MnO2-NG exhibits high capacity of 744 mA h g−1 at 10 mA cm−2, a power density of 82 mW cm−2, OCP of 1.48 V, and discharge voltage of 1.14 V at 25 mA cm−2, which indicates its high performance.Ionic liquid (IL) moiety can increase the conductivity and electrocatalytic activity of rGO while integrating Mn3O4 with rGO-IL can enhance the ORR activity and ZAB performance. Lee et. al. [125] have observed that Mn3O4@rGO-IL exhibits high ORR activity with high ZAB performance. It was prepared by facile solution-based growth mechanism. They observed that rGO-IL exhibits higher ORR activity than rGO, while Mn3O4@rGO-IL exhibits highest ORR activity than rGO-IL. It is composed of crystalline Mn3O4 nanoparticles, which are anchored on the rGO-IL nanosheet, where ionic liquid moiety not only increases the conductivity but also increases the electrocatalytic activity compared with pristine rGO, while the synergic effect between Mn3O4 and rGO-IL can facilitate its catalytic activity, and that can possibly enhance its ORR activity and ZAB performance. It exhibits high ORR activity, while ZAB with Mn3O4@rGO-IL exhibits high power density of 120 mW cm−2.Integrating CoMn2O4 with NrGO can enhance the ORR/OER activity and ZAB performance. Prabu et. al. [124] have observed that CoMn2O4/NrGO exhibits high ORR/OER activity with high ZAB performance. It was prepared by hydrothermal treatment at 150 °C for 3 h. It is composed of CoMn2O4 nanoparticles (~20 nm), which are dispersed on N-doped rGO (Fig. 7 a), where CoMn2O4 possesses tetragonal structure, and it contains mesopores, and it contains pyridinic-N, pyrrolic-N and quaternary-N, where the synergistic effect between the spinel cobalt manganese oxide and nitrogen-doped graphene sheets can facilitate its electrocatalytic activity, and that can possibly enhance its ORR/OER activity and ZAB performance. It exhibits high OER (EJ=10: 1.66 V) and ORR activity (E1/2: 0.75 V; n: 4) with low ΔE of 0.91 V. ZAB with CoMn2O4/NrGO affords initial polarization voltage of 0.7 V with increased polarization voltage of 0.16 V after 100 cycles at 20 mA cm−2 with Zn plate as anode and 6 M KOH as electrolyte, which indicates its high performance. Moreover, Prabu et. al. [37] have observed that CoMn2O4/NrGO exhibits high ZAB performance by consuming oxygen from the atmosphere. ZAB with CoMn2O4/NrGO exhibits high capacity of 610 mA h g−1 and OCP of 1.15 V. ZAB with CoMn2O4/NrGO (Fig. 7 b) affords initial polarization voltage of 0.7 V with increased polarization voltage of 0.36 V after 200 cycles at 20 mA cm−2 with Zn plate as anode and 6 M KOH as electrolyte, which indicates its high performance.Integrating MnCoFeO4 with N-rGO can enhance the ORR/OER activity and ZAB performance. Zhan et. al. [126] have observed that MnCoFeO4-N-rGO exhibits high ORR/OER activity with high ZAB performance. It was prepared by hydrothermal treatment at 150 °C for 3 h. It is composed of MnCoFeO4 nanoparticles (~5 nm), which are uniformly dispersed on N-rGO nanosheets, where MnCoFeO4 is comprised of Mn and Co atoms, which are displaced some Fe atoms from the Fe3O4 lattice. It contains pyridinic N at 399.1 eV and graphitic N at 400.5 eV, and Mn3+, Co2+ and Fe3+, and that can possibly enhance its ORR/OER activity and ZAB performance. It exhibits high OER (EJ=10: 1.71 V) and ORR activity (E1/2: 0.78 V; n: 3.8) with low ΔE of 0.93 V. ZAB with MnCoFeO4-N-rGO exhibits OCP of 1.46 V. ZAB with MnCoFeO4-N-rGO affords initial polarization voltage of ~1.25 V with negligible increased polarization voltage after 75 cycles at 10 mA cm−2 with Zn plate as an anode and 6 M KOH as an electrolyte.Functionalizing graphene oxide with 1-hexyl-3-methylimidazolium chloride molecules can enhance the ionic conductivity and mechanical properties of the solid electrolyte, and that can enhance the ZAB performance. Zarrin et. al. [121] have observed that Co3O4 5-HMIM-GO exhibits high ZAB performance. 5-HMIM-GO was prepared by the following steps (Fig. 7 c): At first, GO nanosheets were functionalized with HMIM in an aqueous solution containing KOH; Finally, 5-HMIM-GO was obtained by vacuum filtration method. 5-HMIM-GO solid electrolyte is composed of graphene oxide, which is functionalized with 1-hexyl-3-methylimidazolium chloride molecules through both covalent and non-covalent bonds, where covalent bond induced by esterification reactions, while non-covalent bond induced by electrostatic πcation–πstacking. It exhibits high hydroxide conductivity at 30% RH and room-temperature, and it is free-standing and flexible membrane, and it possesses high mechanical properties (Tensile strength: 35.49 MPa; Young’s Modulus: 1.7 GPa; Elongation at break: 2.13%; Toughness: 0.37 MPa), where the toughness of 5-HMIM-GO is boosted to ~62% when compared to that of bare GO. It contains C, O, and N, which are uniformly distributed, and it exhibits the I D/I G ratio of 1.6, which suggests the existence of high defects, and it contains 4.89 atomic % of N, 37.67% of O, and 57.43% of C. It contains highly intense C–C/C–H peak at 284.99 eV along with–C–N, CO, OC–OH, N–C–N, and OC–N−, and that can possibly enhance its ZAB performance. ZAB with Co3O4 5-HMIM-GO exhibits OCP of 1.25 V. ZAB with Co3O4 5-HMIM-GO affords initial polarization voltage of ~0.9 V with negligible increased polarization voltage after 60 cycles at 200 mA g−1 with zinc pellet as anode and 5-HMIM-GO as a solid electrolyte, which indicates its much high performance.Integrating Co3O4 with N doped graphene can enhance the ORR activity and ZAB performance. Singh et. al. [122] have observed that Co3O4-NG exhibits high ORR activity with high ZAB performance. It was prepared by hydrothermal treatment at 120 °C for 24 h. It is composed of spinel Co3O4 spherical nanoparticles (~60 nm), which are well dispersed on N doped graphene (Fig. 7 d). It contains pyridinic-N, pyrrolic-N, and quaternary-N, and it exhibits the I D/I G ratio of 1.39, which suggests the existence of high defects, and that can possibly enhance its ORR activity and ZAB performance. It exhibits high ORR activity (E1/2: −0.11 V (vs Hg/HgO); n: 3.7). ZAB with Co3O4-NG exhibits high capacity of ~590 mA h g−1, the energy density of ~840 W h kg−1, the power density of ~190 mW cm−2, and OCP of 1.52 V.Integrating Co3O4 nano-rods with reduced graphene oxide can enhance the conductivity and defects, and that can enhance the ZAB performance. Shen et. al. [118] have observed that rGO-Co3O4 exhibits high ZAB performance. It was prepared by hydrothermal treatment at 120 °C for 5 h followed by annealing at 300 °C for 2 h. They observed that it exhibits higher conductivity than Co3O4. It is composed of crystalline Co3O4 nano-rods, which are dispersed on the surfaces of the reduced graphene oxide, and it exhibits I D/I G ratio of 1.043, which suggests the existence of defects, and it exhibits high conductivity, and that can possibly enhance its ZAB performance. ZAB with rGO-Co3O4 exhibits 47.2 mA cm−2 at 0.8 V. ZAB with rGO-Co3O4 affords initial polarization voltage of ~1.5 V with an increased polarization voltage of ~0.25 V after 100 cycles at 50 mA cm−2 with Zn plate as the anode and 6 M KOH as an electrolyte, which indicates its high performance.Preparing predominant metallic Co with small fraction of its oxides anchored on N-doped reduced graphene oxide can enhance the ORR activity and ZAB performance. Liu et. al. [123] have observed that Co-CoO/NrGO exhibits high ORR/OER activity with high ZAB performance. It was prepared by hydrothermal treatment followed by pyrolysis. They observed that Co-CoO/NrGO exhibits higher ORR activity than CoO/NrGO (Fig. 8 a). It is composed of crystalline Co-CoO nanorods, which are enwrapped with N-rGO, where Co-CoO is comprised of predominant metallic Co with minor CoO. It contains Co0 and Co2+, and that can possibly enhance its ORR/OER activity and ZAB performance. It exhibits high OER (EJ=10: 1.47 V) and ORR activity (E1/2: 0.78 V; n: 3.7 to 3.9) with low ΔE of 0.69 V. ZAB with Co-CoO/NrGO exhibits 35 mA cm−2 at 1.08 V. ZAB with Co-CoO/NrGO affords polarization voltage of 1.26 V at 50 mA cm−2 with zinc plate as an anode and 6.0 M KOH as the electrolyte.Integrating amorphous bimetallic oxide with N-doped reduced graphene oxide can enhance the conductivity, electrochemically active surface area, and create oxygen deficiency, and that can enhance the ORR/OER activity and ZAB performance. Wei et. al. [117] have observed that Fe0.5Co0.5Ox-NrGO exhibits high ORR/OER activity with high ZAB performance. It was prepared from Prussian blue analog nanocrystals by low-temperature (at 300 °C) oxidative decomposition of the Fe a Co1− a /polyethylenimine/GO hybrid (Fig. 8 b). It is composed of amorphous bimetallic oxide (Fe0.5Co0.5Ox) nanoparticles, which are dispersed on N-doped reduced graphene oxide, it exhibits high conductivity, and it exhibits high oxygen deficiency (23.9%). It exhibits the I D/I G ratio of 1.06, which suggests the existence of defects, and it exhibits pyridinic N (18.24%), pyrrolic N (18.75%), graphitic N (47.12%), and oxidized N (15.89%), where the existence of abundance graphitic N can facilitate the ORR activity, and it affords high electrochemically active surface area (6 cm2 at 0.1 mg), and that can possibly enhance its ORR/OER activity and ZAB performance. It shows the better OER and ORR activity with low ΔE of 0.74 V. ZAB with Fe0.5Co0.5Ox-NrGO exhibits the high capacity of 756 mA h g−1 at 10 mA cm−2, the energy density of 904 W h kg−1, the power density of 82 mW cm−2, and OCP of 1.43 to 1.44 V. ZAB with Fe0.5Co0.5Ox-NrGO affords initial polarization voltage of 0.79 V (Fig. 8 c) with increased polarization voltage of 0.1 V after 60 cycles at 10 mA cm−2 with Zn plate as the anode and 6 M KOH with 0.2 M ZnCl2 as an electrolyte.Integrating Fe-doped NiOOH with graphene-encapsulated FeNi3 can enhance the OER activity and ZAB performance. Wang et. al. [119] have observed that FeNi3@G@Fe-NiOOH exhibits high OER activity with high ZAB performance. It was prepared by arc discharging method followed by electrochemical activation. They observed that FeNi3@G@Fe-NiOOH exhibits higher OER activity than FeNi3@Fe-NiOOH, Fe-NiOOH@G, and Fe-NiOOH. It is composed of Fe-doped NiOOH, which are grown on graphene-encapsulated FeNi3 nanodots. It contains Ni0, Ni3+ and Fe3+, and that can possibly enhance its OER activity and ZAB performance. It exhibits high OER activity (EJ=10: 1.52 V). ZAB with FeNi3@G@Fe-NiOOH affords initial polarization voltage of ~0.525 V with an increased polarization voltage of ~0.275 V after 360 cycles at 1 mA cm−2 with Zn foil as the anode and 6 M KOH with 0.2 M Zn(O2CCH3)2 as the electrolyte.Thus, various strategies including preparing MnO2 nanofilm on N-doped hollow graphene [120], integrating Mn3O4 with rGO-IL [125], integrating CoMn2O4 with NrGO [124], integrating MnCoFeO4 with N-rGO [126], functionalizing graphene oxide with 1-hexyl-3-methylimidazolium chloride molecules [121], integrating Co3O4 with N doped graphene [122], integrating Co3O4 nano-rods with reduced graphene oxide [118], preparing predominant metallic Co with small fraction of its oxides anchored on N-doped reduced graphene oxide [123], integrating amorphous bimetallic oxide with N-doped reduced graphene oxide [117], and integrating Fe-doped NiOOH with graphene-encapsulated FeNi3 [119] enhanced the ORR/OER activity and ZAB performance. Table 5 depicts the ORR/OER activity and zinc-air battery performance of various kinds of reported graphene with nitrides, sulfides, carbides, and other carbon composites based air catalysts [127–133]. Integrating Ni3FeN with NrGO can enhance its conductivity, and that can enhance the ORR/OER activity and ZAB performance. Fan et. al. [128] have observed that Ni3FeN-NrGO exhibits high ORR/OER activity with high ZAB performance. It was prepared by heat treatment at 700 °C for 2 h under NH3 atmosphere. They observed that Ni3FeN-NrGO exhibits higher conductivity than Ni3FeN, Ni3N/rGO and Ni3N. It is composed of 2D Ni-Fe nitride nanoplates (~8 nm), which are strongly anchored with the N-doped reduced graphene oxide, and it exhibits high conductivity, and that can possibly enhance its ORR/OER activity and ZAB performance. It exhibits high OER (EJ=10: 1.63 V) and ORR activity (E1/2: ~ 0.7 V; n: 3.8) with low ΔE of ~0.93 V. ZAB with Ni3FeN-NrGO affords initial polarization voltage of 0.77 V (Fig. 9 a) with negligible increased polarization voltage of 0.05 V after 180 cycles at 10 mA cm−2 with Zn as the anode and 4 M KOH as the electrolyte.Preparing CoSx@C3N4 integrated with reduced graphene oxide having porous structure can afford sufficiently exposed active sites, and that can enhance the ORR/OER activity and ZAB performance. Niu et. al. [127] have observed that CoSx@C3N4/rGO exhibits high ORR/OER activity with high ZAB performance. It was prepared by the following steps: At first, porous g-C3N4 was obtained by controlled pyrolysis of Co2+/melamine networks at 500 °C for 1 h under air atmosphere; Finally, CoSx@C3N4/rGO was prepared by mixing porous g-C3N4 with GO followed by heating with sulfur powder at 400 °C for 0.5 h under N2 atmosphere. It is composed of CoSx@C3N4 integrated with reduced graphene oxide, and it possesses a porous structure. It contains –C–O−, −C–C−, −C–C− (sp3) (Fig. 9 b), −C–S−, −NC–N−, Co2+, Co3+, S2−, polymeric S2 2−, −CS−, and– C–S n –C−, where the highly porous morphology with sufficiently exposed active sites can facilitate its catalytic activity, and that can possibly enhance its ORR/OER activity and ZAB performance. It exhibits high OER (EJ=10: 1.57 V) and ORR activity (E1/2: 0.78 V; n: 3.96) with low ΔE of 0.79 V. ZAB with CoSx@C3N4/rGO exhibits OCP of 1.38 V. ZAB with CoSx@C3N4/rGO affords initial polarization voltage of ~1.5 V with negligible increased polarization voltage after 394 cycles at 50 mA with Zn plate as the anode and 6 M KOH with 0.2 M Zn(O2CCH3)2·6H2O as the electrolyte. Thus, various strategies including integrating Ni3FeN with NrGO [128], and preparing CoSx@C3N4 integrated with reduced graphene oxide [127], improved the ORR/OER activity and ZAB performance.Integrating CoS2 with N, S co-doped graphene oxide can enhance the ORR/OER activity and ZAB performance. Ganesan et. al. [130] have observed that CoS2/N, S-GO exhibits high ORR/OER activity with high ZAB performance. It was prepared by solid-state thermolysis approach at 400 °C for 2 h using cobalt thiourea and graphene oxide. It is composed of crystalline cobalt sulfide nanoparticles, which are anchored on N, S co-doped graphene oxide, and that can possibly enhance its ORR/OER activity and ZAB performance. It exhibits high OER (EJ=10: 1.62 V) and ORR activity (E1/2: 0.79 V) with low ΔE of 0.83 V. ZAB with CoS2/N, S-GO exhibits high capacity of 767 mA h g−1. ZAB with CoS2/N, S-GO affords initial polarization voltage of 0.78 V with negligible increased polarization voltage after 70 cycles at 10 mA cm−2 with Zn plate as the anode and 6 M KOH as an electrolyte.Integrating CoSx with N, S co-doped graphene can enhance the ORR/OER activity and ZAB performance. Geng et. al. [131] have observed that CoSx@N, S-G exhibits high ORR/OER activity with high ZAB performance. It was prepared by hydrothermal treatment. It is composed of CoSx nanoparticles (~50 nm), which are anchored on N, S co-doped graphene, and that can possibly enhance its ORR/OER activity and ZAB performance. It exhibits high OER and ORR activity. ZAB with CoSx@N, S-G affords initial polarization voltage of ~0.95 V with negligible increased polarization voltage after 50 cycles at 1.25 mA cm−2, which indicates its high performance.Integrating NiCo2S4 with ultrathin S-doped graphene having high specific surface area and the electrochemically active surface area can enhance the ORR/OER activity and ZAB performance. Liu et. al. [129] have observed that S-G/NiCo2S4 exhibits high ORR/OER activity with high ZAB performance. It was prepared by hydrothermal treatment at 160 °C followed by sulfurization process at 300 °C for 2 h. They observed that S-G/NiCo2S4 exhibits higher specific surface area and electrochemically active surface area than NiCo2S4. It is composed of urchin-like NiCo2S4 microsphere, which is encapsulated by the ultrathin S-doped graphene (Fig. 9 c). The porous 3D-interconnected network contains Ni, Co, S, and C, which are uniformly distributed with high surface area (227 m2 g−1), and high electrochemically active surface area (14.1 mF cm−2). It exhibits I D/I G ratio of 0.91, which is higher than 0.79 of bare G, which suggests the generation of structural distortion and defects, and it contains Co2+, Co3+, Ni2+, Ni3+, C–C/CC, C–O, C–S, CO/OC–O, and C–S–C, and that can possibly enhance its ORR/OER activity and ZAB performance. It exhibits high OER (EJ=10: ~1.56 V) and ORR activity (E1/2: 0.88 V; n: 3.86 to 3.96) with low ΔE of 0.69 V. ZAB with S-G/NiCo2S4 exhibits high power density of 216.3 mW cm−2. ZAB with S-G/NiCo2S4 affords initial polarization voltage of 0.8 V with negligible increased polarization voltage of about 0 V after 150 cycles at 10 mA cm−2 with Zn plate as the anode and 6 M KOH with 0.2 M Zn(O2CCH3)2 as the electrolyte, while the ZAB powers a mini-fan (Fig. 9 d), which indicates its much high performance. Thus, several strategies including integrating CoS2 with N, S co-doped graphene oxide [130], integrating CoSx with N, S co-doped graphene [131], and integrating NiCo2S4 with ultrathin S-doped graphene [129] enhanced the ORR/OER activity and ZAB performance.Preparing Fe3C(Fe) with N-doped graphitic layers as shell and N-doped graphene as an interface layer can enhance the ORR activity and ZAB performance. Yang et. al. [132] have observed that Fe3C(Fe)@N-GrapLy exhibits high ORR activity with high ZAB performance. It was prepared by pyrolyzing a mixture of Prussian blue (PB) and glucose at 850 °C for 6 h under Ar atmosphere. It is composed of metallic Fe-cluster embedded in crystalline Fe3C nanoparticles, which are enwrapped in N-doped graphitic layers (~5 nm), where the interface layer between Fe3C(Fe) and graphitic layers could be the N-doped graphene. It contains pyridinic N, pyrrolic N, graphitic N and oxidized N, where pyridinic N and graphitic N can facilitate the activity for ORR, and it possesses high surface area (418 m2 g−1). The mesopores and macropores also can possibly enhance its ORR activity and ZAB performance. It exhibits high ORR activity. ZAB with Fe3C(Fe)@N-GrapLy exhibits the high capacity of 790 mA h g−1, the power density of 186 mW cm−2, and OCP of 1.53 V. Thus, preparing Fe3C(Fe) with N-doped graphitic layers as shell and N-doped graphene as interface layer [132] improved the ORR/OER activity and ZAB performance.Preparing N-doped vertically aligned carbon nanotubes on graphene foam can enhance the ORR activity and ZAB performance. Cai et. al. [133] have observed that N-CNTs-G exhibits high ORR activity with high ZAB performance. It was prepared by the following steps: At first, graphene was obtained on Ni foam by chemical vapor deposition; Then, CNTs was grown on G/Ni foam by plasma enhanced chemical vapor deposition, and the Ni template was removed by submerging CNTs-G/Ni foam in 1 M/1M FeCl3/HCl solution for overnight, while CNTs-G was activated by HNO3 solution; Finally, N-CNTs-G was obtained by coaxially polymerizing the polyaniline on the sidewalls of the CNTs-G followed by carbonization at 800 °C for 2 h under N2 atmosphere. It is composed of N-doped vertically aligned carbon nanotubes, which are supported by graphene foam, and it exhibits the I D/I G ratio of 0.60, and it exhibits high surface area (101.1 m2 g−1). It contains 91.64 atomic % of C, 2.11 atomic % of N and 6.24 atomic % of O; It contains pyridinic N, pyrrolic N, graphitic N and oxidized N, where pyridinic N and graphitic N can facilitate the activity for ORR, and that can possibly enhance its ORR activity and ZAB performance. It exhibits high ORR activity. ZAB with N-CNTs-G exhibits 58 mA cm−2 at 0.8 V. ZAB with N-CNTs-G affords initial polarization voltage of 0.87 V with negligible increased polarization voltage of ~0.05 V after 240 cycles at 10 mA cm−2 with Zn plate as the anode and 10 M KOH with 0.2 M Zn(O2CCH3)2 as the electrolyte.Fabrication of nitrogen doped graphene nanotube complexes can enhance the ORR activity and ZAB performance. Kong et. al. [134] have observed that N doped graphene nanotube complexes exhibits high ORR activity with high ZAB performance. It was prepared based on the hydrothermal treatment and annealing process, where urea is used as the nitrogen source and boric acid (H3BO3) is used as the splicing agent. It is composed of few-layer graphene and carbon nanotubes, which have been randomly mixed; It possesses 3D porous nanostructure; It contains interconnected macropores; It contains abundant defects; It possesses ultrahigh specific surface area and pore volume; It contains high density of pyridinic-N; It exhibits low charge transfer resistance; Thus, it could modify the electronic structure, afford optimal adsorption with intermediates, expose abundant active sites, facilitate the charge transfer process, enhance the conductivity, and that could enhance the ORR performance, and that could improve the ZAB performance. It exhibits high ORR activity (E1/2 = 0.89 V). ZAB with N doped graphene nanotube complexes exhibits high open circuit potential of 1.54 V, large power density of 149 mW cm−2, high specific capacitance of 873 mAh gZn −1, and high stability.Preparing heteroatoms doped CNT-graphene hybrids can enhance the ORR activity and ZAB performance. Huang et. al. [135] have observed that N, S co-doped CNT-graphene hybrids exhibits high ORR activity with high ZAB performance. It was derived from biomolecule (guanine), where it was obtained by pyrolysis of the guanine-sulfate and OCNT. It is composed of N, S co-doped CNT-graphene hybrids; It possesses 3D hierarchically porous structure; Thus, it could modify the electronic structure, afford optimal adsorption with intermediates, expose abundant active sites, facilitate the charge transfer, enhance the conductivity, and that could enhance the ORR performance, and that could improve the ZAB performance. It exhibits high ORR activity (E1/2 = 0.87 V). ZAB with N, S co-doped CNT-graphene hybrids exhibits high open circuit potential of 1.48 V, large power density of 188 mW cm−2, high specific capacitance of 800 mAh gZn −1, and high stability.Preparing porous carbon materials comprising of few-layer graphene sheets and carbon nanotubes can enhance the ORR activity and ZAB performance. Kong et. al. [136] have observed that carbon-tube-graphene complexes exhibits high ORR activity with high ZAB performance. It was prepared based on the hydrothermal and pyrolysis treatments, where graphene nanosheets and carbon nanotubes are used as building block, while boric acid (H3BO3) is used as splicing agent. It is composed of few-layer graphene sheets and carbon nanotubes, which have been self-assembled; It contains hierarchical porous carbon materials having sponge-like architecture; It contains interconnected macropores; It contains defects; It possesses high surface area and pore volume; It exhibits low charge transfer resistance; Thus, it could expose abundant active sites, facilitate the charge transfer, enhance the conductivity, and that could enhance the ORR performance, and that could improve the ZAB performance. It exhibits high ORR activity (E1/2 = 0.841 V). ZAB with carbon-tube-graphene complexes exhibits high open circuit potential of 1.38 V, large power density of 65 mW cm−2, and high stability.Preparing reduced graphene oxide modified heteroatom-doped ultra-thin hollow carbon spheres can enhance the ORR activity and ZAB performance. Sheng et. al. [137] have observed that rGO@N doped hollow carbon sphere composites exhibits high ORR activity with high ZAB performance. It was prepared based on the self assembly followed by carbonization followed by etching. It is composed of reduced graphene oxide modified nitrogen-doped ultra-thin hollow carbon spheres; It contains defects; It contains high density of pyridinic-N; Thus, it could modify the electronic structure, afford optimal adsorption with intermediates, expose abundant active sites, facilitate the charge transfer, enhance the conductivity, and that could enhance the ORR performance, and that could improve the ZAB performance. It exhibits high ORR activity. ZAB with N doped graphene nanotube complexes exhibits high open circuit potential of 1.54 V, large power density of 142 mW cm−2, high specific capacitance, and high stability.Preparing defect-rich carbon fiber having porous graphene skin can enhance the ORR activity and ZAB performance. Wang et. al. [138] have observed that defect-rich carbon fiber with porous graphene skin exhibits high ORR activity with high ZAB performance. It was prepared using high-temperature H2 etching approach. It is composed of defect-rich carbon fiber with porous graphene skin; Thus, it could modify the electronic structure, afford optimal adsorption with intermediates, expose abundant active sites, facilitate the charge transfer, enhance the conductivity, and that could enhance the ORR performance, and that could improve the ZAB performance. It exhibits high ORR activity. ZAB with N doped graphene nanotube complexes exhibits high open circuit potential of 1.54 V, large power density of 91.4 mW cm−2, high specific capacitance of 707 mAh g−1, and high stability.Thus, various strategies including preparing N-doped vertically aligned carbon nanotubes on graphene foam [133] fabrication of nitrogen doped graphene nanotube complexes [134], preparing heteroatoms doped CNT-graphene hybrids [135], preparing porous carbon materials comprising of few-layer graphene sheets and carbon nanotubes [136], preparing reduced graphene oxide modified heteroatom-doped ultra-thin hollow carbon spheres [137], and preparing defect-rich carbon fiber having porous graphene skin [138] enhanced the ORR/OER activity and ZAB performance.The sluggish kinetics of ORR and OER are often considered as the bottleneck of a variety of electrochemical energy conversion technologies, including water electrolyzers, metal-air batteries, and fuel cells. Hence, development of cheap and efficient ORR and OER catalysts are essential, which not only be the alternative for the expensive noble metal catalysts (ORR: Pt, OER: IrO2 or RuO2, and ORR/-OER: Pt-Ru/C) but also bring the electrochemical energy systems nearer to their theoretical limits. The ORR/OER activity and zinc-air battery performance of several types of graphene-based air catalysts such as graphene with non-metals, non-noble metals, metal oxides, nitrides, sulfides, carbides, and other carbon composites have been identified to develop promising ZABs.Various strategies including creating defects on the graphene [101], doping N with graphene [97], N-doped graphene obtained through carbonization of natural rice [100], preparing N-doped graphene with electron-withdrawing pyridinic N and electron-donating quaternary N [102], preparing N-doped exfoliated graphene [106], S-doped graphene foam prepared from food (idly) [103], preparing S, N co-doped graphene-like electrocatalyst [105], preparing defect enriched S, N co-doped graphene-like carbon [98], B, N co-doped graphene with graphitic N and BC3 [96], integrating holey graphene framework with CNTs [104], and preparing N, P co-doped carbon framework [99] enhanced the ORR/OER activity and ZAB performance.Moreover, several strategies including integrating Fe/Fe3C@C nanoparticles with graphene framework [107], integrating NiFe nanoparticles with graphene [115], integrating Vulcan carbon with CoFe-N-rGO [116], preparing Co nanoclusters on N-doped carbon [108], preparing Co/N/O tri-doped graphene [111], preparing cobalt nanoparticles encapsulated in N-enriched graphene shells with hollow graphene spheres [113], preparing Co, N-co-doped CNT/graphene heterostructure [109], preparing graphene matrix with Cu(I)-N active sites [114], integrating Ag NW with graphene aerogel [112], and preparing Fe–N active sites and integrating CNTs and graphene with MOF [110] improved the ORR/OER activity and ZAB performance.In addition, various strategies including preparing MnO2 nanofilm on N-doped hollow graphene [120], integrating Mn3O4 with rGO-IL [125], integrating CoMn2O4 with NrGO [124], integrating MnCoFeO4 with N-rGO [126], functionalizing graphene oxide with 1-hexyl-3-methylimidazolium chloride molecules [121], integrating Co3O4 with N doped graphene [122], integrating Co3O4 nano-rods with reduced graphene oxide [118], preparing predominant metallic Co with small fraction of its oxides anchored on N-doped reduced graphene oxide [123], integrating amorphous bimetallic oxide with N-doped reduced graphene oxide [117], and integrating Fe-doped NiOOH with graphene-encapsulated FeNi3 [119] enhanced the ORR/OER activity and ZAB performance.Moreover, various strategies including integrating Ni3FeN with NrGO [128], preparing CoSx@C3N4 integrated with reduced graphene oxide [127], integrating CoS2 with N, S co-doped graphene oxide [130], integrating CoSx with N, S co-doped graphene [131], integrating NiCo2S4 with ultrathin S-doped graphene [129], preparing Fe3C(Fe) with N-doped graphitic layers as shell and N-doped graphene as interface layer [132], preparing N-doped vertically aligned carbon nanotubes on graphene foam [133] fabrication of nitrogen doped graphene nanotube complexes [134], preparing heteroatoms doped CNT-graphene hybrids [135], preparing porous carbon materials comprising of few-layer graphene sheets and carbon nanotubes [136], preparing reduced graphene oxide modified heteroatom-doped ultra-thin hollow carbon spheres [137], and preparing defect-rich carbon fiber having porous graphene skin [138] enhanced the ORR/OER activity and ZAB performance.The vital factors governing the performance of ZAB should be considered in future research to provide superior performance in practical applications: 1. Recently significant efforts have been done to fabricate several kinds of high performance air catalysts for ZAB but the high performance air catalysts are very limited because of the poor electrical conductivity, poor thermal stabilities, inferior cycling stability, restricted specific capacity, and slower kinetic diffusion. Hence, additional progresses are indeed necessary by fabricating high performance graphene-based air catalysts including graphene with non-metals, non-noble metals, metal oxides, nitrides, sulfides, carbides, and other carbon composites. 2. Recently, several kinds of high performance graphene-based air catalysts for ZAB have been reported. Nevertheless, very limited facile synthesis route have been explored for the fabrication of graphene-based air catalysts. Therefore, additional efforts are obviously needed by exploring facile and green synthesis route to diminish or alleviate the use or generation of hazardous substances for the fabrication of high performance graphene-based air catalysts for ZAB. Recently significant efforts have been done to fabricate several kinds of high performance air catalysts for ZAB but the high performance air catalysts are very limited because of the poor electrical conductivity, poor thermal stabilities, inferior cycling stability, restricted specific capacity, and slower kinetic diffusion. Hence, additional progresses are indeed necessary by fabricating high performance graphene-based air catalysts including graphene with non-metals, non-noble metals, metal oxides, nitrides, sulfides, carbides, and other carbon composites.Recently, several kinds of high performance graphene-based air catalysts for ZAB have been reported. Nevertheless, very limited facile synthesis route have been explored for the fabrication of graphene-based air catalysts. Therefore, additional efforts are obviously needed by exploring facile and green synthesis route to diminish or alleviate the use or generation of hazardous substances for the fabrication of high performance graphene-based air catalysts for ZAB. Mohammed-Ibrahim Jamesh: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Validation, Visualization, Writing - original draft, Writing - review & editing. Prabu Moni: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Validation, Visualization, Writing - original draft, Writing - review & editing. A.S. Prakash: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Validation, Visualization, Writing - review & editing. Moussab Harb: Funding acquisition, Supervision.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.One of the authors (Dr. M.I.J) thanks to the Science and Engineering Research Board (SERB), Department of Science and Technology, Government of India, for funding under National Post-Doctoral Fellowship scheme with the reference no. PDF/2017/000015. One of the authors, Dr. Prabu Moni grateful to the Department of Science and Technology (DST), New Delhi, India for awarding INSPIRE Faculty Award (DST/INSPIRE/04/2016/000530). One of the authors, Dr. MOUSSAB Harb thanks to the King Abdullah University of Science and Technology (KAUST).

The development of cheap and efficient oxygen reduction and evolution reaction catalysts are important, which not only push the electrochemical energy systems including water electrolyzers, metal-air batteries, and fuel cells nearer to their theoretical limits but also become the substitute for the expensive noble metal catalysts (Pt/C, IrO2 or RuO2 and Pt-Ru/C). In this review, the recently reported potential graphene-based air catalysts such as graphene with non-metals, non-noble metals, metal oxides, nitrides, sulfides, carbides, and other carbon composites are identified in-light-of-their high oxygen reduction reaction/oxygen evolution reaction activity and zinc-air battery performance for the development of high energy density metal-air batteries. Further, the recent progress on the zinc-air batteries including the strategies used to improve the high cycling-performance (stable even up-to 394 cycles), capacity (even up-to 873 mAh g−1), power density (even up-to 350 mW cm−2), and energy density (even up-to 904 W h kg−1) are reviewed. The scientific and engineering knowledge acquired on zinc-air batteries provide conceivable development for practical application in near future.

Electrochemical water splitting is a well-known promising technology for the sustainable production of high-purity hydrogen [1]. The development of efficient electrocatalysts primarily consisting of earth-abundant elements [2,3] for hydrogen evolution reaction (HER) and oxygen evolution reaction (OER) as well, is of great importance for large-scale hydrogen production by water splitting [4,5]. During the last decades, transition-metal-based materials have been and are still considered as alternative HER/OER electrocatalysts due to their earth-abundance, low-cost, and promising activity [6]. Specifically, transition metal alloys [7], nitrides [8], phosphides [9], metal oxides [10], metal hydroxides [11], chalcogenides [12], and other compounds have been proved as active materials for HER/OER electrocatalysis [13]. Furthermore, transition-metal-based materials appear as possible efficient bifunctional electrocatalysts for both HER and OER, which are cost-effective and with higher efficiency in practical applications [14,15]. Since the different reaction mechanisms of HER and OER requires different structural and electronic properties for the electrocatalysts, bimetallic and multi-metallic compounds-based materials appear as more promising candidates for the overall water splitting [16,17]. For example, Chen et al. [18] fabricated porous Fe-Mo oxide hybrid nanorods on nickel foam (NF) as efficient bifunctional electrocatalysts for water splitting. Li et al. [19] synthesized a porous amorphous Ni/Ni-Fe-Mo suboxide nanoplates array on NF. As well known, NF is a popular substrate that can rivet and disperse catalytic components, thus resulting in high loadings of active catalytic components with the consequence in providing abundant catalytic sites [4,20]. Moreover, NF can be used as a substrate and be directly transformed into one of the multiple active components of the catalyst [21]. Fei et al. [22] synthesized ultrathin Fe-doped Ni3S2 arrays on NF for efficient water splitting, which transformed NF into Ni3S2 via a Na2S-induced chemical etching process. Although previous works have made prominent progress, developing low-cost, highly efficient, and durable bifunctional electrocatalysts by combining both the advantages of multimetallic nanomaterials and the utilization of NF is still significant and challengeable [23,24].Herein, we report the development of an efficient bifunctional Ni-Fe/NiMoNx electrocatalyst deposited on NF, using a hydrothermal method followed by NH3 treatment. During the hydrothermal process, Fe and Mo elements were introduced, while partial NF substrate was transformed into Ni(OH)2. The following NH3 treatment resulted in the formation of Ni-Fe/NiMoNx, which showed high-efficiency electrocatalysis toward both HER and OER in alkaline medium. As a consequence, the prepared Ni-Fe/NiMoNx electrocatalyst showed efficient HER and OER performance at low overpotentials of 49 and 260 mV at 20 mA cm−2, respectively. The overall water splitting of Ni-Fe/NiMoNx couple electrodes required a low cell voltage of 1.54 V for 10 mA cm−2.The Nickel foam (NF) was pretreated by ultrasonication in 3.0 M HCl solution, ethanol, and deionized water. Fe(NO3)3·9H2O (1.5 mmol Fe) and (NH4)6Mo7O24·4H2O (1.5 mmol Mo) were dissolved in 60 mL deionized water. After magnetic stirring for 1 h, the solution was transferred into a 100 mL Teflon-lined stainless-steel autoclave, and then one piece of pretreated NF (2 × 5 cm) was immersed into the solution. The autoclave was sealed and maintained at 150 °C for 6 h. After cooled down at room temperature, the obtained NiMoFe-Pre was taken out and cleaned with deionized water and ethanol.The prepared NiMoFe-Pre was transferred into a tubular furnace with NH3 atmosphere flow and annealed at 400 °C for 2 h, and then Ni-Fe/NiMoNx was obtained after natural cooling at room T. The NiMoFe-H400 was prepared in the same tubular furnace at the same temperature but treated in 10 vol% H2/Ar gas atmosphere.Powder X-ray diffraction (XRD) patterns were obtained on Rigaku Smartlab equipment. Scanning electron microscopy (SEM) experiments were performed on a Zeiss Sigma 500 microscope. Transmission electron microscopy (TEM) images were acquired on an FEI Tecnai G2 F30 microscope. X-ray photoelectron spectroscopy (XPS) tests were performed on a Thermos Scientific spectrometer.The electrochemical testing was performed on a VSP-300 (BioLogic, France) electrochemical workstation. A three-electrode system in 1.0 M KOH was used, where the prepared NF-based electrode (0.5 × 1.0 cm) was used as working electrode, Hg/HgO as a reference electrode, and graphite rod and platinum wire were used as counter electrodes for HER and OER, respectively. The iR correction was applied to all the LSV curves, and all potentials were converted into the RHE (E RHE  = E Hg/HgO  + 0.098 V + 0.0592pH). Electrochemical impedance spectroscopy (EIS) tests were recorded in the frequency range between 50 mHz and 100 kHz with an amplitude of 5 mV, and tested at the voltages of −0.022 V for HER and 1.480 V for OER.The electrode was prepared by the hydrothermal method followed by NH3 gas treatment (Fig. 1a). The NiMoFe-Pre nanosheets with smooth surfaces were first grown on the nickel foam using the hydrothermal method. During the hydrothermal process, Mo and Fe elements were introduced, and the substrate NF was partially corroded to become Ni precursor due to the presence of H+ from ammonium hydrolysis and the excess of NO3 − in solution [25]. The powder XRD patterns showed that the formed NiMoFe-Pre precursor was mainly composed of Ni(OH)2, Fe2O3, and MoO3 (Fig. S1), which proved the successful loading of Fe, Mo, and the utilization of NF. To obtain Ni-Fe/NiMoNx, the NiMoFe-Pre precursor nanosheets were further treated in NH3 gas atmosphere. SEM image showed that numerous nanoparticles are formed on the surface of nanosheets after being annealed in NH3 gas atmosphere (Fig. 1a). The presence of nanoparticles should be due to phase separation of different components that are formed by ammonia gas treatment, beneficial to the increase of the specific surface area and to the exposure of active sites.Analysis of TEM images further exhibited that Ni-Fe/NiMoNx were composed of ultrathin nanosheets anchoring with nanoparticles (Fig. 1b). The high-resolution TEM (HRTEM) image of Ni-Fe/NiMoNx showed typical lattice spacings of 0.174 and 0.203 nm on nanoparticles, corresponding to the (200) and (111) planes, respectively, of Ni (Fig. 1c), indicating the formation of Ni nanoparticles on nanosheets surfaces during nitridation. After fast Fourier transformation (FFT) and inverse FFT, the lattice spacing was clearly observed, where 0.215 nm correspond to the (002) plane of Ni3N, and 0.244 nm correspond to the (110) plane of Mo5N6, indicating the formation of nitrides by nitridation. In the selected area electron diffraction (SAED) image, the (200), (111), (110), and (311) planes of Ni and (111) plane of Ni3N were displayed (Fig. 1d). The energy-dispersive X-ray spectroscopy (EDS) images showed that Ni element mainly exist in the nanoparticles and partially on the nanosheets, while Fe, Mo, N, and O elements were uniformly distributed on the nanosheets (Fig. 1e). Combining with HRTEM and EDS results, the Ni nanoparticles and the main elements of nanosheets are preliminarily clarified.The compositions of Ni-Fe/NiMoNx were further investigated by powder XRD patterns (Fig. 2a). They show the existence of Ni, Ni3N, NiO, and Mo5N6. There are no distinct characteristic peaks of the Fe element, which is probably due to the coverage by the adjacent strong characteristic peaks of nickel. XPS analyses were conducted to study the surface chemical states in the Ni-Fe/NiMoNx. The XPS survey scan confirmed the presence of Ni, Mo, Fe, O and N elements in the catalyst (Fig. 2b). The high-resolution spectra of Ni 2p (Fig. 2b) evidenced the presence of Ni2+ and Ni0 due to NiO and Ni phases, whereas Ni1+ appeared is due to the Ni3N phase. The coexistence of NiO, Ni, and Ni3N indicates the incomplete nitridation after the NH3 gas treatment.In the case of Fe 2p XPS spectra, the Fe0 signal proved the existence of metallic Fe (Fig. 2d). The presence of Fe0 and Fe2+ further proved the incomplete nitridation process by NH3 gas treatment. In the case of Mo 3d spectra, peaks of Mo6+, Mo4+, and Mo3+ were located in the regions of Mo 3d3/2 and Mo 3d5/2 (Fig. 2e). Besides, there is another peak which is located in the region of Mo 3p (Fig. 2f). Characteristic peaks of metal-N and NH appeared in the region of N 1 s. The metal-N bonding is obviously from Ni3N and Mo5N6. The formation of the NH bonding might be due to the hydrogen adsorption properties of metal nitrides. Specifically, binding with N atoms alters the d-band structure of the host metal, thereby contracting the d-band of the metal [8]. This alteration changes the coupling state between the adsorbed hydrogen s-band and the metal d-band, thus the adsorption of hydrogen tends to the HER process [26,27]. All these XPS results combined with HRTEM and powder XRD data showed the incomplete reduction and nitridation occurred during NH3 gas treatment, and which resulted in the formation of Ni-Fe/NiMoNx hybrid nanosheets.The HER performance of Ni-Fe/NiMoNx and other samples was characterized in 1.0 M KOH solution. As can be observed from the linear sweep voltammetry (LSV) curves (Fig. 3a), Ni-Fe/NiMoNx shows low overpotentials of 49 and 107 mV to achieve the hydrogen evolving currents of 20 and 100 mA cm−2, respectively (Fig. 3a), which is comparable to Pt/C/NF (η20 = 26 mV) and much lower than NiMoFe-H400 (η20 = 139 mV), NiMoFe-Pre (η20 = 190 mV), and NF (η20 = 280 mV). Ni-Fe/NiMoNx has also a low Tafel slope at 70.74 mV dec−1 (Fig. 3b), better than NiMoFe-H400, NiMoFe-Pre, and NF.The HER performance is comparable and even better than recently reported HER electrocatalysts (Table S1, Supporting Information). Nyquist plots were obtained from the EIS tests (Fig. 3c). On the basis of equivalent circuit model, the charge transfer resistance increases in the order: Ni-Fe/NiMoNx (Rct = 1.7 Ω) < NiMoFe-H400 (Rct = 11.8 Ω) < NF (Rct = 57.55 Ω) < NiMoFe-Pre (Rct = 220.9 Ω). The lower impedance value means a higher charge transfer rate and faster electrode kinetics for HER. Previous works have proved that nitrides exhibit low electrical resistance and bind exceptionally to both water molecules and hydrogen atoms [28,29]. Furthermore, the nitrides and metallic Ni and Fe are capable to permit rapid electron transfer between the active surface sites of the catalyst and the NF current conductor [10]. In order to highlight the importance of nitrides for HER, NiMoFe-H400 was prepared with 10 vol% H2/Ar gas treatment instead of NH3. Their powder XRD and XPS results proved the presence of similar components such as Ni, NiO, and Fe, which is due to the incomplete reduction, except the case of nitrides (Figs. S2 and S3, Supporting Information). The Ni-Fe/NiMoNx possesses lower overpotential, Tafel slope, and impedance than the NiMoFe-H400 solid. This indicates the presence of Ni3N and Mo5N6 which can significantly improve charge transfer rates, thus highlighting the HER enhancement after NH3 gas pre-treatment.NH3 gas treatment at different temperatures exhibited different HER performance. SEM and powder XRD patterns indicated different degrees of nitridation at different temperatures, ca. 300, 400, and 500 °C (Figs. S4 and S5, Supporting Information). Compared with Ni-Fe/NiMoNx (N400), the powder XRD patterns of NiMoFe-N300 and NiMoFe-N500 showed significant diffraction peaks of NiO and Mo5N6, respectively. This indicates that higher temperatures of NH3 gas treatment can lead to a higher degree of nitridation. As shown in Fig. S6 (Supporting Information), Ni-Fe/NiMoNx (N400) showed lower overpotential and lower impedance than NiMoFe-N300 (η20 = 245 mV, Rct = 113.2 Ω) and NiMoFe-N500 (η20 = 56 mV, Rct = 4.92 Ω). Comparative analyses of these catalysts led us to conclude that Ni-Fe/NiMoNx has the best HER performance (Fig. 3d). Furthermore, to calculate the double-layer capacitance (Cdl), the cyclic voltammograms at various scan rates for samples in the non-faradaic capacitance current range were obtained (Fig. S7, Supporting Information). The Cdl of Ni-Fe/NiMoNx was 22.03 mF cm−2, which is much higher than those obtained over NiMoFe-H400, NiMoFe-Pre, NF, and NiMoFe-N300 (Fig. 3e and Fig. S6, Supporting Information). Since Cdl is positively associated with electrochemical surface area (ECSA), it is concluded that Ni-Fe/NiMoNx has higher ECSA and thus more exposed electrocatalytic active sites.Stability is another critical parameter for the electrocatalytic performance evaluation of the solid. In Fig. 3f, there is no significant difference in LSV curves before and after 1000 cycles of voltammetry scanning. The electrocatalyst also exhibits durability at the current density of −100 mA cm−2 for 40 h with a decline of only 15 mV. The SEM image after long-term HER showed well-defined nanosheet geometry, further confirming the good structural stability of electrocatalyst (Fig. S8, Supporting Information).Ni-Fe/NiMoNx achieved the oxygen-evolving current of 20 and 100 mA cm−2 at overpotentials of 260 and 292 mV, respectively (Fig. 4a), with Tafel slope of 39.26 mV dec−1 (Fig. 4b). The overpotential is significantly lower than that of NiMoFe-Pre (η20 = 280 mV) and NF (η20 = 356 mV). Notably, NiMoFe-H400 exhibits similar overpotential (η20 = 262 mV) and Tafel slope (39.26 mV dec−1) for Ni-Fe/NiMoNx. The reason behind this is the similar chemical states and components formed by the incomplete reduction, and the collective effect of these multiple components makes the easy formation of metal (oxy)hydroxide active sites for OER, and the consequent adsorption of *O, *OH, and *OOH intermediates neither too strong nor too weak. Thus, it enhanced the rate of processes of both adsorption of intermediates and evolution of O2, resulting in rapid OER kinetics with obviously lower Tafel slopes after the incomplete nitridation/reduction [30,31]. Both the low overpotential and Tafel slope indicate the good electrocatalytic activity of Ni-Fe/NiMoNx for OER, which outperforms the recently reported non-precious metal electrodes (Table S1, Supporting Information). EIS tests show the high charge transfer capability of Ni-Fe/NiMoNx (Rct = 7.9 Ω), which is smaller than that in other samples.The OER performances of NiMoFe-N300 and NiMoFe-N500 were also tested to investigate the effect of the annealing temperature (Fig. S9, Supporting Information). It was found that calcination temperature in the NH3 gas atmosphere can affect the degree of nitridation and alter the OER performance. Comparative analysis of the OER performance of different catalysts showed that Ni-Fe/NiMoNx has the best OER performance (Fig. 4d). Besides, Ni-Fe/NiMoNx has excellent stability in OER. The LSV curves before and after 1000 cycles of voltammetry scanning are almost coincident as seen in Fig. 4e. After 40 h long-term testing at 100 mA cm−2, the potential has a small fluctuation of only 9 mV (Fig. 4f) [31]. After long-term OER testing, the SEM image was obtained which confirms the good structural stability (Fig. S10, Supporting Information).The overall water splitting performance was performed in an alkaline electrolyzer with a two-electrode configuration, utilizing two Ni-Fe/NiMoNx electrodes as the anode and cathode. The polarization curves showed that Ni-Fe/NiMoNx-based electrolyzer requires a cell voltage of only 1.54 V at a current density of 10 mA cm−2, whereas 1.83 V is required for NF||NF. The water splitting performance of the present catalyst appears better than many reported electrocatalysts, such as P-Co3O4/NF (η10 = 1.63 V) [32], ZnCo2S4/NF (η10 = 1.66 V) [33], and NiCo2S4/NF (η10 = 1.61 V) [34]. The LSV curves before and after 1000 cycles of voltammetry scanning were almost coincident, as shown in Fig. 5c. The durability test showed a stable potential at 100 mA cm−2 for over 40 h, both exhibiting remarkable stability for water splitting (Fig. 5d). Therefore, the above results and discussion led us to confirm that Ni-Fe/NiMoNx is a highly efficient and durable catalyst for water splitting.An efficient and durable HER and OER bifunctional Ni-Fe/NiMoNx electrocatalyst was successfully designed and prepared. The Ni-Fe/NiMoNx showed low overpotentials of 49 and 260 mV for HER and OER at a current density of 20 mA cm−2, respectively, with Tafel slopes of 70.74 and 39.26 mV dec−1. For overall water splitting, Ni-Fe/NiMoNx-based electrolyzer requires low cells voltage, ca. 1.54 V for a current density of 10 mA cm−2 in 1.0 M KOH, which is durable at a constant current density of 100 mA cm−2 for 40 h. Enhanced HER and OER performance is due to the combined effect of the presence of multiple active electrocatalytic components, after NH3 gas pre-treatment. Besides, this work highlights the effect of NF and provides a new strategy for the rational design of bifunctional water electrolysis catalysts. Yu Qiu: Methodology, Formal analysis, Investigation, Data curation, Writing – original draft. Mengxiao Sun: Methodology, Formal analysis, Investigation, Writing – original draft. Jia Cheng: Formal analysis, Investigation. Junwei Sun: Formal analysis, Investigation. Deshuai Sun: Methodology, Formal analysis, Resources, Writing – review & editing, Supervision. Lixue Zhang: Conceptualization, Methodology, Formal analysis, Resources, Writing – review & editing, Supervision, Project administration, Funding acquisition.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This study was financially supported by the National Natural Science Foundation of China (No. 22075159), Taishan Scholar Program (No. tsqn202103058), and the Youth Innovation Team Project of Shandong Provincial Education Department (No. 2019KJC023). Supplementary material Image 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.catcom.2022.106426.

Fabricating bifunctional electrocatalysts for hydrogen evolution reaction (HER) and oxygen evolution reaction (OER) is challengeable. Herein, an efficient and durable bifunctional Ni-Fe/NiMoNx electrocatalyst has been synthesized on nickel foam. This electrocatalytic system exhibits significant HER and OER performances with overpotentials of 49 and 260 mV at 20 mA cm−2, respectively. For overall water splitting, Ni-Fe/NiMoNx electrodes require only 1.54 V for 10 mA cm−2. This enhanced HER and OER electrocatalytic performance is due to the combined effect of multiple active components formed after NH3 treatment. This work provides a new strategy for the rational design of bifunctional electrocatalysts for overall water splitting.

Ever since the beginning of the twentieth century, the demand for energy and technological advances have been intercorrelated with the rapid development of new processes for the conversion of different forms of energy. The development and expansion of the car generated the increased demand for the liquid instead of previously dominant solid fuels. Most fractions of crude oil, the only abundant resource of liquid energy carriers, however, are not suitable for combustion in engines. Following the demand of lower molecular weight products, thermal cracking, the decomposition of longer into shorter hydrocarbons was developed. If a solid acid catalyst is added, the reaction can occur at significantly lower temperatures, and the selectivity to fuels with a higher octane number is increased. Catalysts, therefore, act by facilitating the overall reaction as well as favoring certain products, resulting in the reduced energy consumption of the process. This and other closely related catalytic processes nowadays are used to produce almost all available petrol, jet, and diesel fuels. Beyond hydrocarbon cracking, catalytic materials have been used for a variety of reactions such as the Haber-Bosch process for the fixation of atmospheric nitrogen, the production of hydrogen from natural gas, the synthesis of plastics, and commodity and fine chemicals—most of which rely on the use of metallic transition metals on metal oxides. 1 As shown in Figure 1 A, the energy consumption based on nuclear and coal resources are not expected to increase any further in the next few decades. The demand for petroleum and other liquid fuels will continue to increase in the upcoming decades while renewable resources and natural gas, both of which are much more directly related to catalysis than nuclear energy and coal, are projected to play a dominant role by 2050. It becomes apparent that the industrial sector is currently responsible for more than half of the global energy consumption, with a projected slight decrease of 4.2% in 2050 (Figure 1B). Similarly, the transportation sector accounts for around another quarter of the global energy consumption and will moderately rise. The commercial and residential energy consumption makes up less than 25% in 2012, and will continue to be less than a quarter by 2050. When taking a closer look at the energy consumption of industrial subsectors (Figure 1C), catalysis-related processes such as the production of basic chemicals will witness an overall growth in significance until 2040. The same trend is obvious for the transportation subsectors with an obvious increase in the use of natural gas and jet fuels whereas the consumption of diesel and motor gas are projected to plummet. 2–4 Despite its already high significance in current energy-related industries, we expect catalytic processes to grow in importance in the future for the utilization of both conventional and unconventional energy resources. In particular, a number of above-mentioned chemical transformations have already become active research topics in SAC.Reports on atomically dispersed ionic transition metals have been reported earlier. 5–11 Then the term “single-atom catalysis” (SAC) was coined in 2011 by Zhang, Li, Liu, and colleagues, and has spurred tremendous developments. The evolution of related concepts and name conventions have been detailed recently, 12 and herein, we will focus the scope of this review predominantly on energy-related applications published in the recent decade. 13 Comprehensive introduction on the synthetic strategies of SACs, 14–16 specific types of support materials, 17–19 electrocatalysis, 16,18,20 and photocatalysis 21,22 could be found in excellent reviews that were recently published. In contrast to nanoparticle-based catalysts, where a significant fraction of metal atoms is buried below the surface and is thus inaccessible to reactants, SACs offer the maximum possible atom economy and maximized metal-support interactions. SACs exhibit electronic states ranging from positively to negatively charged and carry the absence of neighboring metallic atoms, significantly different from metallic nanoparticles. Furthermore, nanoparticles tune properties through d-band, while SACs display unique tunability because of their homogeneous-like ability to control the frontier orbital geometry and energy of the active sites. Meanwhile, the geometric structures of SACs are also different from that of homogeneous catalysts. 23 The metal species of homogeneous catalysts are coordinated by ligands or substrates. The geometric structures are flexible and the coordination environment can be changed easily during the reaction conditions. For single-atom catalysts, the geometric structures are partially restricted by the support, which could be regarded as a rigid, bulky ligand to coordinate single-atom metals. Those unparalleled properties give rise to unique reactivity patterns and have thus enabled their potential use in a range of energy applications. 12,16,24 Figure 2 depicts the conversion of energy carriers, in particular, fossil resources such as natural gas, crude oil, and coal as well as renewable energy such as solar, wind, hydro, and biomass over SACs. The products of those reactions might be the final products, or represent energy carriers that are used for grid-scale or transportation-scale energy storage including hydrocarbon, oxygenate, hydrogen, and ammonia fuels as well as a variety of chemicals and electrochemical energy storage systems. Each of those sectors will be reviewed and assessed individually in the context of SAC, and a general future perspective for the field of SAC for energy applications will be provided.Hydrocarbon fuels are currently produced from fossil fuels (coal, petroleum, and natural gas). 25 They are likely to remain as the principal sources of transportation fuels for the next decades. According to the International Energy Outlook 2017 released by the U.S. Energy Information Administration, the world primary fossil fuels consumption reached 514 quadrillion Btu in 2017, accounting for 82% of the global primary energy consumption. The global fossil fuels consumption is projected to reach 739 quadrillion Btu in 2040. Due to the diminishing fossil fuel reserves and continuous increase in CO2 emission from human activity, it becomes desirable to develop efficient catalytic systems to produce hydrocarbon fuels from sustainable resources such as biomass and CO2. 26 As the most abundant renewable organic carbon source, biomass is an attractive alternative feedstock for fuels. In many developing countries, a large percentage of fuels consumed is derived from biomass. 27 The transformation of oxygen-rich biomass to hydrocarbon fuels requires oxygen removal reactions such as hydrodeoxygenation (HDO) to form molecules that have desirable properties for upgrading. The traditional HDO catalysts generally suffer from poor catalytic performance and fast deactivation at high temperature. To improve the activity and stability of HDO catalysts, Tsang and co-workers developed a single-atom Co/MoS2 catalyst, in which isolated Co atoms were anchored covalently to sulfur vacancies of MoS2 monolayers. 28 The single Co atoms were observed in the HAADF-STEM images (Figure 3 A), and DFT calculation was carried out to simulate the geometric structure of Co (Figure 3B). The prepared Co/MoS2 catalyst with a large number of Co-S-Mo active sites exhibited excellent performance in the selective HDO of lignin-derived 4-methylphenol to toluene. The high catalytic performance of Co/MoS2 allowed the reaction to proceed at 180°C, which normally requires 300°C to occur. The lower operation temperature triggers energy saving, as heating is one of the main costs in large-scale industry applications, potentially pushing the hydrocarbon fuels production from biomass to commercial viability. Later, Tsang and co-workers designed a bifunctional catalyst consisting of atomically dispersed Pd and ultrasmall molybdenum phosphate nanoparticles for the conversion of phenolic monomers as well as wood and bark-derived oligomers into liquid hydrocarbons. 29 The prepared catalyst showed almost complete conversion of phenol to cyclohexane at 383 K. It was proposed that atomically dispersed Pd species promoted the hydrogenation of phenol to cyclohexanol, which was dehydrated to cyclohexene by Brønsted and Lewis acid sites on MoO3-P2O5 nanoparticles. Finally, the formed cyclohexene was hydrogenated to cyclohexane. The bifunctional catalyst also displayed state-of-the-art activity for the production of hydrocarbon fuels from water-insoluble bio-oil with a yield of 29.6 wt % under mild conditions.The production of hydrocarbon fuels from CO2 reduction may both mitigate global climate change and secure energy supply, but one shall keep in mind that every step for the transformation of CO2 needs energy, including capture, storage, and conversion. The cost of capturing CO2 from the air was estimated to range from $94 to $232 per ton of CO2. 30 The advancement of techniques is likely to further bring down the cost to make CO2 a viable carbon source. In parallel, a cheap and renewable energy source such as electricity or H2 is needed to upgrade CO2. Another critical issue is to develop highly active catalysts that are able to lower the reaction temperature and pressure without compromising activity, thus decreasing the energy consumption. Besides, catalysts with excellent selectivity would avoid the energy-intensive separation and purification process. Recent studies suggest that SACs are promising catalysts for CO2 transformation into hydrocarbons.The desirable CO2 reduction pathway is the direct one-step conversion into hydrocarbons and oxygenates (vide infra). Most of the currently available catalysts, however, can only reduce CO2 to CO, which is the first step towards the production of hydrocarbon fuels, 31 but recent advances suggest SACs might be exceptional. SACs showed tunable selectivity towards CO 32–34 and hydrocarbons. 35 Single atoms Pt/Pd supported on g-C3N4 for the photocatalytic reduction of CO2 was studied using DFT calculation by Du and co-workers. 36 In the presence of Pt or Pd single atoms, the absorption edge of g-C3N4 was extended from 2.7 to 0.2 eV as a result of the electron excitation from d band of the metal to the conduction band of g-C3N4. For Pd1/g-C3N4, HCOOH is the favored product with an activation barrier of 0.66 eV; however, Pt1/g-C3N4 preferred the formation of CH4 with a barrier of 1.16 eV. Head-Gordon and co-workers further examined in silico the electrocatalytic reduction of CO2 to hydrocarbons on 28 single-atom alloys (SAAs), in which single-atom M species (M = Cu, Ni, Pd, Pt, Co, Rh, and Ir) were dispersed on the host Au or Ag. 37 The SAAs with M = Co, Rh and Ir, showed desirable performance in the conversion of CO2 to methane. It was calculated that the host metals Au or Ag were responsible for the reduction of CO2 to CO, which was captured and further converted to methane by the nearby single-atom species. The performance of transition metals in CO2 electrocatalytic reduction was limited by the scaling relationship. 38 To break it, Jung and co-workers studied the catalytic performance of TiC as well as single metal atoms doped-TiC during CO2 hydrogenation by DFT. 39 As shown in Figures 3C and 3D, the binding energy of both *COOH and *CHO on pure metals depends linearly on the binding energy of *CO; however, a nonlinear relationship was observed on TiC-supported SACs. These calculations offer encouraging motivation to develop and characterize SACs for CO2 conversion. The theoretical calculation predicted that single-atom Cu dispersed on the CeO2 (110) surface could induce the formation of three oxygen vacancies in each neighboring Cu atom, and promote the conversion of CO2 to methane. 40 Experimental results were consistent with the prediction that an excellent selectivity towards CH4 was achieved on mesoporous CeO2 nanorod supported single-atom Cu species.In 2016, Ye and co-workers reported an efficient single-atom Co catalyst supported on porphyrin-based MOF for the conversion of CO2 to CH4 under visible light, mimicking the photosynthetic process in nature. 35 Both experiment and theory demonstrated that atomically dispersed Co atoms improved the electron-hole separation efficiency in the metal-organic framework (MOF). The photogenerated electrons were transferred efficiently from MOFs to single Co atoms. Compared with the MOF without Co, the incorporation of single-atom Co to MOF increased the formation rate of CH4 by up to 6 times.Methane is the main component of natural gas. According to the U.S. Energy Information Administration, the global methane consumption was 130 quadrillion Btu in 2017, and is projected to rise to 218 quadrillion Btu in 2050. 2–4 Around 90% of global natural gas production is used in the direct combustion for electricity generation and industrial heating, 41 and less than 1% natural gas is being used as transportation fuels in the U.S. 42 Industry is interested in the transformation of natural gas to easily transportable vehicle fuels such as aromatics and higher-value compounds to reduce the need for coal and petroleum. 43 Both indirect and direct strategies for the conversion of methane to higher hydrocarbon fuels have been explored. 44 The indirect methods require the formation of syngas (a mixture of H2 and CO), which is then transformed to hydrocarbons by Fischer-Tropsch synthesis. Syngas production is an energy-intensive process, accounting for over 60% of the total capital cost of gas to liquid plants. 45 The direct transformation of methane circumvents the low efficiency and high capital cost syngas production step, although it suffers from severe thermodynamics limitations at low temperature. 46,47 In 2014, Bao and co-workers reported the direct methane conversion to higher hydrocarbon products in oxygen-free conditions at high temperature. 48 Single-atom Fe embedded in the silica matrix (Fe/SiO2) exhibited 48.1% methane conversion, and 99% total hydrocarbon (such as ethylene and aromatics) selectivity at 1,090°C in a single pass experiment. No coke deposition was detected due to the absence of adjacent Fe species, which are considered necessary for oligomerization. In 2018, the nonoxidative conversion of methane to higher hydrocarbons was also reported on single-atom Pt/CeO2 catalysts at 900–1,000°C, 49 and the onset temperature of methane activation (less than 900°C) was a bit lower than that over Fe/SiO2 (less than 950°C). It was also demonstrated that, under mild conditions, single-atom Rh/ZrO2 was efficient for the conversion of methane to ethane by O2 in the gas phase, while only CO2 was formed over Rh nanoparticles. 50 The stabilization of CH3 intermediates over single-atom Rh was crucial for methanol and ethane formation, while the C–H bonds of adsorbed CH3 species were successively dissociated on Rh nanoparticles. Although the conversion of methane was limited, this work opens up a new way for the production of value-added fuels from methane under mild condition.Apart from making higher hydrocarbons, the conversion of methane in fuel cells using SACs has been demonstrated. Compared to the conventional combustion-based technologies, fuel cells are able to convert the chemical energy of the fuels to electricity with less pollution and higher efficiency. Although H2 is the most ideal energy source in fuel cells, the cheaper and more readily available hydrocarbon fuels such as methane are more attractive in the immediate future. The potential of using hydrocarbon fuels in the solid oxide fuel cells (SOFCs) has been investigated, either via the direct electrochemical oxidation of hydrocarbons or via formation of H2 and CO as the first step. 51,52 However, the extremely high operating temperature (800–1,000°C) hampered the practical application of hydrocarbon fuel-cells. 53 It is desirable to develop hydrocarbon fuel cells working at intermediate temperature.Along this line, Liu and co-workers reported a robust methane fuel cell, in which a catalyst with atomically dispersed Ru and Ni on CeO2 was coated on the anode. 54 The fuel cells enabled the direct electrochemical oxidation of methane containing 3.5% H2O at 500°C with no apparent coke formation after operation for 550 h. The authors attributed the outstanding activity and stability to the synergistic effect of single-atom Ni and Ru species in activating methane and H2O. As shown in Figures 3E and 3F, DFT simulations showed that single-atom Ni is responsible for the activation of C–H bond in methane. To form CO, an oxygen atom was removed from CeO2, thus creating an oxygen vacancy, together with which single-atom Ru species participated in the activation of H2O. Currently, there are very few publications available on single-atom catalyzed methane oxidation in the fuel cell, and much more work needs to be done in this area.Compared with traditional gasoline fuel, oxygenates have lower toxicity, reduced CO2 emissions, higher octane rating, and are more environmentally friendly and sustainable. 55,56 For example, the blending of 20% ethanol (one of the most common oxygenated fuels) in gasoline decreased the emission of CO, hydrocarbons, and NOx by 60%, 40%, and 20%, respectively. 57 It was also reported that every 10% addition of ethanol to gasoline could increase the octane number by 5 units. 58 According to the Renewable Fuels Association (RFA), the world ethanol production increased more than 100% from 13,123 million gallons in 2007 to 27,050 million gallons in 2017.Currently, ethanol is mostly produced via fermentation of sugars derived glucose. The production of ethanol from lignocellulosic biomass is receiving major research attention due to its easy availability, low cost, and minimal competition with food production. 59 Zhang and co-workers proposed a two-step transformations of lignocellulose to ethanol 60 : lignocellulose was firstly converted to methyl glycolate (MG), which then underwent hydrogenation to form ethanol on copper-based catalysts. To further improve the efficiency of the second step, a single-atom Pt@Cu alloy catalyst was employed. The single-atom Pt species improved Cu dispersion, promoted H2 activation, and minimized C–C bond cleavage, thus enhancing the activity and selectivity of ethanol. Li and co-workers developed a single-atom Ru/C3N4 catalyst, which showed temperature-dependent selectivity in the conversion of biomass-derived vanillin—a typical lignin model compound as a precursor for fuel additives. 61 The selective hydrodeoxygenation of vanillin to 2-methoxy-p-cresol is a widely studied model reaction for the much more complicated real biomass deoxygenation to fuel additives. Removal of oxygen is a critical issue for biomass upgrading, as the high oxygen content is the cause for low energy density, instability, corrosiveness, and high viscosity. Lower reaction temperature facilitated the hydrogenation of vanillin to form vanillyl alcohol, while higher temperature promoted the production of 2-methoxy-p-cresol via hydrodeoxygenation.The direct conversion of methane to oxygenated products using H2O2 as an oxidant in the aqueous phase was performed on SACs. Tao and co-workers designed a single-atom Pd catalyst with 2.0% CuO on ZSM-5, which showed a TOF of 2.78 s−1 with 86% selectivity toward methanol at 95°C. 62 Lee and co-workers reported five times recycling of single-atom Rh/ZrO2 catalyst in the oxidation of methane to methanol by H2O2 at 70°C without significant deactivation. 50 The graphene-confined single iron atoms were even active under ambient temperature for the direct conversion of methane to oxygenated fuels following a radical pathway. 63 Among various transition metals evaluated, only single-atom Fe was active due to the unique O–FeN4–O structure (Figure 4 A). To understand the origin of activity, the methane activation rate as a function of formation energy (Gf) of O-MN4-O (M = Cr, Mn, Co, Ni, Cu) was determined (Figure 4B). Compared with other O-MN4-O species, O-FeN4-O has a moderate free energy of formation (Gf) and shows the best ability to compromise all the energy barriers, and thus exhibited the best performance for methane activation. Methane was also converted to methanol and other oxygenates using molecular O2, despite of the lower activity. In 2017, Flytzani-Stephanopoulos and co-workers reported that single Rh atoms dispersed on TiO2 and ZSM-5 showed high performance in catalyzing the formation of acetic acid and methanol from methane using O2 and CO under mild conditions. 64 After 3 h of reaction at 150°C, 21,295 micromoles of acetic acid and 230 micromoles of methanol per gram of catalysts were produced on Rh/ZSM-5 and Rh/TiO2, respectively. Tao and co-workers also reported similar results that single-atom Rh anchored on the wall of microporous ZSM-5 transferred methane to methanol, formic acid, and acetic acid through the coupling of methane, O2, and CO. 65 SACs catalyzed CO2 hydrogenation into oxygenates such as methanol, 66–69 and formic acid 36,70,71 has been studied. The catalytic performance and reaction pathway of Pt single atoms Pt1@MIL and nanoparticle Ptn@MIL (MIL is a typical MOF which consists of μ3-oxo bridged Cr(III)trimers cross-linked by terephthalic acid) were compared during CO2 hydrogenation to methanol. 68 The TOF and selectivity for methanol over Pt1@MIL was 5.6 times and 6.8 times higher, respectively, than that over Ptn@MIL. A significant amount of byproduct CO was produced over Ptn@MIL (Figure 4C). The reaction mechanism showed that COOH* was the main intermediate on the Ptn@MIL catalyst, while on Pt1@MIL the key intermediate was identified as HCOO*. The distinct pathway over Pt1@MIL offered a lower energy barrier as well as the high selectivity for methanol production. The synergetic interaction between neighboring single metal atoms affects the reaction pathway as well as the activation energy in CO2 hydrogenation. 72 Individual single-atom Pt preferred the hydrogenation of CO2 into methanol, while both methanol and formic acid were formed over neighboring single Pt atoms. Zhang and co-workers designed a porous organic polymer (POP) with aminopyridine functionalities to anchor single-atom Ir for the conversion of CO2 to formate (Figure 4D). 71 The formed polymeric framework (denoted as AP-POP) with electron-donating aminopyridine functional groups was used as support to disperse single-atom Ir via wet impregnation followed by reduction. The fabricated single-atom Ir/AP-POP with an analogous structure to that of the homogeneous catalyst showed a turnover number (TON) of 25,135, representing one of the most active heterogeneous catalysts so far for formate synthesis from CO2 hydrogenation.Formic acid electrocatalytic oxidation as the anodic reaction in the fuel cell has attracted intensive research activities. The formic acid oxidation reaction follows two pathways: in the direct pathway formic acid is converted to CO2 (Equation 1); while in the indirect pathway CO is generated (Equation 2), leading to the deactivation of catalyst due to the strong affinity of CO to metal. 73 (Equation 1) HCOOH → CO2 + 2H+ + 2e− (Equation 2) HCOOH → COads + H2O → CO2 + 2H+ + 2e− In 2013, Lee and co-workers finely controlled the amount of Pt on Au nano-octahedra from single Pt atoms to Pt overlayers, and the performance of different Pt species in formic acid electrocatalytic oxidation was compared. 74 The single-atom Pt showed a mass activity of 62.6 A/mgPt, which was almost 10 times higher than that over Pt overlayers. The single-atom Pt preferred the reaction pathway towards direct oxidation due to the absence of Pt nanoparticles and the bifunctional effects of Pt−Au sites. In 2018, a series of bimetallic PtAu nanoparticles with various Pt loading from 4% to 96% for formic acid oxidation were reported. 75 A similar conclusion was reached that single-atom Pt showed a higher resistance to CO poisoning, and exhibited orders of magnitude improvement in the oxidation of formic acid compared with Pt nanoparticles. DFT analysis indicated that the CO adsorption on single-atom Pt is weaker as a result of the electronic effects induced by the Pt-Au binding interaction as well as the discrete Pt active sites. To meet the requirements of practical application of single metal atoms catalysts in formic acid fuel cell, the loading of Pt was increased to 8 wt % while maintaining the atomic dispersion. 76 Single Pt atoms anchored on antimony-doped tin oxide (Pt1/ATO) maintained superior formic acid oxidation activity to the conventional Pt/C catalyst even after 1,800 cycles.H2 is regarded as an ideal fuel for the future. The weight energy density of H2 is 122 kJ/g, 2.75 times higher than that of hydrocarbon fuels. 77 According to the “Hydrogen Generation-Global Market Outlook (2017–2026)” report, the global H2 production market is projected to reach $207.48 billion by 2026 at an annual growth rate of 8.1% from the starting point of $103.20 billion in 2017. It is predicted that 1 in 12 cars in South Korea, California, Japan, and Germany may be powered by hydrogen by 2030. 78 Hydrogen element is abundant in nature in the form of H2O, hydrocarbons, and biomass. However, a separate energy source such as electricity, light or heat is needed to extract H out of these sources in the form of H2. The main commercial H2 production relies on steam reforming, oil reforming, coal gasification, and water electrolysis, accounting for 50%, 30%, 18%, and 2% respectively. 79 The overall challenge of using H2 as fuel compounds comes from the ability to produce H2 efficiently at low cost. Besides, compression energy for H2 storage accounts for 10%–15% of the H2 energy content. Single-atom metal catalysts make efficient use of the noble metal atoms and have found applications in H2 production from methane or methanol reforming, water-gas shift reaction (WGSR), hydrogen evolution reaction (HER) and photocatalysis.Steam methane reforming is the most commonly used method to produce H2 in a large scale. One ton of H2 production forms 9–12 tons of CO2 in the process, 80  making it a significant contributor to CO2 emission on earth. 81 McFarland and co-workers designed a stable molten Ni-Bi metal alloy catalyst, the active sites of which were atomically dispersed for the conversion of methane to H2 and carbon without CO2 and other byproduct formation at 1,065°C (Figures 5A–5D). 82 The previously used solid catalyst suffered severe deactivation due to carbon deposition, while the molten metal alloy showed stable performance in seven days of continuous operation. The formed carbon floated to the surface of the molten metal alloy where it was skimmed off easily. The H2 production from methane at 1,100°C was also catalyzed by a single atom Fe@SiO2 catalyst. 48 The concentration of H2 in the effluent varied from 10.9% to 51.2% with the generation of value-added hydrocarbons (ethylene and aromatics) as by-products. The H2 fuel was generated from methanol steam or aqueous-phase reforming. Ma and co-workers reported that atomically dispersed Pt on α-MoC exhibited superior low-temperature H2 production activity as well as stability in aqueous phase methanol reforming, with an average activity of 18,046 mole of H2 per moles of Pt per hour. 83 Due to the strong interaction between Pt and α-MoC, electron-deficient single Pt atoms were highly dispersed on the support, facilitating the adsorption and activation of methanol. Besides, the α-MoC promoted the dissociation of H2O to form abundant surface-bound hydroxyls, which benefited the reforming of active intermediates at the interfaces between α-MoC and single-atom Pt.Water-gas shift reaction (WGSR), an important industrial reaction to produce H2 for energy application and to remove CO impurity in H2 fuel cell, has been intensively studied on single-atom Au, 8,84–86 Pt, 8,87–90 Ir, 91 and Pd catalysts. 92 Flytzani-Stephanopoulos and co-workers proposed that positively charged Au and Pt SACs were active in WGS, while Au or Pt metal nanoparticles did not contribute significantly to the reaction as the removal of the particles by cyanide did not affect the activity. 8,84 Later, they reported that the addition of alkali ions (Na, K) helps to stabilize mononuclear Au or Pt atoms on zeolite KLTL and MCM-41 for low-temperature WGSR. 85,90 Besides, the activation energy of single-atom Au in WGS was independent of supports, regardless of inert supports such KLTL and MCM-41 or reducible supports including TiO2, CeO2, and Fe2O3 (Figure 5 F). Zhang and co-workers also reported that single-atom Ir/FeOx catalyst showed more than 10 times activity of the Ir nanoparticles in WGSR. 91 The single-atom Ir improved the reducibility of FeOx and promoted the formation of oxygen vacancies, resulting in the excellent catalytic activity of single-atom Ir/FeOx catalyst.The hydrogen evolution reaction (HER), the cathodic half-reaction of water splitting, offers a reliable solution for the sustainable H2 production. HER occurs through the reduction of protons (Equation 3) in acidic electrolytes or the reduction of water (Equation 4) in alkaline electrolytes. 93 (Equation 3) 2H+ + 2e- → H2 (Equation 4) 2H2O + 2e- → H2 + 2OH- Under standard temperature and pressure conditions, the enthalpy change for the formation of H2 is 286 kJ/mol, which corresponds to a voltage of 1.23 V for the reversible electrolysis cell. 94 Under ideal conditions, an external potential of 1.23 V should be sufficient to drive the HER reaction. However, applying an overpotential is required to overcome the activation barriers and drive the electrochemical reaction. Efficient catalysts, mostly based on noble metals, can lower the overpotential. 95,96 To maximize the noble metal utilization efficiency, lower the catalyst cost, and improve the activity, selectivity as well as stability, the HER reaction has been carried out on single-atom noble metals, 97 such as Pt, 98–100 Pd, 101 Ru. 102,103 Besides, the HER reaction is also effectively promoted by non-noble metal SACs including Co 104 and Ni. 105,106 Among the noble metal SACs in H2 production from HER, single Pt atoms are the most widely studied. Single Pt atoms dispersed on nitrogen-doped graphene nanosheets (NGNs) by atomic layer deposition (ALD) technique exhibited as much as 37 times higher activity than the commercial Pt/C in HER, due to the partially unoccupied 5d states of single Pt atoms. 107 Wu and co-workers designed an ultra-low temperature (−60°C) ultraviolet photochemical method to prepare single-atom Pt and suppress the nucleation process of Pt atoms. 108 The prepared single-atom Pt catalyst exhibited lower overpotential (55 mV at 100 mA cm−2) and excellent stability in 5,000 cyclic voltammetry cycles. To understand the relationship between the coordination environment of single metal atoms and their catalytic performance, single-atom Pt catalysts with tunable coordination environment were dispersed on graphdiyne (GDY). 109 The four-coordinated Pt species (Pt-GDY1) were 3.3 times more active than five-coordinated Pt species (Pt-GDY2) in HER, due to the higher total unoccupied density of states of Pt 5d orbital and near to zero hydrogen adsorption Gibbs free energy on Pt-GDY2. Sun and co-workers reported that the electronic structure of single Pt atoms was modified by the coordination of nitrogen in aniline, showing superior HER performance and stability. 110 CO is often regarded as a poison ligand for Pt in heterogeneous catalysis, but Hyuck Choi and co-workers observed an unexpected improvement effect of CO on the performance of single-atom Pt in HER reaction. 111 The CO-ligation on the single-atom Pt promoted the dissociation of water to form Hads on Pt, thus enhancing the HER performance. Single Pt atoms were also used as co-catalysts to improve the HER performance. Two dimensional (2D) MoS2, as a potential alternative to Pt, has been studied in HER reaction. However, the performance of 2D-MoS2 needs to be improved as only the edge sites of the 2D-MoS2 contribute to the reaction while most sites at in-plane positions are inactive. Bao and co-workers doped single Pt atoms into the 2D MoS2 via the substitution of Mo sites to trigger the HER activity of MoS2. 112 The doped single-atom Pt tuned the H atoms adsorption behavior on the neighboring S atoms, leading to a significant improvement of HER activity on MoS2.Although noble metal SACs have shown excellent performance in H2 production from HER, it is not undesirable to replace noble metals with non-noble metals to make H2 a competitive energy carrier. The single Ni atoms dispersed on defective graphene showed a TOF of 0.3 s−1. 113 Although this value is almost one order of magnitude lower than that of the commercial Pt/C (2.30 s−1) at 50 mV overpotential, 114 platinum is four orders of magnitude more expensive than nickel. 115 The defective graphene offered a high density of anchoring sites through the efficient electron transfer between single-atom Ni and the 2π antibonding state of the adjacent carbon atoms. 113 Chen and co-workers dispersed single Ni atoms on nanoporous graphene for HER. 116 The unique sp-d orbital charge transfer between single atom Ni and the neighboring carbon atom resulted in a low overpotential around 50 mV. The dynamic structure of single-atom Co under alkaline HER condition was studied using operando X-ray absorption spectroscopy (Figure 5G). 117 The adsorption edge of the single-atom Co under open-circuit conditions was shifted towards higher energy side in comparison to ex situ samples; besides, a further shift was observed when the potentials of −0.04 and −0.1 V were applied (Figure 5H), indicating the oxidation state change of single-atom Co under working conditions. The operando EXAFS of single-atom Co showed different oscillation frequencies compared with the ex situ sample, and the intensity of Co–O/N peaks also changed when potentials were applied, suggesting a structural change of the single-atom Co under working conditions.The photocatalytic H2 production under light irradiation is considered as a type of artificial photosynthesis. 118 Generally, the photocatalytic H2 production involved the adsorption of light to generate electron-hole pairs, followed by charge separation and surface reaction. The overall photocatalytic performance is determined by both the thermodynamics and kinetics of the above steps. 119 Photocatalysts usually suffer from lower efficiency and selectivity towards H2 evolution under solar energy due to the high probability of charge-hole recombination events. The utilization of single metal atoms, as a new form of co-catalyst, can suppress electron-hole recombination, thus increasing the H2 photocatalytic production efficiency. 120–123 It was reported that single-atom Pt dispersed on C3N4 dramatically enhanced the photocatalytic H2 formation. 124,125 Ultrafast transient absorption spectroscopy indicated that the change of the intrinsic surface trap states in the support induced by the single Pt atoms contributed to the performance enhancement. 125 Due to the improved hydrogen binding energy, single-atom Pt confined into the metal-organic framework (MOF) exhibited a TOF 30 times greater than Pt nanoparticles. 126 Single-atom Pd/g-CN was reported to show a TOF of 417 h−1, which was much better than that of benchmark Pt/g-CN (76 h−1) for photocatalytic H2 evolution reaction. 101 Hyeon and co-workers reported the design of highly active hollow TiO2 photocatalyst with single-atom Cu anchored in the Ti vacancies. 127 The oxidation state change of single-atom Cu induced by the atomic localization of photogenerated electrons promoted the activation of neighboring TiO2, and improved the H2 production performance dramatically.H2 fuel cells convert H2 and O2 to electricity with H2O and heat as the by-products. In comparison to other energy converters such as internal combustion engines and power plants, H2 fuel cells are free of emissions besides H2O. Even if H2 is produced by existing technology from non-renewable natural gas, the overall pollutant emission will be decreased by 30% for cars and trucks driven by H2 in comparison to gasoline-powered counterparts. 128 Development of cost-effective and high-performance catalysts for the electrocatalytic oxygen reduction reaction (ORR) is key to realizing the large-scale application of H2 fuel cells. The ORR occurring at the cathode of electrochemical energy devices proceeds via either a two-electron (2e−) or four-electron (4e−) pathway. The four-electron pathway that reduces oxygen directly into the water is highly preferred for batteries because of the high energy-conversion efficiency. The catalysts are the “heart” of the H2 fuel cells, and noble metals, particularly Pt are the essential elements for the ORR catalysts. Increasing the utilization efficiency of noble metals by making them atomically dispersed, while not compromising the catalytic performance, have become a potential solution. 129–131 There is also considerable incentive to develop non-precious metal catalysts, such as Fe, 132–134 Co, 135,136 Mn, 137,138 Cu, 139 and Zn 140 to replace Pt-based ORR catalysts. 141 The single Au atoms dispersed on TiC were fabricated for the ORR. 114 The TOF of single-atom Au/TiC (1.57 s−1) was almost 3 times higher than that of Au nanoparticles supported on TiC (0.54 s−1) during ORR in acidic solution at 0.2 V. A quasi-Pt-allotrope ORR catalyst consisting of hollow Pt3Co nanosphere as the core and N-doped carbon with single-atom Pt as the shell exhibited stable 4e− ORR over 10,000 cycles. 142 Single Ru atoms dispersed on N-doped graphene by forming Ru-N4 moieties offered better resistance toward methanol, and CO poisoning than commercial Pt/C catalyst. 143 Among the earth-abundant transition metal SACs, metal-nitrogen-carbon (M = Fe or Co) based catalysts have been regarded as one of the most promising candidates in ORR. 136,144–147 Li and co-workers reported single-atom Co catalysts, in which cobalt atoms were anchored in hierarchically porous N-doped carbon 148 and hollow N-doped carbon spheres. 149 A half-wave potential of 0.892 V—53 mV more positive than that of commercial Pt/C—was obtained on the designed single Co atoms. The promotional effect was attributed to the synergistic contribution from both isolated Co atoms and the unique 3D hierarchical porous structure of the carbon support. 148 Lin and co-workers compared the ORR performance of hierarchically porous Co–N–C and Fe–N–C SACs. 150 The Fe–N–C offered a half-wave potential of 0.972 V, which was 49 mV higher than that on Co–N–C, as the single-atom Fe–N–C promoted the release of OH* intermediate, thus improving the ORR performance.Fe-based catalysts have attracted great interest for ORR. Single-atom Fe species were anchored on graphene hollow nanospheres using SiO2 as the template and Fe phthalocyanine as the precursor. 151 The rigid planar macrocycle structure of Fe precursor and the strong π–π interaction between Fe precursor and graphene oxide were beneficial for the dispersion of Fe. The atomically dispersed Fe species showed excellent activity, stability for ORR and tolerance toward methanol, NOx, and SO2 poisoning. The incorporation of S to single-atom Fe dispersed on nitrogen-doped carbon further improved the ORR activity due to the formation of thiophene-like structure (C–S–C) that decreased the electron localization of single-atom Fe. 152 The properties of the supports also have a great influence on the activity of the single-atom Fe catalysts. 153 Fe anchored on nitrogen-doped graphene with identical FeN4C12 moieties were prepared by the pyrolysis in Ar or NH3. The ORR performance over NH3-pyrolyzed catalyst was much higher than that over Ar-pyrolyzed one, due to the formed basic N-groups in the NH3 pyrolysis process. While some studies referred to the formation of Fe- pyrrolic-N structures as the origin of high performance of single-atom Fe catalyst confined in carbon supports, 154 another study suggested that the size of supports was also critical for the ORR activity. 155 In the range of 20 to 1000 nm, the best ORR activity was achieved at a particle size of 50 nm. Single-atom Fe bonded to graphdiyne through the formation of Fe–C also showed comparable activity as commercial Pt/C during ORR, in which single-atom Fe species promoted the reduction of oxygen directly into water while suppressing the formation of H2O2. 156 Batteries have been regarded as promising candidates for sustainable energy storage and conversion because of the high energy density and low cost. ORR is a relevant process for both fuel cells and batteries. The rate-determining step of SACs in ORR is complicated and still under debate. While O2 adsorption was identified as the rate-limiting step for single Zn atoms, 140 the reduction of adsorbed O2 to OOH*, 143 and the desorption of OH were proposed as the slowest step on single-atom Ru and Pt, respectively. 157 The SACs showing exceptional performance in fuel-cell might also have the potential to be applied in batteries. 158,159 For example, single Fe atoms dispersed on the hollow carbon polyhedron containing N, P, and S as dopants were tested in H2 fuel cell and Zn-air battery. 159 The designed catalysts delivered a superior current density of 400 mW/cm2 at 0.40 eV in the H2-air-fuel test, comparable to that of commercial Pt/C catalyst. Moreover, the single-atom Pt, as the air cathode for the Zn-air battery, showed negligible voltage change after 500 cycle tests with 200,000 s; whereas a significant voltage decreased was observed on commercial Pt/C catalyst. In this section, we touch upon recent progress made on the utilization of SACs in batteries.Single-atom Fe 160–164 and Co 165 have been integrated in Zn-air batteries in lab-scale and showed better stability than commercial Pt/C catalyst. Deng and co-workers reported that in comparison to nanoclusters and nanoparticles, single-atom Co showed the best activity, durability, and reversibility in Zn-air batteries. 166 Single-atom FeN4 species dispersed on open-mesoporous N-doped-carbon nanofibers were used as the electrode in Mg-air batteries. 167 The prepared electrode offered high open-circuit voltage, long operating life, and excellent flexibility, which showed a potential application in wearable and bio-adaptable Mg-air batteries. Single-atom Co embedded on N-doped graphene was applied as a cathode in Zn-air batteries. 168 The formed Co–N–C moieties promoted the formation of Li2S in discharge process as well as the decomposition in the charge process.The development of the Haber-Bosch process—the catalytic hydrogenation of N2 into ammonia—has enabled the strong population growth worldwide since the beginning of the twentieth century. For the first time, fertilizer could be produced synthetically on a large scale. For the last 100 years, the Haber-Bosch process has essentially not witnessed any major improvements and continues to constitute a major climate change driver consuming 1%–2% of global energy, predominantly because of the production of H2 as the reducing agent. Mainly due to slow reaction kinetics, high temperatures (400–500°C) and pressure (150–250 bar) must be employed, and the reaction mixture must be passed through catalyst beds multiple times to achieve favorable conversions. As the equilibrium conversion is higher at low temperatures but the activation of the strong N–N triple bond is challenging, the development of efficient catalysts could significantly impact the ecological footprint of fertilizer production. SACs based on bimetallic catalyst structures have been proposed for this daunting task, and the development could further improve our knowledge on the active site structure during reaction conditions. Beyond thermal catalysis, more recent advances in photo- and electrocatalysis with SACs could prove promising as ambient pressure and temperatures are sufficient to achieve reasonable reaction rates for the fixation of nitrogen gas. Due to its high gravimetric hydrogen content of 17.8%, ammonia is often regarded as a potential hydrogen storage compound. Ammonia can be compressed under much lower pressure (10 bar) and temperature (−33°C) compared to hydrogen (−240°C). Furthermore, the concentration of N2 in the air is 78.1% while that of CO2 is 0.04% by volume rendering nitrogen-containing hydrogen storage compounds more straight-forward. Both the decomposition of ammonia into carbon-free hydrogen for hydrogen fuel cells and the use of direct ammonia fuel cells are envisioned. 169,170 Especially, energy storage on a small scale based on the conversion of stranded energy resources into a chemical storage compound may rely on the electrocatalytic production of ammonia. 171 Bimetallic alloys where one metal is atomically dispersed in a solid ‘solution’ of another metal are popular catalysts. On the very small side of this approach are bimetallic single-cluster catalysts like Rh1/Co3 supported on cobalt oxide. Although this has been realized experimentally for the thermal reduction of NO to N2 and N2O, it was recently proposed based on DFT calculations that among other single-cluster catalysts, Rh1/Co3 would also be capable of reducing nitrogen into ammonia. 172 The capacity of the metal surrounding the atomically dispersed element to buffer charges and contribute to the catalytic reaction synergistically are believed to mainly contribute to the predicted catalytic performance (Figure 6 A).Different experimental studies have validated the use of SACs for the electrocatalytic nitrogen reduction reaction (NRR)—mostly based on Fe, Mo, and Ru. A nitrogen-doped carbon nanotube-supported Fe-based SAC synthesized by the pyrolysis of an iron-containing MOF reduced nitrogen to ammonia at −0.2 V versus RHE with Faradaic efficiencies of 9.28%, and a production rate of 34.83 μg h−1 mgcat. − 1 An iron species surrounded by 3 nitrogen atoms was proposed as the main active species and poisoning with thiocyanate salts showed that the nitrogen reduction activity was prohibited by the presence of Lewis bases. The mechanism was proposed to follow a distal pathway where dinitrogen binds in an end-on fashion on the Fe atoms, and the first ammonia production reaction occurs on the nitrogen atom more distant to Fe. 173 A Mo-based SAC supported on nitrogen-doped carbon material was synthesized in a similar pyrolysis procedure to the above-mentioned Fe SAC. Faradaic efficiencies of 14.6% ± 1.6% with an ammonia formation rate of 34.0 ± 3.6 μg h−1 mgcat −1 were achieved at −0.3 V versus RHE in 0.1 mol L−1 KOH solution. With around 10 wt % Mo loading, the catalyst does not exhibit a significant decrease in activity in 14 h and no formation of Mo clusters after the reaction was observed by HAADF-STEM. 22 Similar to the thermal catalytic reduction of nitrogen, Ru exhibits the best reaction rates achieved so far for the NRR using SACs. Using a MOF-pyrolysis procedure (Figure 6B), atomically dispersed Ru supported on nitrogen-doped carbon achieves Faradaic efficiencies of 29.6% at −0.2 V versus RHE with reaction rates of 120.9 μg h−1 mgcat −1 in 0.05 mol L−1 sulfuric acid. Again, the triple nitrogen-coordinated structure was proposed to be the active site. 174 The addition of ZrO2 to Ru SACs supported on nitrogen-doped carbon was found to be sufficient to suppress the competing hydrogen evolution reaction, and ammonia Faradaic efficiencies of 21% were achieved at −0.21 V versus RHE with maximum ammonia formation rates of 3.67 mg h−1 mgRu −1. A duration test over 60 h indicated high stability of the SAC under reaction conditions (Figure 6C). Plausibly, the presence of nitrogen in the carbon material and the addition of ZrO2 are essential to not only enhance the stability and activity for electrochemical reactions but also improve the Faradaic efficiencies for ammonia production. 175 Due to the importance of alternative ways to convert nitrogen gas into ammonia and the scarcity of experimentally reported catalysts, many DFT-based studies have been conducted to guide the rational catalyst design and screening. Similar to the best experimental systems, isolated Ru atoms supported on different nanoporous carbon materials have been predicted to be stable and active for the NRR, although the competing HER increases the necessary overpotential. 176 A systematic study of different transition metals on nitrogen-doped carbon employing three defined properties including stability, the competitive adsorption of dinitrogen against dihydrogen molecules, and the competition of the first dinitrogen protonation against hydrogen adsorption on metal sites. Based on this analysis, Co- and Cr-containing SACs are predicted to yield the highest activity and selectivity for ammonia production at low overpotentials. 177 Besides carbon-based materials, boron has been proposed as a powerful support and even active site for NRR. Upon binding of dinitrogen to sp3-hybridized boron atoms, the B-to-N π-back bonding populates N–N π* orbitals and thus activates the notoriously strong N–N bond (Figure 6D). Depending on the support for isolated boron atoms, the NRR activity can be improved while the HER activity can be suppressed (Figure 6E). Mo SACs supported on defective boron nitride with a boron monovacancy were calculated to surpass equivalent noble metal-based catalysts which were assigned to the unique ability of Mo to stabilize N2H* and destabilize NH2* species (Figure. 6F). 178 One of the challenges of utmost importance is the suppression of the HER under reaction conditions relevant for NRR which has been addressed recently by the computational comparison of 120 transition metal SACs supported on different nitrogen and carbon-containing scaffolds. The authors found that Ti and V have the strongest ability to activate dinitrogen as well as the lowest free energy barriers for the NRR while exhibiting little predicted HER activity. 179 The direct reduction of nitrogen into ammonia using sunlight as the sole energy source would be favorable but the development of efficient catalysts is a major challenge. Atomically dispersed copper on carbon nitride were shown to generate ammonia under the illumination of visible light (420 nm) with quantum efficiencies of around 1% and reaction rates of 186 μg h−1 gcat −1 around 7 times higher than pure carbon nitride. 65 Doping isolated low-valent Mo atoms into W18O49 nanowires was sufficient to enhance the catalytic activity by around 7 times compared to the un-doped material and can achieve ammonia formation rates of 195.5 μg h−1 gcat −1 with an apparent quantum efficiency of 0.028% under simulated AM 1.5 light irradiation. The interface between W and Mo was calculated to be the active site, and the reason for the enhanced catalytic performance (Figure 6G). 180 Similar to the NRR, boron atoms have been predicted to efficiently convert dinitrogen into ammonia on a semiconductor material such as carbon nitride. Besides the activation of dinitrogen molecules, boron atoms can enhance the visible light absorption of carbon nitride and thus are expected to improve the photocatalytic nitrogen reduction. 181 Beyond the above-mentioned categories, the production of commodity and fine chemicals is closely connected to energy-consumption, providing access to agrochemicals, pharmaceuticals, polymers, fragrances, food additives, adhesives, lubricants, among others. According to the process intensification workshop held by U.S. Department of Energy in 2015, the overall US manufacturing sector in 2010 reached 19.24 quadrillion British thermal units (quads), where the chemical production processed consumed 1.15 quads. 182 Developing more efficient chemical process will reduce the energy consumption of the chemical sector and greenhouse gas emission. It becomes increasingly difficult to achieve full conversion while maintaining high selectivity for more complex chemicals, and therefore, laborious post-treatment becomes inevitable. Based on estimates by the Oak Ridge National Laboratory, separation processes account for around 15% of the total annual US energy consumption and for approximately 40%–50% of the total energy consumption in chemical processes. 183,184 More challenging separations such as in the pharmaceutical industries caused by very rigorous purity requirements and complex separation tasks such as the resolution of enantiomers would increase the energy consumption more significantly. Improving the selectivity of chemical reactions as well as replacing homogenous catalysts with suitable recyclable catalysts are thus imperative if the energy for separation is to be decreased. Besides offering the opportunity to conduct chemical reactions under milder and thus less energy-intensive conditions, SACs are also able to improve the reaction selectivity. For selective hydrogenation reactions where isomeric products or mixtures of alkynes, alkenes, and alkanes are particularly difficult to separate, SACs have proven to be excellent selective catalysts surpassing their nanoparticle counterparts. On the bridge between homogeneous and heterogeneous catalysis, SACs have been shown to combine activity and selectivity for certain coupling and hydrofunctionalization reactions well beyond other heterogeneous catalysts while they are easily removed from the reaction solution by filtration. Besides, the catalysts can be used continuously when fixed bed reactor is applied.Due to the absence of adjacent metal atoms, the activation of hydrogen and the subsequent hydrogenation reaction will occur much more selectively. One such example is the selective hydrogenation of acetylene – a major impurity hampering the ethylene polymerization reaction – to ethylene without promoting the complete hydrogenation to ethane. Similarly, the hydrogenation of butadiene, which is a strong poison for alkene polymerization catalysts, into the butene isomers requires the development of highly selective catalysts. Both positively charged SACs, as well as SAAs, have been used based on Pd, 185–190 Pt, 191–194 and Au 9,10,195 all of which showed selectivities for the semi-hydrogenation products far exceeding nanoparticle-based catalysts. Several other selective hydrogenation reactions have also been achieved, such as the chemoselective conversion of nitroaromatics to amines 159,191,196,197 and azo compounds 198,199 or the semi-hydrogenation of quinoline. 200 Compared to nanoparticle-catalysts, several SACs have been proven to be CO-tolerant hydrogenation catalysts probably due to the weak adsorption of CO on positively charged noble metals, 201,202 allowing the direct use of industrial-grade hydrogen gas as feedstock. Of note, the CO adsorption strength on single-atom Pt is still under debate. While some reports suggest CO adsorption on Pt1 is much weaker than that on Pt nanoparticles, 203 other studies provide evidence for the strong adsorption of CO on single-atom Pt. 204,205 Additionally, a Pt1/α-MoC catalyst offers high activity in the water-gas shift reaction so that water can be used as hydrogen source (Figure 7 A). 201 SACs have been shown to show great promise for several hydrofunctionalizations reactions such as hydroformylation, hydrosilylation, and hydrochlorination reactions—all industrially relevant reactions where nanoparticle-based heterogeneous catalysts are inferior compared to homogeneous catalysts. Rh SACs supported on ZnO 206 or CoO 207 show high activity and simultaneously high selectivity of up to 95% toward a certain isomeric aldehyde in stark contrast to Rh clusters of higher nuclearity and most homogenous catalysts (Figure 7B). 207 The authors ascribe this enhanced reactivity to the dynamics of Rh atoms on the CoO support or the charge transfer from Zn to Rh on ZnO yielding almost metallic atomically dispersed Rh. Another application for atomically dispersed positively charged Pt atoms is the alkene hydrosilylation reaction, the arguably most important industrial application for homogenous Pt catalysts. Several Pt SACs and an SAA have been demonstrated for the hydrosilylation of different alkenes. 208–211 The high activity is normally attributed to either the high valence and thus the facile insertion of Pt into the C–H bond 210 or the charge transfer of Pd to Au in dilute Pd-Au alloys. 211 Recycling studies revealed that the SACs could be used up to 5 times without significant loss of activity with TONs of up to 105. For the production of polyvinylchloride, the production of its monomer—vinylchloride—by the hydrochlorination reaction of acetylene is inevitable. The conventional heterogeneous industrial catalyst is based on toxic Hg, but recently single-site gold catalysts have been identified as a viable alternative. The reaction mechanism on carbon-supported Au has been experimentally proven to be based on an Au(I)-Au(III) redox cycle (Figure 7C). 212,213 In contrast, CeO2-supported Au catalysts follow an Au(0)-Au(I) redox cycle because of electronic coupling with a Ce(IV)/Ce(III) cycle. The authors also show that catalytically inactive Au nanoparticles can decompose into isolated Au atoms when exposed to a C2H2/HCl mixture under reaction conditions. 214 Coupling reactions belong to the most important reactions in the synthesis of complex chemicals such as those in the pharmaceutical industries. Traditionally, heterogenous nanoparticle catalysts are neither particularly active nor selective, and thus the sector mostly relies on the use of homogeneous catalysts which are inherently difficult and energetically expensive to recycle. Recently, different SACs have been shown to be active in the Ullmann, 215 Sonogashira, 216,217 Heck, 216 and Suzuki 216,218,219 couplings. Besides the positive charge of noble metals in SACs resembling homogeneous metal complexes, the mobility of metal ions in supports such as carbon nitride was reported to be the reason for the catalytic activity sometimes surpassing homogeneous complexes. Recycling and flow reactor stability studies reveal that the SACs sustained coupling reactions over a long-time period in stark contrast to the stability of homogeneous complexes (Figure 7D). 219 Light olefins belong to the most crucial building blocks in the chemical industries. The recent exploitation of shale gas deposits spurred interest in the dehydrogenation reaction of light paraffins such as propane. Harsh reaction conditions resulting in catalyst stability issues as well as the formation of coke and other side products plague the development of suitable SACs. Pt SACs on CeO2 are stable under propane dehydrogenation reaction conditions, but the selectivity towards propylene was negligible. This was assigned to the facile C–C bond cleavage on Pt1 sites. 220 More recent calculations, however, indicated that SAAs with Pt diluted in more abundant metals such as Cu seem to combine both the excellent C–H activation capabilities of the noble metal and the low first dehydrogenation reaction barrier but prevent the further dehydrogenation of propylene to undesired side products. In fact, Pt/Cu SAAs are capable of breaking the scaling relationships between the propane dehydrogenation activity and selectivity commonly observed for single metal and alloy catalysts. 221 Similar turnover frequency (TOF) values of 0.72 s−1 for Pt nanoparticles and 0.56 s−1 for Pt/C SAAs at 520°C under otherwise identical reaction conditions were observed. Of note, the propylene selectivity was around 3.2 times higher for the SAA (90%) indicating a significantly better performance of SAA catalysts. 221 It was predicted based on DFT calculations that Pd/Cu SAAs would also exhibit favorable performance in the dehydrogenation of propane. 222 As shown in the sections above and summarized in Table 1 , SACs have been reported to show superior catalytic activity to their nanoparticle counterparts in a wide range of catalytic applications, including hydroformylation to selective hydrogenation, 185–194 dehydrogenation, 221,222 water-gas shift reaction, 8,84,91 and hydrogen evolution reaction. 98,107,108 It is not unreasonable to propose, as a rule of thumb, that SACs may be superior to NPs in the reactions that are conventionally more successful using metal complexes as catalysts. Likewise, the design of SACs should learn from the wisdom in homogeneous catalysis to fine-tune the frontier orbital geometry and energy of the active sites.Meanwhile, SACs are not as active as nanoparticles in some other reactions. SACs might even be completely inactive in case that two or more neighboring metal atoms are required to activate a reactant. It is well-known that the electrooxidation of methanol in fuel cell dominantly involves three or four Pt atoms to accommodate the formed CHxO intermediate. 20 The single-atom Pt dispersed on thiolated multiwalled carbon nanotubes (S-MWNTs) was almost inactive, while Pt nanoparticles were favorable for the methanol oxidation. 223 Similarly, Pt nanoclusters Pt4 and Pt10 anchored on indium tin oxide (ITO) showed excellent ethanol oxidation performance, while single-atom Pt1/ITO was much less efficient. 224 Similarly, Pd ensembles rather than single-atom were proposed to be responsible for the ethanol oxidation reaction. 225 SACs could also be inferior to their nanoparticle counterparts in terms of reaction selectivity. For example, Pt SACs on CeO2, when employed in propane dehydrogenation, exhibited negligible selectivity towards propylene due to facile C–C bond cleavage. 220 Although SACs have received intensive research activities, deep understanding of the working mechanism of SACs is still under development. The debate remains regarding whether SACs are active or not in certain reactions, and if yes, whether they are more active than nanoparticle counterparts. Examples include CO oxidation, methane activation, and N2 hydrogenation. In the case of methane oxidation, Lee and co-workers argued that Rh single-atoms promoted the conversion of CH4 to methanol using O2 in the gas phase or H2O2 in the aqueous solution, 50 while single Pd atoms were reported to be inactive for the same reaction. 226 For N2 hydrogenation, both experiment, and DFT simulation indicated that N2 dissociation on Ru(0001) was dominantly determined by the step sites, 227 whereas N2 reduction to ammonium was predicted to be feasible on single-atom Ru where step sites were not available. 172,174,175 Single-atom catalysis emerges from the in-depth study of supported metal nanoparticle catalysts that already found wide industrial applications in oil refining, coal transformation, fertilizer production, and many more. Thanks to the technological advances in the spatial and temporal resolution of analytical tools, within merely a few years’ time hundreds of reports generated in labs around the world authenticated the existence of the isolated single-atom species on various supports and their active participation in catalytic reactions. This does not only fundamentally change the way we view the structure and function of metal-based catalysts, but also provides grand opportunities for a more efficient usage of fossil resources, less energy-intensive processes for chemicals production, more effective energy storage and the novel transformations of alternative energy sources.The development of efficient, selective, and stable catalysts with low cost is crucial for energy-related applications. Due to the maximized atom utilization efficiency and unparalleled electronic and geometric features, SACs have exhibited exciting technological, and fundamental significance in nearly every field of energy transformation and storage. In this review, the recent advances of SACs in the transformation of hydrocarbons, oxygenates, H2 fuel, batteries, ammonia, and fine chemicals have been summarized. Particular attention was paid to structure-performance relationship and the advantages of SACs in comparison to traditional nanoparticle or commercial catalysts in energy-related catalytic reactions. The prospect of using SACs in energy application looks promising, and enormous advances have been achieved to date. However, future research should be devoted to the following aspects to foster further growth of the area, and potentially push the SACs for practical energy application.Noble metal-based single atoms catalysts account for about two-thirds of published articles in the past five years. 228 Nevertheless, the non-precious SACs exhibiting comparable activity as noble-atom SACs are more attractive, given that performance and cost of catalysts are two important factors affecting the energy conversion. As discussed in the review, 3-d metal-based SACs exhibited comparable or even superior performances to noble-metal catalysts in several photocatalytic and electrocatalytic reactions. Future work should be directed to nitride or carbide supports combined with non-noble single metal atoms, which may offer unique electronic interactions with the metals generating improved performance.Considering that many energy-related applications require harsh operation condition, the development of industrial-scale manufacturing methods that offer stable and high metal loading SACs at affordable cost is essential. Although several strategies have been reported along this line, these methods rely on strong anchoring sites on particular supports and therefore to a certain extent suffer from a lack of general applicability. Universal stabilization strategies for the synthesis of a wide range of SACs are pressingly needed. A possible approach is to learn from strategies to make stable colloidal nanoparticles, 228 such as electrostatic interaction and steric hindrance, which have been well studied and even quantitatively described in the past decades. On another note, despite that more than 80% of all the heterogeneous catalysts are fabricated by wet-chemical impregnation or precipitation, 229 it may not be ideal for every type of SACs synthesis since low loadings are necessary to keep the metals atomically dispersed. Very recently, a facile shockwave method was developed to synthesize thermally highly stable SACs. 230 Solid-state syntheses such as this one and other less conventional methods may find unique advantages in making SACs in the future.At present, the identification of single atoms is mainly achieved by the combined use of HAADF-STEM, CO-DRIFT-IR adsorption, and XAFS. While these techniques generate a clear picture of the structure of dominant metal species in a catalyst, none provides accurate electronic structure and coordination environment of the single metal atoms with spatial resolution under the working state. Therefore, the current understanding of the structures of active sites in complex heterogeneous SACs and their working mechanism in catalytic reactions are largely based on postulations derived from statistically averaged properties. A potential solution to the problem is the development of single-atom electron spectroscopy, which would enable structural identification of individual metal species under a microscope. This emerging technique has been successfully applied in the revealing localized electronic structure of single atoms, 231 but its usefulness to help understand the catalytic function of isolated metal atoms remains to be explored.Another challenging and critical task in SACs is to develop a technique that does not only microscopically or spectroscopically image various metal species, but also differentiates which ones are active in catalysis and which ones are not. Often, various sized metal species coexist in a working heterogeneous catalyst. The contributions of all these species in catalysis are hard to disentangle. Even for single-atom species, their structure and catalytic property are likely to be non-identical. Considering most catalytic reactions are associated with heat effect, active sites will induce a significant change of the local temperature. As such, we envisage sub-nanometer resolution thermometry combined with atomic resolution electron microscopy would offer a powerful tool to contrast metal species that are more active from the ones that are less active or completely inert. Recently, plasmons 232 and phonons 233 have been used to probe the temperature of nano-objects in the electron microscope. Leveraging on these advances, a thermometry-microscopy system for the above-mentioned application may become a reality.Although SACs have been intensively studied in various energy transformations, more research work should be devoted to expanding the application of SACs in even broader areas. We propose several reactions where SACs deserve further exploration: (1) Fischer-Tropsch synthesis. In industry, cobalt and iron nanoparticles are widely used. Metallic single-atom cobalt, iron or ruthenium alloys might offer unique selectivity in Fischer-Tropsch reaction; (2) Hydrocracking of heavy oil. Heavy oil hydrocracking is currently promoted by Pt nanoparticles supported on zeolites. A few single-atom Pt alloy catalysts have been successfully prepared in the literature. 60,234 It would be interesting to test the performance of these catalysts in hydrocracking despite their stability potentially representing an issue. (3) C-H activation. While methane activation has been realized by SACs, C–H activation of larger molecules such as benzene derivatives has rarely been reported. Considering that the Palladium complex is widely used in the C–H activation, 235 it deserves more effort to expand the application of SACs in C–H functionalization of more complicated substrates. (4) N≡N activation. Several DFT simulations for the hydrogenation of N2 to NH3 have been performed on SACs. It was predicted that SACs are promising for the conversion of N2 to NH3. However, experimental validation of these reports is rare at the moment. Provided low temperature, pressure ammonia synthesis become viable, one could envisage decentralized facilities for NH3 production and point distribution.Along with the catalytic application of SACs getting increasingly broad, standardized operation protocols should be established for various reactions using SACs. When preparing this review, we realized that the performance of catalysts in most cases is measured under various conditions, making the comparison of catalytic behavior of SACs challenging. A good practice is surfacing. For instance, McCrory and co-workers developed a benchmarking protocol, using the potential increase after 2 h of galvanostatic polarization at 10 mA/cm2 per geometric surface area to test the stability of OER catalysts. 236,237 More such efforts should be spent for the rational comparison of SACs in a broad range of energy-related applications.The past few decades have witnessed a growing synergy between theoretical simulation and experimental investigations in catalysis. Traditionally, the DFT calculations were carried out within the concept of potential energy surface, in which a simplified model under idealized conditions (ultra-high vacuum and −273°C) was considered. 238 The fast development of hardware and software makes it possible to simulate the catalytic reactions under realistic conditions. A deeper understanding of the reaction mechanism and structure-performance relationship under realistic conditions will benefit the rational design of single-atom catalysts for the specific energy transformation process.The development of data science has enabled the big data strategies to discover the underlying correlations and making predictions. Machine learning for data analysis is spreading rapidly in catalysis, 238 and it is mainly focused on two aspects in heterogeneous catalysis: (1) the direct prediction of catalytic performance and (2) developing a model to estimate the reaction rate indirectly. Very recently, single-atom transition metals anchored on graphdiyne with outstanding electron transfer ability were identified using a deep-learning algorithm and big-data technique. 239 We anticipate growing employment of machine leaning in guiding the design of particular SACs for energy transformation. Ideally, it is combined with fast synthetic platforms and high throughput performance screening techniques that have already been commercialized.While hundreds of papers are available for SACs, only a handful of cases have been reported for dinuclear and multi-nuclear species without organic ligands as active sites. 240–242 There is a clear gap between SACs and well-studied nanoparticle catalysts. An atom-by-atom approach to synthesize active sites ranging from single-atoms to atomically precise metal clusters on the same support is highly desirable. In this regard, the concept of single-atom catalyst has been recently extended to single-cluster catalyst (SCC), 172 i.e., each catalyst bears only one type of Mx (x ≥ 1) species with a specific number of x. In this manner, the nuclearity effect in heterogeneous catalysis could be systematically studied, understood, and rationalized. The well-known B5 sites of Ru(0001) and the more recently proposed Fe3 sites on θ-Al2O3(010) 243 for ammonia synthesis could both be considered as multi-nuclear metal sites. We expect research along this line will provide important new discovery in structure-activity correlations, which will ultimately benefit the identification of the best catalyst in each energy application.We thank the National University of Singapore Flagship Green Energy Program (R-279-000-553-646 and R-279-000-553-731) for financial support.N.Y. and J.P.-R. conceived and supervised the preparation of the review. S.D., M.J.H., and N.Y. collected references and wrote the manuscript. N.Y. and J.P.-R. revised and finalized the manuscript. All authors approved the final version of the manuscript.

Almost a quarter of the energy consumed globally is directly or indirectly related to the use of a catalytic process. Conventional nanoparticle-based catalysts recently witnessed the dawn of its potential successor—heterogeneous single-atom catalysts (SACs), which allow the maximum possible dispersion of metal atoms on the catalyst surface, possess unparalleled electronic structure and geometric configuration, and exhibit, otherwise, exceptional performance in a range of energy-related applications. Herein, we critically review the use of heterogeneous SACs in the generation and conversion of hydrocarbons, oxygenates, H2 fuel, ammonia, commodity and fine chemicals, and the electrochemical energy storage in and release from batteries. We describe the importance of those energy-related compounds in the current energy infrastructure and discuss how catalysis—in particular, single-atom catalysis—can be used more effectively in each application. At last, general guidelines and trends guiding the future design of stable and efficient single-atom catalysts for sustainable energy transformations are provided.

The constant growth in global energy demand has entailed a massive consumption and excessive depletion of fossil fuels, leading to an energy crisis as well as a health crisis due to greenhouse gas emissions, responsible for climate change and global temperature increase. The atmospheric CO2 level (one of the main greenhouse gases contributors) has also increased, reaching a value of 416.47 ppm (May 2020) [1] and is forecast to reach up to 570 ppm by 2100 [2], while the safety limit is estimated at 350 ppm [3]. As a result, there is an urgent need to control and mitigate these emissions by pursuing low-carbon alternatives including CO2 capture, sequestration, and utilization [1]. Indeed, research efforts are focused on the use of CO2 as carbon pool, given its potential to become genuine feedstock to produce value-added products hence turning a problem into a virtue.Nonetheless, conversion of CO2 is an arduous task given its high thermodynamic stability due to the two strong equivalent CO linear bonds; bonds that possess a much higher bonding energy (750 kJ mol−1) than that of other carbon bonds (C–H, 411 kJ mol−1; C–C, 336 kJ mol−1; C–O, 327 kJ mol−1) [4]. Therefore, a suitable CO2 transformation route must overcome these energy barriers requiring high energy input, preferably coming from carbon-neutral sources, as well as the use of an active catalyst or high pressures and temperatures [5–7]. Among all the possible techniques to convert CO2, including thermochemical, electrochemical, photochemical and biological processes [1,7–9], electrochemical reduction of CO2 has attained a growing interest due to its multiple possible uses in the energy sector and chemical industries, producing value-added fuels and chemicals at mild conditions and in a carbon-neutral way [10–12].Electrochemical reduction of CO2 allows tuning the selectivity of the value-added products obtained [13]. Traditionally, research on electrocatalytic reduction of CO2 has mainly focused on the formation of C1 products (such as CO, formate and methanol) since these simple 2e− transfer reactions are kinetically more favourable [14,15]. However, the production of C2+ species would be rather interesting from the application perspective [16]. C2+ alcohols contain higher energy densities, lower toxicity and corrosiveness compared to methanol, being more suitable as blending or even pure fuels in existing internal combustion engines; and short-chain alkanes can be directly injected into gas-distribution grids enhancing the calorific value of natural gas or biogas. They can also be regarded as entry platform chemicals for current value chains, e.g. light olefins for the production of polymers [9,17]. Moreover, and taking acetate as an example, its direct production from the electrochemical reduction of CO2 is a more efficient and effective process in terms of energy and steps compared to the traditional industrial multistep process from fossil fuels, besides more environmentally friendly from a CO2 emissions point of view [14].However, C2+ products reaction pathways are complex and strongly influenced by the catalyst surface, electrode materials, reaction medium (buffer strength and electrolyte solution), design of electrochemical cell or working conditions such as temperature, pH or pressure and concentration of CO2 [18,19]. The high C–C coupling activation energy required joint to its bond formation competition with C–H and C–O bond formations limit the efficiency towards C2+ products. The later along with the overpotential gap between the essential CO intermediates formation and that of C2+ species are the main impediments in the practical application of CO2 reduction to C2+ species in commercial electrolysers. Indeed, these aspects are regarded as the major bottlenecks accounting for the low energy efficiency of C2+ products compared to C1 counterparts [19,20]. Still the commercial appetite and their extraordinary added value makes necessary to strengthen the research efforts towards direct CO2 to C2+ products.In order to achieve the practical viability and implementation of this technology, higher efficiency and energy requirements as well as lower operational costs must be met. Electricity plays a key role in the profitability of the CO2 reduction reaction (CO2RR), being the main factor in the operational costs [21,22]. Nonetheless, one of the advantages of the electrochemical reduction of CO2 is that it can be approached as an efficient way to store all excess electrical green energy (generated from unpredictable and intermittent sources such as wind and solar) as transportable chemicals and fuels [23]. Regarding the efficiency requirements, commercial electrolysers require current densities above 200 mA cm−2 as well as long-term durability catalysts [21,24,25]. For these reasons, research on the development of efficient catalysts is encouraged.Recently, several efforts have been made to obtain suitable electrocatalysts, able to improve catalytic activity and selectivity by controlling their chemical states, size, morphology, surface defects, crystal facets, porosity or by creating heterostructures [6,20,26–28]. Generally, catalysts explored for the electrochemical reduction of CO2 are based on metals. While formic acid is the main product using Sn, Pb, In, Hg or Bi as catalysts; Zn, Ag, Au and Pd have been found to be selective towards CO [10,23,29]. In the case of C2+ products, Cu has demonstrated a unique ability to facilitate C–C coupling, although it is not a selective catalyst since numerous C1–C3 hydrocarbon and oxygenate products have been observed on Cu surface [19,30].In this scenario and given the research gaps and motivation for this appealing CO2 conversion route, herein we analyse the recent advances and efforts in the electrochemical CO2 reduction to C2+ products. Beyond summarising the fundamental aspects for the development of a suitable catalyst as well as the mechanistic pathways of the most industrially desired C2+ species, this review makes emphasis on key aspects to design highly selective electrocatalysts. Additionally, as a very important aspect not frequently addressed in electrocatalysis literature, we bring techno-economical requirements into discussion targeting potential industrial implementation.Electrocatalytic CO2 reduction is especially appealing due to the wide variety of important products derived from it and their multiple advantages. The process can be carried out in neutral pH, at room temperature and atmospheric pressure and it can be controlled by the reaction temperature and the electrode potential. The chemical consumption can be minimised since electrolytes can be completely recycled and the intermittent renewable energy can be converted into stable chemical energy. Moreover, the system is modular, compact and easy to scale-up [8,9,13,31]. However, despite all the advantages, some drawbacks must be overcome for the technology to fully take off at commercial level. The activation of the very stable CO2 molecule is the first step in its electrochemical reduction and requires high overpotential. In fact, the formation via single-electron transfer of CO2*- radical intermediate (E = −1.9 V vs. SHE) is considered as the rate-determining step on most transition metal-based catalysts [3,32]. In general, heterogeneous electrocatalytic reactions take place at the catalyst surface-electrolyte interface [33], so the main CO2 electroreduction process could be simplified as follows: CO2 chemisorption and bond forming interaction on the catalyst surface, electron and/or proton transfers leading to C–H, C–O bonds or C–C coupling; and product species rearrangement and desorption from the catalyst surface into the electrolyte [3,20,34], While formate and CO generation from C–O and C–H has been further studied and can be achieved upon implementing the right electrocatalyst at low overpotential [32], hydrocarbon or alcohol production is a more complex task. A summary of the most accepted activation routes for the electrochemical CO2 reduction to C2+ products is given in Fig. 1 [12,19,20,32].Due to the multiple reaction pathways that can be conducted in parallel and competitive ways as well as more complex pathways that are not elucidated yet, a wide-ranging product distribution is generally obtained. The reaction pathways and, consequently, product distributions strongly depend on the electrocatalyst. Noble metals (Ag, Au and Pd) have been found as high selective catalysts for C1 products like CO and formic acid, while activity towards more than 2e− products such as methanol and methane has been obtained with Cu-based catalysts [12]. Besides the choice of a selective electrocatalyst, product distribution can be also adjusted by tuning external parameters such as the electrode characteristics (surface, morphology, facets, …), operation conditions (applied overpotential, electrolyte, anodic reaction) or the cell design itself.In general, electrolytes should provide stable pH at bulk, good ionic conductivity and moderate to high CO2 solubility [22]. The type, concentration and composition of electrolytes have been found to affect CO2 electrochemical reduction. One of the most common and favourable electrolytes is alkaline aqueous solution because it stands lower overpotential than its neutral counterpart [22,35], helps to supress the hydrogen evolution reaction (Eq. (1)) and provides a high conductivity which reduces ohmic losses [35]. Hydrogen evolution reaction (HER) (equation (1)) is the main competing reaction to CO2 reduction. Due to the proton insufficiency on the catalyst surface, a basic media helps to supress this undesired reaction boosting CO2 electrochemical reduction products [35,36]. (1) 2 H + + 2 e − → H 2 In the study of Verma et al. [37] it was observed with different alkaline aqueous solutions higher current densities as increasing the solutions concentrations and through electrochemical impedance spectroscopies demonstrated the decrease of the cell resistance due to the ionic conductivity increase while increased the concentrations. Moreover, as Gabardo et al. reported [35], a 240-mV positive shift of the onset potential was achieved while using a 10 M KOH solution as electrolyte instead of a 1 M KOH one.The CO2 solubility can be increased using organic solvents instead of water solutions as electrolytes. Although CO2 is a nonpolar molecule, it possesses an appreciable polarizability and an ability to accept hydrogen bonds from suitable donor solvent [38]. Most of the organic electrolytes are polar solvents and allow the electrochemical CO2 reduction in a wider potential window [20]. The use of an aprotic solvent (such as acetonitrile or dimethylformamide) enables the *CO–CO dimerization while CH4 production is favoured in a protic solvent. Nonetheless, electrochemical CO2 reduction can be tuned by adding protic compounds to aprotic solvents in order to facilitate the generation of H-containing products [20]. Some research on the use of ionic liquids (added to aqueous or organic solvents) has been made looking for a higher conductivity and hence, a decrease in overpotential [39–41]. Ionic liquids (ILs) could be proper electrolytes since they have good thermal stability, high CO2 solubility, wide electrochemical window, low vapor pressure and partially HER inhibition [20,42]. However, ILs also present some drawbacks concerning their high cost, the environmental toxicity embedded in their production, liquid products extraction struggling [43] or cathodic corrosion [44]. The latest works where ILs has been used as electrolyte are summarised in Table 1 . Cathodic catalysts can suffer from deactivation due to the presence of trace impurities (metal ions or organics) coming not only from noble metal anodic catalysts dissolved because of operating conditions [45], but from the electrolyte, being the main deactivation cause in Cu, Ag and Au catalysts [46]. However, deactivation due to impurities can be avoided if a pre-electrolysis with a sacrificial electrode or an irreversible coordination between metal ions and a chelating agent (e.g. ethylenediaminetetraacetic acid) are conducted [46].As depicted in Fig. 1, the formation of C2+ products through different reaction pathways depends on the protonation process. The protonation transfer can affect the product distribution in electrochemical CO2 reduction, therefore, pH near the catalyst electrode surface, bulk pH, and local pH were found to affect the final product distribution [6].Due to the consumption of protons near the electrode surface as a result of proton and water reduction, a local pH (usually more alkaline than bulk pH) can be generated [32]. This difference in pH is caused by mass transport limitations [61] and depends on the operation conditions: current density, bulk pH, types of cations or anions, and on the electrode morphology [6,32,61]. Providing the adequate electrolyte with the right pH is a key factor to control the reaction pathway and the stability of the reaction intermediates [19,62]. As an example, it has been reported that selectivity towards ethylene on Cu electrocatalysts can be fine-tuned by modifying the buffering capacity. At high local pH, ethylene production through *CO–CO dimerization pathway is both kinetically and thermodynamically favourable, whereas the C1 pathway is suppressed [63,64].In accordance with pH effect, H+ and OH− concentration has been shown to influence both, the activity and selectivity of the electrochemical CO2 reduction. Comparing several electrolytes containing potassium precursors (KCl, K2SO4, KClO4), phosphate buffer and KHCO3 aqueous solution diluted and concentrated, it was found that diluted KHCO3 KCl, K2SO4, KClO4 solutions favoured the formation of C2H4 and alcohols [65,66]. The presence of OH− anions produced from electrochemical reactions provokes a pH increase in the electrocatalyst surface enhancing C2 selectivity by suppressing HER [63,67]. Concentrated electrolytes (concentrated KHCO3 and phosphate buffer) are capable of neutralising this OH− anions. The local pH does not differ from the bulk pH and CH4 formation is favoured [65].Besides H+ and OH− ions, the electrolyte cations and anions nature was also reported to have a significant effect on product selectivity. Increasing size of alkali cation from Li+ to Cs+ usually leads to higher electrochemical CO2 reduction rates and higher C2/C1 products. This behaviour was attributed to differences in the local pH and changes in the electrochemical potential in the outer Helmholtz plane (OHP) due to the cation size [68]. The presence of hydrated alkali metal cations located at the edge of the Helmholtz plane stabilises the adsorption of surface intermediates such as *CO2, which is the intermediate precursor to the formation of C2 products through C–C coupling. Slighter cations are strongly hydrated, hindering cation-specific adsorption on the electrode surface [12]. Larger cations are more energetically favoured at the OHP than smaller ones, hence an increase in cation size typically leads to a large cations coverage [68].The influence of halide anions on electrochemical CO2 reduction has also been studied. The halide adsorption on the catalyst surface affects to the activity, selectivity, and the electronic structure, depending on the halide size and its concentration [69,70]. It has been reported that adsorbed halide anions could limit the proton adsorption, inhibiting HER [71]. While using Cu as electrocatalyst, halides are able to stabilize the *CO intermediate through the formation of a covalent Cu-halide interaction [69]. Due to this increase in *CO population on the catalyst surface, adding Cl−, Br− and I− to the electrolyte can lead to an increase in CO selectivity and hence a large methane production [69], or a lower overpotential and an increase in CO2 reduction rate while keeping C2–C3 faradaic efficiencies [70]. Among the mentioned halides, KI electrolyte led to the highest *CO formation on the catalyst surface and then, the highest C2 selectivity [72].Working at higher pressures encompasses the possibility of operating at higher temperatures and an increase in the CO2 solubility according to Henry's law [73]. This increase favours the adsorption of CO2 species on the catalyst surface due to the higher CO2 concentration in the electrolyte [22], and some works have reported higher current densities by increasing CO2 partial pressure [73,74]. Therefore, changing the CO2 partial pressure can tune the relative surface coverage and hence, the stability of CO2 reduction intermediates and the product selectivity [35]. The combination of both, an increase in the reaction pressure and a highly alkaline environment could improve CO selectivity and energy efficiency reaching industry relevant current density values [35]. Nevertheless, working at high pressures imposes some drawbacks since an unbalanced pressure could result in drying the catalyst surface or flooding of the catholyte causing operational issues [22].Regarding the temperature effect, it has rarely been studied and most studies were conducted at ambient temperature. An increase of temperature implies a lower CO2 solubility, so the undesired HER will prevail over the CO2 reduction process. Moreover, the temperature increase is also limited by the possible membrane degradation, the operational window is rather narrow, although it could be widened if coupled with a pressurised system [22]. In general, higher temperatures lead to higher currents since ionic conductivity increases and the diffusion coefficient also rises, thus the CO2 transport is more efficient [75]. The CO2 solubility problem could be overcome while supplying CO2 and water vapor in a gas phase electrolyser provided with a proper gas diffusion electrode (GDE). That way, the reaction kinetics would be enhanced due to the higher temperature and would compensate the poor solubility [22,75]. Nonetheless, further studies of temperature effects should be conducted in order to achieve the optimal operation conditions and assess its economic viability.The effects of pressure and temperature strongly depend on the cell design, i.e., their permissible and adequate values are certainly disparate if a gas phase system or a liquid phase system is used, and additional studies are needed. However, and considering the limited available studies, it could be stated that slightly higher than ambient conditions of pressure and temperature would be favourable in terms of product selectivity and reaction rate [22].The energy barrier between the onset potential and the standard reduction potential, the overpotential, is highly dependent on the working electrode. The different ranges of the applied overpotential can affect the preferable pathways to form C–C coupling. The *CO–COH coupling is dominant at high overpotentials, while the *CO–CO dimerization is favoured at low overpotentials [6,19].Up-to-date, CO2 cathodic reaction has usually been coupled with anodic water oxidation since water electrolysis is a well-known process without mass transport limitations, and all accomplishments and knowledge achieved in water electrolysers can be adapted to CO2 electrolysers [76]. However, oxygen evolution reaction (OER) requires a high overpotential that limits the energy efficiency of the whole electrolyser (OER can consume up to 90% electricity in CO2RR [77]) and oxygen gas could corrode and oxidise metallic catalysts or cell components [22]. Replacing this reaction for another with lower cell voltage requirements and providing high-value anodic products while maintaining a free of emissions process could be of interest. The work of Vass et al. [76] reviews and summarises different interesting alternatives to the widely used OER, including the CO2RR coupled with an already existing technology, the use of industrial waste as sacrificial agents or the production of raw and fine chemicals [78–80]. In the case of Li et al. [77], possible anodic oxidation reactions have been summarised and classified attending to the desired cathodic and/or anodic products. Some of the valued-added anode processes that have been lately paired up with CO2RR are summarised in Table 2 . Moreover, first technoeconomic analysis have shown promising results concerning CO2RR coupled with organic compounds oxidation [81], achieving an electricity consumption save up to 53% in the case of the glycerol oxidation [78]. Notwithstanding, the complexity of coupling CO2 cathodic reaction with these anodic reactions in terms of operation conditions, cell design and structure, product separation is still a challenge that requires further research (see Table 3).Significant efforts and improvements have been made on developing a suitable electrocatalytic reaction system that enhances the electrochemical CO2 reduction reaction [13,22,25]. As represented in Fig. 2 , electrochemical reaction systems can be categorised into H-cell systems, flow cell systems and microfluidic reactors.H-type cells are widely used because of their simple assembly, operation and products separation, their versatile configurations and low cost. In this system, working and reference electrodes are fixed in the cathodic reaction chamber while counter reaction is fixed in the anodic one. Both chambers are prefilled with the electrolyte and work without recycling. Their main disadvantage is the low CO2 solubility in the liquid electrolyte, which limits the current density [98], as well as the large distance between electrodes [99]. Moreover, selectivity towards C2+ products has always been diminished in favour of HER [13]. For all mentioned above, H-type cells are a suitable batch reactor for studying and comparing different electrocatalysts and products in lab-scale, although they do not fit in further industrial applications.The main component of the flow cell systems is the membrane-electrode assembly (MEA), where the electrodes are assembled to the solid polymer electrolyte or membrane, which acts a separation barrier of the chambers and is in charge of ion transportation between them. Depending on the type of ion transported, membranes can be cation exchange membrane (CEM), i.e., protons are transported from the anolyte to the cathodic chamber; anion exchange membrane (AEM), where hydroxide ions are transported from the anode to the cathode or the combination of both, bipolar membrane (BPM).According to the way of feeding the CO2 into the cell, two cell systems can be distinguished: gas phase and liquid phase electrolyser. In the former type, a humidified gaseous CO2 stream can be directly used as feedstock at cathode, enhancing this way mass diffusion and production rates [13]. The use of this cell design has also been reported to increase CO selectivity [100], to improve partial current density and stability for formate production [101], and to selectively generate C2+ products [14,102].In the liquid phase electrolysers, liquid electrolyte is in both electrodes. For that reason, the system can be pressurised and CO2 can be fed without further humidification [35]. In general, these flow cells have shown a better electrochemical CO2 reduction performance compared with H-type cells, mainly due to the enhanced CO2 diffusion and the local gas-electrolyte-catalyst interface [13].Microfluidic cells are an attractive alternative electrolyser configuration where, in some cases, the membrane is replaced by a thin gap between the electrodes filled with flowing electrolyte stream. The CO2 molecules are easily diffused into the electrode-electrolyte interface. Besides avoiding issues related to membrane degradation and cost, its absence allows wider pH and reaction temperature windows [103]. However, the existence of crossover could result in the products oxidation at some extent, although a multichannel design, a nanoporous separator [104] or a dual-electrolyte system [105] could solve this problem.The electrode surface and morphology has been reported to strongly affect in the activity and selectivity on the electrochemical CO2 reduction. Formation of defect sites, nano or complex structures, subsurface oxygen or crystal facets (schemed in Fig. 3 ) are the dependant factors [13,32].The formation of defect sites from vacancies or grain boundaries has been pointed out as the responsible of increasing activity and selectivity in electrochemical CO2 reduction [106,107]. Grain boundaries can alter the surfer properties of particles, generating active sites and inducing a large surface catalytic area [26]. These modifications of the particles and the catalyst surface can lead to a lower energy barrier for CO2 reduction or to an enhancement of CO adsorption and the C–C bond formation with a consequent stabilization of C2 intermediates and higher selectivity for this C2+ products [107,108].Several efforts have been made to create nanostructured catalysts more active and selective than bulk or foil electrocatalysts. It was found that roughened Cu electrodes could provide a high local pH, which promotes C2H4 generation over CH4 [109,110]. Catalyst structure is capable of controlling C2+ products selectivity by tuning the pore size and depth of nanostructures such as nanowire arrays, nanofoams, nanoparticles or nanocubes [111–113]. Smaller and deeper pores increase ion concentration and intermediates residence times, enhancing C2+ products selectivity. Moreover, intermediates can be confined into the nanostructure generating long chain molecules [26,114].Besides physical factors, the intrinsic catalyst properties could also affect the activity and selectivity [19] and several studies have proved the strong facet dependence on electrochemical CO2 reduction product selectivity [6,19]. As an example, it has been observed in Cu catalysts that Cu(100) favours selectivity towards ethylene and the C–C coupling improving C2+ products selectivity while Cu(111) favours the methane and formate formation [67,115,116].The electrocatalyst performance could also be adjusted by tuning the oxidation states via oxide-derived materials. Using again the copper catalysts as an example, it was observed that Cu oxidation states can easily change during electrochemical reaction conditions and oxide-derived catalysts enhance the activity and selectivity towards C2+ products [6,19]. The presence of residual oxygen, subsurface oxygen or oxidized cooper states favours the stabilization of the intermediates thus enhancing C–C coupling [117,118].As mentioned before, due to the multiple possible reaction pathways, numerous products can be originated. However, product distribution can be tuned by designing an active and selective electrocatalyst. The complex formation of C2+ products is gaining attention due to their higher energy density and economic value [119], and several research projects are currently focused on their production [120–122]. The numerous and assorted applications, where these products can be valuable, are summarised in Fig. 4 . The search of a catalyst with adequate electronic properties to be selective for the C–C coupling is vital to generate species with one or more C–C bonds. Among metals, copper has been found particularly active for the formation of C2+ compounds, although a few non copper-based catalysts such as heteroatom-doped carbon materials [123], NiP [124], NiGa [125] or PdAu [126] have shown significant formation of C2+ species. In the following subsections we will discuss the catalysts design to successfully facilitate the production of key C2 and C2+ products which are highly appealing for the chemical industry and whose electrochemical production might result in a game-changing approach.3.1.1. Oxalate. Oxalate or oxalic acid is the simplest dicarboxylic acid widely used in the chemical industry in applications such as pharmaceuticals and textiles manufacturing, rare earth extraction, oil refining or metal processing [127–130]. Oxalate is a precursor to glycols, which are in turn precursors to valuable synthetic materials [131,132]. Some research has been done concerning the beneficial effect of its application to improve quality of some vegetable and fruits and to extend their storage time and prevent its early degradation [133–135] and its role in the recovery of valuable metal ions from lithium-ion batteries [136,137]. Regarding current investigations, oxalic acid effect is being subject of study in a treatment to a honeybee's disease [138].The electrochemical CO2 reduction products strongly depend on both the chemical nature of the electrode and the reaction medium. In a protic solvent as water, formate would be the main product [139,140]. Since the solubility of CO2 in aqueous solutions is really low (about 30 mM), water molecules would be available nearby the electrode surface and HER would be favoured. In fact, only hydrogen is yielded while water is the only component of the electrolyte, the charge transfer proceeds favourable to water [141]. Using aprotic solutions that provide a higher solubility (about 240 mM in acetonitrile as an example [141]) as solvents would lead to CO2*- intermediate disproportionation to carbon monoxide and carbonate or dimerization to oxalate ions [139]. Besides their higher CO2 solubility, aprotic nonaqueous solvents such as acetonitrile, dimethyl formamide, dimethyl sulfoxide or propylene carbonate have been widely used because of their large cathodic window and the inhibition of HER [130,142].In this aprotic medium, CO and carbonate were reported as the major products using metals such as Pt, Pd, In, Zn, Sn or Au as catalysts, while oxalic has been reported as the major product using a Pb, Sn or Hg electrode [143,144].It is known that the first and primary step on the electrochemical CO2 reduction is the formation of the intermediate radical ion CO2*-, a change and mass-transfer controlled reaction that must be conducted at a sufficiently cathodically polarized electrode. In a nonaqueous aprotic medium, two main competitive pathways have been repeatedly described in literature: nucleophilic coupling of CO2*- with nearby CO2, that could produce CO and carbonate, or oxalate through ECE mechanism; and the purely dimerization of two radical ions following an EC mechanism, which would also lead to an oxalate anion production [131,140,141,143,145–147]. Dimerization reaction is more favourable at high concentrations of CO2*- radical anions while the reductive disproportionation of CO2 could be conducted at low CO2*- concentrations [148]. The reaction mechanism for oxalate production is depicted in Fig. 5 .Lead has traditionally been chosen as the working cathode due to its high HER overpotential and its low cost, which favours CO2 reduction [143,144,147,149]. In nonaqueous medium, oxalate has been reported as the main product and oxalate faradaic efficiencies (FEs) exceeding 80% were registered from the very early studies [143,144]. The work of Eneau-Innocent et al. studied the role of Pb electrode in a propylene carbonate electrolyte [142]. Propylene carbonate large electrochemical window, CO2 solubility and relative low toxicity made this aprotic solvent suitable for the electrochemical reduction. It was found that Pb electrodes are very selective towards oxalate via direct dimerization route. In situ IR spectroscopy measurements revealed that unstable intermediate CO2*- radical ions evolve quickly to adsorbed Pb–CO2*-, which dimerizes, and oxalate is finally desorbed.Some studies have also explored Pb or stainless steel electrodes coupled with sacrificial anodic electrodes in order to easily remove oxalate product from the electrochemical system [130,147,148]. Zinc has been widely used as a sacrificial anode because it produces zinc cation which fast reacts with oxalate anion towards insoluble zinc oxalate, preventing the further reduction or oxidation reaction of oxalate anion [147]. Lv et al. achieved a 96.8% FE for producing oxalate in 0.1 M tetrabutylammonium perchlorate in acetonitrile with a lead cathode and a zinc sacrificial anode [147]. Such a high efficiency was a result of a combination of two factors: water total elimination and low temperature (5 °C), as shown in Fig. 6 . As observed in the reaction mechanism (Fig. 5), the presence of water in the electrochemical CO2 reduction negatively affects to oxalate selectivity (see Fig. 6) [60]. Therefore, in this study [147], the electrochemical system was subjected to a pre-electrolysis and CO2 desiccation process in order to remove possible trace of water from the electrolyte and from CO2. It was found a decrease up to 40% in oxalate FE where 1.25 v/v% of water was added to the electrolyte solution, confirming the interferences of water in electrochemical CO2 reduction. In the case of temperature effect, solubility of CO2 increases while decreasing the temperature, enhancing CO2 availability on the electrode surface for its reduction.The advantages of incorporating ILs as electrolytes were also explored for oxalate production in Pb electrodes. A novel aromatic ester anion functionalized IL of 4-(methoxycarbonyl) phenol tetraethylammonium ([TEA][4-MF-PhO]) was designed based on catalytic and stability properties of aromatic esters and quaternary ammonium salts, respectively [60]. DFT calculations were performed in order to investigate the interaction between aromatic ester anion of [4-MF-PhO]- and CO2 revealing the ability of the anion with phenoxy and ester double active sites to bend the stable CO2 molecule to CO2*-. H+ cations provided by anolyte were combined with the anion-radical generating [4-MF-PhO-COOH]- that eventually dimerized to oxalic acid (Fig. 6). High partial current densities (9.03 mA cm−2) and acceptable FEs (86%) obtained on Pb electrode in [TEA][4-MF-PhO]-acetonitrile electrolyte assert the integration of IL in this aprotic electroreduction systems for oxalate production and open a potential path of research and improvement.Besides all progress related to Pb or stainless steel electrodes, cathodic materials such as metal organic frameworks (MOFs), metal-complexes or carbon supported metals have also been explored to perform the electrochemical CO2 reduction to oxalate [132,150–152]. Metal organic frameworks are porous materials with a crystalline ordered structure capable of storing CO2, which have also presented attractive features such as high porosity or thermal stability and adjustable chemical functionality [153–155]. They are indeed an emerging family showing promising features for electro and thermal catalysis [156] and their careful implementation in CO2 to oxalate process merits further studies.Shentil Kumar et al. coated Cu-BTC on a glassy carbon electrode to prove it as an efficient electrocatalyst for the oxalate production from CO2 [151]. The electrochemically synthesized MOF was capable of sequestrating and electro reducing CO2 simultaneously. The electrochemical reduction can be conducted inside the pores through heterogeneous electron transfer between the MOFs and the previously adsorbed and compressed CO2 molecules, generating a 90% pure oxalic acid with a FE of 51%.Silver was also tested as cathodic catalyst in a carbon-silver hybrid configuration [150]. Babassu coconut, an important Brazilian vegetation that has been previously used for nanocarbon production [157], has been hydrothermally carbonised in the presence of silver nitrate solution, obtaining the Ag@C catalytic system (see Fig. 7 ). Comparing the results obtained with this system with those in absence of silver nanoparticles, it can be concluded that silver nanoparticles enabled the charge transfer for the CO2 reduction besides enhancing the electrocatalytic activity. Moreover, synthesis with longer residence times resulted in larger deposits of silver particles that influenced the size of the carbon spheres and increased oxalic acid production.As opposed to the previous works, Paris et al. reported the use of a Cr–Ga electrode supported on glassy carbon to produce oxalate in a KCl aqueous solution electrolyte [158]. This catalytic system consisted of a Cr2O3–Ga2O3 thin alloy film and, whereas neither Cr2O3 nor Ga2O3 films alone could produce oxalate, the alloy of the two oxides led to oxalate FE of 59% at a pH of 4.1. Organic solvent or those with low proton availability reported the production of the oxalate dianion; however, in this aqueous system, a combination of monoanionic and dianionic species was generated. The Cr–Ga system originates a more appealing product distribution since protonated oxalates are more desired species from an industrial point of view [158].Oxalate was obtained at a more positive potential than that required for CO2*- intermediate generation, i.e., this process provides a CO2*- independent and lower energetic pathway than those previously reported (Fig. 5). This low energetic reaction pathway could involve a C1 product as oxalate precursor, and experiments feeding CO, formate, methanol or their combinations instead of CO2 were conducted. It was observed that CO-methanol combination did lead to oxalate production. Labelling experiments showed that carbon atoms oxalated derived from CO, although both methanol and CO are required for its production. Results obtained from this report encourage additional studies to advance in oxalate production in aqueous solutions [158].3.1.2. Acetate. As already mentioned, the traditional synthesis of acetic acid involves a three-step energy-intensive process (methane or coal to syngas, syngas to methanol and methanol carbonylation to acetic acid) using methane or coal as feedstocks [159,160]. Therefore, an efficient direct production of acetic acid can revolutionise the production of this key chemical. Indeed, acetate is not only an important end-product, but a versatile platform chemical capable of producing medium-chain fatty acids, alcohols and bioplastics, e.g., vinyl acetate monomers, esters and acetic anhydride [161,162]. Further, some industrial processes like denitrification or biological phosphorus removal also use acetate as carbon substrate [163].Unfortunately, acetate has not received a lot of attention as product of electrochemical CO2 reduction. The reaction pathway for its formation remains certainly unclear, although several possible routes have been suggested (as depicted in Fig. 8 ). At high potentials and high local pH, ethanol and acetate could be directly produced from acetaldehyde via Cannizzaro-type reaction [164]. Garza et al. pointed out that the ratio ethanol/acetate did not follow the proposed reaction [165]. They suggested the existence of additional pathways, such as the acetate production as a by-product on the ethylene pathway through *OCH2COH dimerization to a three-membered ring compound and further reduction.In general, the formation of CO2*- radical anion was accepted as the first step in the electrochemical CO2 reduction [16,123]. Genovese et al. provided experimental evidence for the acetic acid formation from a nucleophilic attack of *CH3 adsorbed species by this intermediate radical anion [16]. Once the radical anion was adsorbed, it can be reduced at the electrode surface until a –CH2OH specie is originated, and then generating methanol or proceeding with reduction until –CH3 species generation. In the case of conducting the electrochemical reduction on a N-doped nanodiamond/Si rod array electrocatalysts, the in situ infrared spectroscopy identified OOC-COO as intermediate [123]. Based on the assumption that kinetics C–C coupling is faster than CO2*- protonation [166], the pathway proposed involves the combination of two radical anions to generate the OOC-COO intermediate that would be further protonated and reduced producing acetate. Following the same assumption and while using Cu(I)/C- doped boron nitride as catalyst in the presence of 1-ethyl-3-methylimidazolium tetrafluoroborate ([Emim]BF4)–LiI solution, Sun et al. proposed the formation of a [Emim–CO2]+ complex that reduces the electron transfer barrier to CO adsorption [160]. CO adsorbed is then reduced and protonated until methanol formation (detected product) that would couple with CO2*- to generate acetate. This process is promoted by the presence of the Lewis acidic cation Li+ and the strong nucleophilicity of I−, although a more detailed mechanism involving these species needs further investigation.Alternatives to the above describe route has also been proposed ruling out CO2*- generation. This is the case of the works of Panglipur et al. [89] and Grace et al. [167] In the former one, acetic acid is believed to be generated by the formation and adsorption of CO and H+ species, which would generate CH2 that, in turn, could react with CO and be hydrated until acetic acid generation [89]. In the second work, due to the ANH groups of polyaniline catalyst and the use of methanol as electrolyte, H adatoms are generated on the electrode surface and then transferred to the CO2 molecules generating formic acid that could be attacked by the methanol and form acetic acid [167].Although copper-based catalysts are the most suitable for C2+ products from CO2, bulk copper catalysts face limitations (poor selectivity, high energy input, low activity) that must be overcome tuning these copper catalysts. More selective catalysts can be generated with smaller nanoparticles, defect sites, core-shell structures, oxidation states or heteroatomic doping [168]. In that sense, several tuned copper-based catalysts have reported high faradaic efficiencies toward acetic acid or acetate.The choice of a suitable support could be a key to improve bulk copper performance. The incorporation of conducting polymers such as polyaniline have been proved to decrease the electrochemical CO2 reduction overpotential [167]. Cu2O nanoparticles, with oxygen species to easily adsorb CO2 and Cu + ions promoting CO2 reduction as active sites, were disposed in a polyaniline matrix. This configuration led to a synergistic effect which allows the generation of acetic acid as the main product with an efficiency of 63% [167].Boron-doped diamond (BDD) has been widely used in electrochemical application due to its adequate characteristics for those purposes [169]. Although it is able to produce formaldehyde on its own under ambient conditions [89], as support of copper particles, it resulted in a selective electrode to acetic acid generation. Acetic acid was not observed while using copper or BDD as electrodes separately [89]. Same behaviour was observed in an electrocatalyst based on copper nanoparticles supported on carbon nanotubes (CNTs), achieving a production of acetic acid with a selectivity close to 60% [16].Pure titanium oxide does not present activity in the electrochemical CO2 reduction process itself [170], but its ability to control the local pH and its good conductivity make nanotubular TiO2 array (TNA) a suitable support. In the work of Zang et al. TNA was applied as the support of a modified polyoxometalate Cu catalyst [168]. Polyoxometalates (a type of transition metal oxygen anion clusters) were induced to modify the local electronic and protonic environment of Cu nanoparticles conforming a Mo8 modified cubic fragmented Cu submicron particles on TNA (Mo8@Cu/TNA) catalytic system. The skeleton structure and SEM images of the catalytic system can be observed in Fig. 9 .Acetate is originated following the single carbon coupling pathway (schemed in Fig. 8), i.e., CO2 molecules were coupled with formed *CH3. Experiments on mechanism investigation showed the synergistic effects of Cu planes and polyoxometalate cluster promoting the generation of the intermediate *CH3 tuning the selectivity towards acetate. Moreover, Mo oxide clusters modified Cu catalyst obtaining a fragmented Cu surface rich on grain boundaries. Ag was selected as metal to substitute Cu, but in view of selectivity and FE obtained while using Ag instead of Cu, these control experiments just served to highlight the synergistic effect between Cu surface and Mo oxides.Wang et al. also incorporated Ag in the electrocatalyst, but combined with Cu, forming clusters (∼ 6 nm) of Cu and Ag on the surfaces of electropolymerized films, (Cu)m, (Ag)n/polymer/GCE [171]. The increase in the acetate formation as Ag was added to the nanoparticles evidenced the important role of surface metal composition. As observed in Fig. 10 , it is believed that the presence of Ag could modified reaction mechanism although is still uncertain. Ag catalysts are able to promote CO2 reduction to CO [172], so Ag sites could facilitate the CO formation that would be further captured on neighbouring Cu sites for C–C bond formation and later acetate production (Fig. 10).N-doped systems also represent an appealing family of materials for electrochemical CO2 conversion. Indeed, they are chemically and active stable compounds that present an overpotential for HER higher than most reported electrocatalysts [173]. This way electrochemical CO2 reduction could be conducted suppressing H2O reduction and hence, enhancing the selectivity for the desired product [174]. Boron nitride (BN) promotes the CO2 chemisorption in the presence of electrons, thus the role of a C-doped BN as support of a Cu(I) complex Cu(I)/BN-C for acetic acid production was evaluated achieving a FE of 80.3% [160]. The support and the catalyst complex behaviours were evaluated separately, finding that bulk BN is not active for CO2 reduction, but BN-C can generate large amounts of formic acid since N-doped carbon materials can absorb CO2 and stabilize CO2*- [175]. The high selectivity to acetic acid is then related to a synergistic effect between catalyst complex and BN-C. CO2 is first adsorbed and converted into CO2*- due to BN-C and then protonated and reduced into methanol that would generate the desired acetic acid via C–C coupling with the help of the Cu metal center [160].Due to the observed proper characteristics of N-doped materials, N-doped nanodiamond supported on a Si rod array (NDD/Si RA) (schemed in Fig. 11 ) was tested for electrochemical CO2 reduction showing high efficiency, fast kinetics and good selectivity towards acetate [123]. As observed in Fig. 11, an increase in the N content led to an enhance in acetate and formate production rates due to the defects sites presence and the carbon atoms polarized by the N doping [176]. Defects sites and polarized carbon atoms may promote CO2 adsorption and CO2*- stabilization and rod array structure provides a large surface area that facilitates electron transfer. The combination of all these effects would explain the promising results obtained [123].As mentioned before, Cu-based catalysts have shown excellent performance on electrochemical CO2 reduction and high selectivity towards C2+ products, although they are not the only ones and, as in the case of the N-doped nanodiamond catalyst, modified Fe, Mn, Au, Pd or In metals have demonstrated its activity and selectivity towards acetic acid [15,177–180].Fe oxyhydroxide nanostructures (Ferihydrite-like clusters, Fh-FeOOH) were also supported on N-doped graphitic materials (Fe/N–C), and FE values were up to 97% with a high selectivity to acetic acid [177]. Results supporting the same clusters on O-doped carbon were not so encouraging, revealing the favourable Fe–N interaction. N-dopants allowed the stabilization of Fe(II) species instead of Fe0 (as in the case of using O-dopants) that promotes HER. Acetic acid formation was attributed to the presence of these Fe(II) species adjacent to N sites, while Fe(II) species are believed to reduce HCO3 − species, N sites enable the C–C coupling.Indium metal favours the formation of C1 products, although modifying its electronic properties inducing transitions metal oxygen anion clusters such as polyoxometalate (which has also reported beneficial effects modifying Cu electrodes) [168] can tune its selectivity towards C2+ products. Li et al. [180] and Zha et al. [15] works have incorporated polyoxometalates as electrolyte to assist In for the electrochemical CO2 reduction. As expected, the electrocatalytic performance of In resulted in the formation of hydrogen and formic acid as the only products. The addition of SiW11Mn to the electrolyte gave a result the appearance of acetic acid as well as hydrogen and formic acid, according to ion chromatography, gas chromatography and mass spectrometry analyses [180]. The presence of SiW11Mn reduced the overpotential and enriched In0 species since it enabled the re-reduction of elemental In.In the case of adding SiW9V3 to the electrolyte, the reaction mechanism was investigated through IR monitoring experiments and gas chromatography [15]. Control experiments (electrolyte without adding SiW9V3, and SiW9V3 electrolyte without In metal as catalyst) reported that SiW9V3 was necessary for the acetic acid formation (as in the previous case using SiW11Mn) and In for the CO2 reduction since the first step on the mechanism was the In-facilitated formation of In–CO3 - (instead of CO2*- intermediate). Besides, XPS spectra indicated that the V-center of SiW9V3 participates in the electron transfer process decreasing the overpotential, i.e., V-center efficiently catalyses the reduction of CO2. Faradaic efficiencies in this system reached values up to 96% with nearly 96% selectivity towards acetic acid.The synthesis of C2+ products from CO2, such as ethylene or ethanol, is of high importance due to the essential role of these products in the chemical and energy industry [119,181–183]. Ethylene is widely used as a building block in the production of many raw materials such as polyethylene, ethylene oxide, vinyl acetate, and ethylene glycol. Usually, ethylene is obtained from steam cracking of naphtha under harsh production conditions (800–900 °C). Although steam cracking is the industry standard for ethylene production, it presents different disadvantages. This process is non-catalytic and non-selective and is high energy and capital intensive, yielding into many by-products which require extensive separations and purification [184,185]. In recent years, the ethylene electrochemical synthesis from CO2RR is gaining attention since this approach offers mild conditions and an environmental pathway for ethylene production [186,187]. However, the use of CO2RR for the highly selective production of compounds of economic interest such as ethylene is still a challenge [186]. Concerning the catalyst, in the electrochemical environment of this reaction, i.e., abundant protons and negative electrode polarization, different catalytic behaviours have been observed: Ni, Fe, Pt, and Ti cathodes preferentially produce H2 over CO/CO2 production. Post-transition metals such as Pb, In, Sn, and Tl mostly produce formates. Additionally, Ag, Au, Pd and Zn reduce CO2 only to CO. On the other hand, Cu possess the outstanding ability to reduce CO2 or CO to CH4, C2H4, C2H5OH, and a variety of products [188–190].Copper catalysts have been extensively explored in the electrocatalytic synthesis of ethylene. It has been proved that tuning the intermediates' stabilities can favour a desirable reaction pathway and by consequence, improve the selectivity [190–192] However, the mechanistic pathway for the formation of ethylene over Cu catalyst is still under academic discussion [19]. While the C–C coupling of two carbenes (*CH2) has been proposed as the determinant step for the synthesis of the C2 product from CO2RR over Cu catalyst [193], more recent, studies concluded that C2 product formations are more attributed to CO dimerization.Different theoretical and experimental studies have supported that the CO dimerization pathway is the limiting step for ethylene production [192]. As shown in Fig. 12 , in the CO dimerization, *CO + *CO (Fig. 12a), *C2O2 − intermediate can exist either from carbon and oxygen atoms from *C2O2 − intermediate, which are bounded to the catalyst surface in a bridging mode (Fig. 12b) or by the *C2O2 − intermediate which is bounded to the surface by two C atoms (Fig. 12c). However, both ways proceed to further protonation of the *C2O2 − intermediate giving place to more stable intermediates such as *CO–COH (Fig. 12d) of *COCHO [194]. Alternatively, Goodpaster et al. proposed that while at low overpotential, the formation of the C–C bond takes place in a reaction of two *CO bounded to the surface, at higher overpotential, the reaction between *CO and *CHO (Fig. 12e and f) is favoured as a result of the larger activation barrier to the formation of the CO dimer [195]. This intermediate is followed by the reaction with *CO to form *COCHO (Fig. 12g) and finally ethylene formation [164,165].Copper has been selected as an ideal catalyst in the electrochemical conversion of CO2 due to the thereof mentioned characteristics. Its reaction performance has been widely explored in CO2RR by controlling morphology [113], grain boundaries [166], facets [196], oxidation states [197], molecule decoration [188–190], and dopants [198]. Among these methods, the design of Cu metallic nanostructures for C2+ products seems more promising due to the simple synthesis and the easy study of the structure-activity relationship of the catalyst. For instance, Roberts et al. showed that modifying the structure of the copper surface is a novel pathway to improve ethylene efficiency and selectivity. These authors obtained nanotubes-covered copper (CuCube) surface with high selectivity and low overpotential to ethylene formation, but also probed the effect of (100) sites in the C–C coupling as a strategy to target multicarbon products [199].Recently Zhang et al. reported the design of Cu nanosheets as an electrocatalyst for ethylene synthesis from CO2RR. These nanosheets (Fig. 13 a) present defects in the size of 2–14 nm, which were observed to be strengths in the adsorption, enrichment, and confinement for reaction intermediates and hydroxyl ions on the catalyst. As shown in Fig. 13b, the maximum FE reached with these copper nanosheets was up to 83.2%, which is the highest value among all the studied electrocatalysts to date [200].Loiudice et al. explored the effect of shape and size of Cu nanocrystals (NCs) in ethylene production. They explored crystal cube and spheres of Cu NCs, observing an increase in the activity as the crystal size decreased [116].The modification of oxidative copper states has also been proved to be an alternative to enhance the selectivity and efficiency of C2+ products [197,201]. Anodized-copper was investigated in CO2RR as a tool to improve ethylene selectivity. It was observed that compared with a Cu foil, this anodized-Cu catalyst presented a two-fold improvement [202]. But most importantly, the selectivity continued stable over 40 h of the experiment. As it is observed in Fig. 14 a, the catalyst performance remained stable, retaining an average of 38.1% FE for ethylene production. In this experiment, it was also observed the electrochemical treatment of the as-synthesized catalyst as a critical parameter in the ethylene selectivity (Fig. 14b). When the material was exposed to electrochemical reduction treatment at mild biased potential, the species were observed to suffer a reduction, quickly decreasing the ethylene selectivity. Meanwhile, when the treatment was performed at highly biased harsh conditions, the catalyst presents much extended durability for selectivity C–C coupling. These observations give evidence of the relationship between Cu–O-containing surface states and the durability of ethylene production, which are of high importance to develop new strategies for controlling selectivity and durability of O–Cu catalyst for electrochemical synthesis of ethylene [202].The Cu-based alloys such as Ag–Cu, Au–Cu, Au–Pd, and Cu–Pt have been demonstrated to present a high efficiency in the C1 production via stabilization of the intermediates [203,204]. However, the effect of these alloys in the production of C2+ hydrocarbons seem to be more complex [205]. For instance, Chang et al. reported an atomically dispersed Cu–Ag bimetallic catalyst where it was observed that Cu-rich zones preferred the production of hydrocarbons. In contrast, Ag-rich zones are dominated by CO. These experiments highlighted the importance of the atomic ratio for CO2RR electrocatalysts [206]. Similar studies using Cu–Ag alloys from additive-controlled electrodeposition showed that Ag sites were believed to act as a promoter for CO formation during the electrocatalytic CO2 reduction, helping the C–C coupling in the neighbouring Cu due to the availability of CO intermediate [207].Molecular distribution in Pd–Cu catalyst has also proved to play an important role in improving the selectivity of C2+ compounds. For instance, Ma et al. demonstrated that a phase-separated Pd–Cu catalyst offered an ethylene selectivity up to 50% compared with its disordered and ordered counterparts (Fig. 15 a and b) [208]. Since according to the d-band theory, typically, lower d-band center transition metals show weaker binding on the in-situ generated intermediates on the metal surface [209]. The Pd–Cu alloy exhibited a similar catalytic activity and selectivity for CO that the ones obtained with Cu nanoparticles (NPs). It offered a wide d-band difference (Fig. 15c) which helps to conclude that the geometry and structure effect may play a more important role than the electronic effect for enhancing the selectivity of hydrocarbons in phase-separated Cu–Pd alloys case [208].The nature of the electrocatalytic CO2 reduction imposes the need of catalysts with specific active sites for the reactants adsorption and the further transformation of multiple intermediates, thus makes primary importance the increasing of specific active sites in the design of catalyst. As discussed in the oxalate section, MOFs have emerged as an alternative to act as both, as electrocatalysts or precursors to derive in different heterostructured catalysts due to their extraordinary properties and could be useful for electrochemical ethylene production from CO2. In fact, from an electrochemical point of view, the permanent porosity, ultrahigh surface area, coordinatively unsaturated metal sites, adjustable pore size, and active sites homogeneous dispersion makes them highly attractive for this kind of reactions [210].Post-synthesis modification of MOFs is a strategy of high relevance in the design of electrocatalysts since it could assemble the functional metal fractions such as metal porphyrin and metal complex that would enhance the CO2 reduction in MOFs, particularly in those with coordinatively unsaturated metal sites, which interact with CO2 working as Lewis acid [29,211,212]. Cu-BTC (HKUST-1), which presents open metal sites, has been widely studied in the CO2RR due to its structural features that enhance the catalytic performance. Nam et al. reported the strategy involving the MOF-regulated Cu cluster formation (Fig. 16 a) as a tool to optimize the selectivity of ethylene in comparison with MOF-based active carbon. They obtained an efficiency of up to 45% (Fig. 16b) [213].MOFs derivates such as metal oxides, porous carbons, carbides, phosphides, and nitrides have appeared to overcome the main limitations of MOFs related to the poor stability and conductivity in electrochemical reactions [214–217]. For instance, Zheng et al. studied N-doped nanoporous carbon obtained from ZIF-8 for ERC, emphasizing the calcination temperature in the reaction and the mechanism. According to the catalytic performance, higher pyrolysis temperature resulted in a higher activity with a maximum FE for CO formation of 95.4% [218].The design of atomically dispersed metals in activated carbon (AC) has also been a powerful tool in the electrochemical conversion of CO2. As shown in Fig. 17 , Guan et al. boosted the CO2 electroreduction to CH4 and C2H4 by tuning the neighbouring single-copper sites in a dopped-AC showing the Cu-coping concentration as a tool to direct the selectivity to the desired product. For instance, they showed that at a Cu high concentration, the distance between Cu-Nx species was close enough to enable C–C coupling and, by consequence, produce C2H4. In contrast, a Cu concentration lower than 2.4% mol, the distance between species was larger to the formation of CH4 was favoured [219].The electrocatalytic conversion of CO2 to higher-value hydrocarbons beyond the C1 products is an area of extraordinary importance for applications such as transportation, fuels, energy storage, and the chemical industry [220]. With a worldwide production of ca. 88.5 Mt/y and a projection of US$105.2 billion by 2025, ethanol is an important organic chemical for the biofuel and food industries [221,222]. For instance, approximately 80% of global ethanol produced is used as fuel, followed by food, pharmaceutical, and cosmetics applications [223].In recent years, electrochemical production of ethanol has been proposed as a green route to obtain this high-value chemical. As observed in Fig. 12, the electrochemical production of ethanol takes place in a similar pathway to ethylene, where the C–C coupling activity is considered a critical step in ethanol production, similarly to other C2+ products. Garza et al. proposed an ethanol pathway over Cu catalyst where it was observed that *COCHO intermediate (Fig. 12g) is the key to determinate the selectivity between ethylene and ethanol [165]. The glyoxal can be reduced to acetaldehyde and ethanol (Fig. 12h), followed by a further reduction to glycolaldehyde (Fig. 12i) and ethylene glycol/vinyl alcohol (Fig. 12j) or acetaldehyde, and finally to ethanol [164,189].The main goal for CO2RR to ethanol is to improve selectivity at high conversion rates, which are influenced by the catalyst and the process conditions. By far, Cu-based catalysts have been the most studied to catalyse CO2RR for ethanol production [201,224,225]. In this sense and similarly to ethylene, different strategies have been studied to enhance ethanol selectivity in Cu-based catalysts, such as nanoparticle morphology [226,227], oxidation states [228], and Cu-based alloys [208,229]. For instance, Duan et al. studied the crystallinity of Cu nanoparticles as a driven parameter in the ethanol selectivity, observing that amorphous nanoparticles enhance the adsorption of CO2 at room temperature, which plays a key role in the CO2RR [230].Several studies have demonstrated the strong relationship between product selectivity of CO2RR and the crystal facets of Cu [231]. For example, Cu(100) increases the selectivity for ethylene, while in Cu(111) catalyst, methane is the main hydrocarbon product [67,115,116,232]. Jiang et al. developed and studied an efficient nanocube-shaped catalyst with significant improvement in C2+ selectivity. It was reported that Cu(100) and (211) facets favour ethanol and other C2+ products over Cu(111) via dimerization of *CO to form *OCCO and the subsequent proton-electron transfer and surface hydrogenation to form *OCCHO [196].The oxidation states of copper can vary among Cu0, Cu+, and Cu2+, and their states change reversibly during electrochemical reaction conditions. Copper oxide-based catalysts improve the activity and selectivity of C2+ products. Generally, copper oxide-based catalysts are synthesized by the growing of Cu2O from Cu-based precursors followed by high-temperature treatment and consequently reducing Cu2O to form Cu0 sites. However, the elucidation of the pathways of this enhancement toward C2 formation is still under debate [19]. Tentatively, the enhancement of the selectivity for ethanol of such catalyst could be attributed to (i) the presence of residual oxygen atoms close to the surface that favour the modification of the electronic structure of Cu atoms and the increase of the CO binding energy, which by consequence promotes C–C couplings [233]; (ii) the strong adsorption of H2O molecules due to the presence of residual Cu + which may work synergically with Cu0 sites. These H2O molecules favour the CO2 conversion to CO [117,118]. For instance, Handoko et al. reported a mechanistic study electroreduction of CO2 to ethanol and other C2+ hydrocarbons on Cu2O-derived films. They synthesized Cu2O films with five different morphologies and proved a relationship between the crystal size and the selectivity to C2 products being able to reach a FE of c. a. 20% in middle-sized particles [234].A wide variety of morphologies of copper oxide-based materials has been reported as a determinant factor to enhance ethanol production. For instance, Daiyan et al. reported the synthesis of nanowires of Cu2O/CuO/Cu foam (Fig. 18 a) that exhibited a FE for ethanol formation of 31% [235]. Also, 3D dendritic Cu–Cu2O/Cu catalyst (Fig. 18b) were synthesized and proved a FE up to 39.2% attributed to the high density of exposed active sites and favourable Cu2O:Cu ratio [220].Copper alloys have attracted considerable attention as a strategy to change the composition of Cu single crystal catalysts via combining it with a secondary metal able to produce H2 (i.e., Fe, Ni, Ag, Au, and Pd) and to tune the activity and selectivity through optimization of the binding strength of the key intermediates on the surface of the catalyst [6]. For instance, Shen et al. described the synthesis of submicron arrays based on the alloy CuAu reaching a FE up to 29%. It showed the critical role of Cu:Au ratio content in the selectivity ethanol/ethylene [236]. Interestingly, in the same way as metal ratios, the phase distribution of the metal involved in the alloys has been proved to be a determinant factor in directing the selectivity to ethylene or ethanol. As observed in Fig. 19 a, which shows the ethanol/ethylene selectivity in an Cu/Ag catalyst, the selectivity is three times higher in the catalyst with phase blended, probably related to the effect of the dopant in the surrounding Cu atoms but also to the effect of Ag–Cu biface boundaries that suppresses the HER and favours the formation of mobile CO on Ag sites and its further reaction to a residual intermediate on Cu sites (Fig. 19b and c) [225].In the last decade, MOFs, and especially those formed by Cu metal ions or clusters, have served as active electrocatalyst in CO2RR but also as precursors for highly dispersed Cu over N-doped carbon catalysts [237,238]. For instance, Zhao et al. synthesized a porous Cu/C catalyst consisting of Cu2O and metal Cu particles embedded in a porous carbon matrix through the pyrolysis of Cu-based MOF (HKUST-1) (Fig. 20 ). A maximum FE for ethanol production of 34.8% at a potential of −0.5 V was obtained using this selective catalyst [239].The electrochemical reduction of CO2 to C3+ products remains a challenge. Among the C1–C3 alcohols, n-propanol possesses the highest energy-mass density (30.94 kJ/g) and an octane number up to 118 [240,241]. These relevant properties make the efficient and green production of n-propanol a target point nowadays.The n-propanol production by a CO2 reduction reaction has been only observed in Cu-based catalysts. It has been hypothesized the pathways for n-propanol production as a result of the transformation of acetaldehyde into vinyl alcohol by a tautomerization equilibrium [227,242]. As observed in Fig. 21 , the adjacent intermediates CO and *CH2 are inserted into *CH3–CH in a similar way. Then, the reduction to propionaldehyde (CH3–CH2–CHO) and, consequently, to n-propanol takes place [20].Ren et al. reported a mechanistic study to n-propanol electrosynthesis via the design of nanocrystals agglomerates with unprecedented catalytic activity. The defect sites generated in the catalyst surface was responsible for the improved activity of the Cu nanocrystals and, hence, the n-propanol formation [227].In contrast with the previous sections, in which each product has been reviewed separately, here we give a general overview of the market aspects and techno-economic analysis carried out to date. The market size of the targeted products is critical from both a commercial perspective and CO2 utilization potential. In this vein, Fig. 22 presents the approximate market size of the C2-products here considered [243–249]. As shown, for example, propanol and acetic acid market sizes are very low in comparison with other potential CO2 utilization routes. For instance, other relevant CO2 utilization alternatives such as CaCO3 production, have a market sizes one hundred times higher than propanol and acetic acid (116 Mton/a market size) [250]. Nonetheless, their production from CO2 should not be dismissed as the production capacities may fit with the capacities of small-medium CO2 emitters. A fair example of these emitters is biogas upgrading, that could achieve the category of negative CO2 emissions technology if propanol or acetic acid are produced from the CO2 contained in this bio-resource. Oxalic acid is a very interesting C2 product from a utilization perspective since it is demanded by pharmaceuticals, textiles manufacturing, rare earth extraction, oil refining or metal processing industries. Nonetheless, worldwide oxalic acid production is above 0.230 Mton/a [249], which places oxalic acid in the same range than propanol and acetic acid. Therefore, this option cannot provide an alternative for CO2 utilization of large emitters. Ethylene and ethanol present market sizes considerably greater, and as explained in previous sections, the production of these chemicals is crucial for the end-products consumed by our society nowadays.On the other hand, Fig. 22 also presents the market prices for the products studied. From a market price point of view, ethanol, ethylene and propanol are the most attractive products to be achieved via CO2 electrocatalytic reduction. Although rapidly booming, CO2 utilization technologies are currently far from being competitive with traditional technologies, hence focusing on products with elevated market prices could be a good strategy to balance economic performance. Following this reasoning, the production of oxalic and acetic acid from CO2 would be difficult to become economically profitable with the current market prices in the short term. However, the growing projection for CO2 emission taxes along with the ongoing commitments to pursue a net-zero emissions opens the possibility for the viable production of all these chemicals using electrochemical routes.Concerning techno-economic works carried out to date, only a few studies are reported in literature. Economic assessment of CO2 electrocatalytic reduction are scarce probably because of the complexity of the analysis and the lack of reliable commercial scale data. In this sense, this criticism aims to be an encouraging call for experts in techno-economic and profitability evaluations. Focusing on the works performed, Table 1 gathers the most important characteristics and findings of the techno-economic analysis available in the literature. Perhaps the most impacting study carried out to date was performed by Jouny et al. [243] In this work, authors present the end-of-life net present value (NPV) of a 100 ton/day plant for various CO2 reduced products: propanol, formic acid, carbon monoxide, ethanol, ethylene and methanol. From the products targeted in their study, here we focus on propanol, ethanol and ethylene. Under the current conditions, authors conclude that the production of these C2-products from CO2 is not profitable. Nevertheless, the result could be reverted if reasonable electrocatalytic performance benchmarks are achieved. In agreement with these authors, these performances must be 300 mA cm−2 and 0.5 V overpotential at 70% FE. Another very important work was presented by Kibria et al. [244]. In their work, a techno-economic analysis was carried out with the aim of ensuring the economic viability of the process by the identification of profitable CO2 electrocatalytic reduced products as well as the performance targets that should be met to achieve it. As in the previous discussed work, this study includes parameters such as current density, energy and faradaic efficiencies, and stability. The most interesting point of this work is the prices predicted for ethanol and ethylene to achieve profitable scenarios. According to these authors analysis, ethanol price should achieve 1400 $/ton to get a net present value equal to zero (revenues equal to costs at the end of the plant life). In comparison with the current market price for ethanol (around 1000 $/ton, see Fig. 22), the difference is 40%. This fact shows the great technological challenge that we face to make electrocatalysis economically attractive in the context of direct CO2 conversion to C2 products. In the same study, a price of 1700 $/ton is predicted for ethylene to be a profitable alternative. Again, in comparison with current market prices (approximately 1300 $/ton), the threshold needed is remarkably higher. Interestingly, the authors present a sensitivity analysis based on a tornado plot, revealing that current density is the main parameter with room for improvement for all the products considered.The transition towards sustainable modern societies relies on the implementation of disruptive technologies for CO2 utilization. Among these technologies, electrocatalytic CO2 conversion to added value products will play a major role given its advantages compared to traditional thermal catalysis. In particular, the fact the electrochemical reactions can take place at very mild conditions represents a major bright side of this approach. So far, most the academic works are focused on the conversion of CO2 to C1 products since the reaction is “electrochemically cheaper”. Nevertheless, the production of C2 and C2+ is more appealing for the chemical industry given the broad market applications of these advanced products. Herein, challenges on the catalysts design to achieve high faradaic efficiencies and end-product selectivity are identified as the main bottlenecks for the electrochemical CO2 conversion to C2 and C2+ compounds. Among the different studied heterogenous catalysts, Cu-based formulations outstand showcasing the best activity/selectivity balance. However, in many cases their performance is still below the threshold to be considered as commercially viable options. In this sense, new formulations of advanced materials including multi-alloy systems, N-doped catalysts, optimized porous MOFs structures among many others are under development showing promising results. In this sense, our review provides an end-product guided perspective of the progress within catalyst design to deliver C2 and C2+ from electrochemical CO2 conversion routes. Beyond offering an illustrative analysis for experts and a straightforward starting point to newcomers in the field, our work would also like to emphasize the need to strengthen the research efforts within the catalysis and energy communities in the conversion of CO2 to advanced products. More precisely we advocate for a C2 and C2+ production using a new generation of advanced electrocatalytic materials.Along with the catalysts design and electrochemical processes considerations, market studies are crucial to ascertain the viability of the CO2 electrochemical conversion routes. Our analysis indicates that products with a broad market such as ethanol, ethylene and propanol are worth exploring in the short term while electrochemical production of other key products such as oxalic and acetic acids are not yet economically appealing. Still, progress on catalyst design pushing forward products selectivity and overall CO2 conversion will certainly help to make these options also viable. In any case, the production of advanced products using CO2 as carbon pool and electrocatalysts design seem to share a common destiny whose convergence will result in a remarkable contribution to decarbonise the chemical industry, opening new routes for the desired low-carbon future.The authors equally contributed to this work.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.Financial support for this work was gathered from Spanish Ministry of Science and Spanish Ministry of Science and Innovation through the projects RTI2018-096294-B-C33 and RYC2018-024387-I. This work was also partially funded by the University of Seville via the VI PPIT grant scheme for talented researchers. Support from CO2Chem UK through the EPSRC grant EP/P026435/1 is also acknowledged. Financial support from the European Commission through the H2020-MSCA-RISE-2020 BIOALL project (Grant Agreement: 101008058) is also acknowledge.

The energy crisis caused by the incessant growth in global energy demand joint to its associated greenhouse emissions motivates the urgent need to control and mitigate atmospheric CO2 levels. Leveraging CO2 as carbon pool to produce value-added products represents a cornerstone of the circular economy. Among the CO2 utilization strategies, electrochemical reduction of CO2 conversion to produce fuels and chemicals is booming due to its versatility and end-product flexibility. Herein most of the studies focused on C1 products although C2 and C2+ compounds are chemically and economically more appealing targets requiring advanced catalytic materials. Still, despite the complex pathways for C2+ products formation, their multiple and assorted applications have motivated the search of suitable electrocatalysts. In this review, we gather and analyse in a comprehensive manner the progress made regarding C2+ products considering not only the catalyst design and the electrochemistry features but also techno-economic aspects in order to envisage the most profitable scenarios. This state-of-the-art analysis showcases that electrochemical reduction of CO2 to C2 products will play a key role in the decarbonisation of the chemical industry paving the way towards a low-carbon future.

Efficient and sustainable electrocatalysts are needed to increase the competitivity of hydrogen energy over conventional, polluting power generation. Especially in acidic environments, such as proton exchange membrane fuel cells (PEMFC), where the stability of the catalyst is critical, material selection and design is very important [1]. Since Pt is the most active but also expensive catalyst, efforts are made in order to find a compromise between cost and performance of the catalyst [2]. Alternatively, water electrolysis and fuel cells can operate in alkaline media as a less aggressive environment for the catalyst [3].Apart from the selection of a catalyst material, the structure of catalysts is optimised in order to maximise their surface area while minimising the amount of material [4]. This can be achieved by synthesising materials with high surface-to-volume ratios such as porous structures, nanoparticles, or nanotubes. Thin films, in spite of their lower surface area, typically show higher specific electrochemical activity compared to nanoparticles [5].Ni is a good candidate to partially replace Pt via alloying. Several features related to the Gibbs free energy, electronegativity and lattice mismatch point to a good electrocatalytic activity of Ni-Pt alloys towards HER [6]. Ni is commonly used for the protection of electrical contacts, and electrodeposited Ni coatings are often used for corrosion protection [7,8]. For example, electrodeposited Ni-based Ni-Co-B coatings have shown to improve the corrosion resistance in a fuel cell environment compared to uncoated stainless steel and Al 6061 alloy [9]. Many Ni-based compounds have been investigated for HER, however almost exclusively in alkaline media [10–13].A critical issue for Ni-rich alloys is hydrogen embrittlement, which can occur especially in environments containing H2S, and even during the hydrogen evolution reaction [14]. Electrodeposited Ni films are especially susceptible to hydrogen embrittlement due to interstitial, monoatomic hydrogen dissolved in the crystal lattice [15]. However, this form of hydrogen embrittlement is reversible [16].Single-phase mesoporous Ni-Pt thin films exploit both the increase of surface area provided by the porous structure, as well as a reduced usage of Pt via alloying with Ni. In a previous study, Ni-rich films were produced by electrodeposition and thoroughly characterised, showing excellent performance at HER in 0.5 M H2SO4 [17]. A good stability over 200 cycles of HER was reported for Ni-Pt films with different compositions, ranging from 99 at% Ni (1 at% Pt) to 61 at% Ni (39 at% Pt). Yet, their electrochemical behaviour in acidic and alkaline media remains to be explored. In view of their potential integration in PEMFCs, their long-term stability must be further assessed by corrosion studies to determine if the mesoporous Ni-Pt thin films can be safely used in acidic media, or if an alkaline electrolyte is more favourable for long-term stability instead.Also, the effect of the composition on the electrochemical behaviour must be investigated. The selective dissolution in acidic media, also known as leaching, is a common issue for Pt alloyed with a transition metal [18], but may be hindered in single-phase alloys, where the noble metal can protect the transition metal atoms if its content is sufficiently high. Other strategies with the intent to minimise the leaching of Ni include the synthesis of core-shell structures with a Pt-rich surface protecting the Ni-rich core [19]. The corrosion resistance of Ni-Pt alloys is expected to increase with the Pt content. However, any effect of the mesoporosity on the electrochemical behaviour needs to be investigated. For instance, the porosity may lead to crevice corrosion due to differential aeration inside the pores [20].This work focusses on the effect of composition (Ni and Pt contents) and mesoporosity on the observed electrochemical behaviour, corrosion properties, and long-term stability of Ni-Pt alloy thin films. In order to assess the electrochemical properties of Ni-rich Ni-Pt alloy thin films, the behaviour is studied both in acidic (0.5 M H2SO4) and alkaline (1 M NaOH) electrolytes, two media which are commonly used to study materials for acidic and alkaline fuel cells. The studies were conducted by cyclic voltammetry (CV) and with the use of the electrochemical microcell technique (EMT), which allows for multiple measurements on a single sample, with the aim to identify the oxidation and reduction reactions occurring. In addition, electrochemical impedance spectroscopy (EIS) was used to provide an understanding of the corrosion resistance of the Ni-Pt films in acidic media. Althoughcathodic potentials, which are usually not critical in terms of corrosion, are applied during HER, the electrode potential can rise well into the range of oxidising potentials when not in operation [21].All electrochemical tests were performed on single-phase, nanocrystalline Ni-Pt thin films with varying composition, both with and without mesoporosity, which had been structurally characterised elsewhere [17,22]. A series of dense and mesoporous Ni-Pt films, ranging from 61 at% to 99 at% Ni, are investigated in the present study (Table 1 ).The samples consisted of a Si wafer sputter-deposited with a Ti adhesion layer and a Cu seed layer, onto which Ni-Pt films were grown (Fig. 1 ). The films were potentiostatically deposited from an aqueous electrolyte, using a micelle-forming surfactant in the case of the mesoporous films [17]. TEM analyses were performed on a Jeol JEM-2011 at 200 kV acceleration voltage. Sample preparation was done by grinding, polishing and Ar ion milling (for mesoporous Ni92Pt8) as well as by cutting with a focused ion beam (FIB, for dense Ni91Pt9).An Autolab 302N potentiostat/galvanostat was used to perform all electrochemical tests. Initial CVs were performed in a conventional three-electrode set-up using an Ag|AgCl reference electrode (RE) and a platinum spiral as counter electrode (CE). HER in 0.5 M H2SO4 was investigated by linear sweep voltammetry (LSV), sweeping the potential from –0.15 V to –0.5 V vs. Ag|AgCl in an identical set-up but with the use of a graphite rod as CE. For evaluation of the long-term stability at HER in 0.5 M H2SO4, a 24 h long electrolysis was performed at a geometric current density of –10 mA/cm2. Results obtained from these methods, which deal with the characterisation of the whole films (as opposed to local measurements) are hereafter denoted as global scale tests.The EMT was used to study the electrochemical behaviour of the Ni-Pt thin films in detail. Contrarily to its usual scope to study single inclusions, grains, phases etc. [23], and due to the homogeneous, single-phase and nanocrystalline character of the Ni-Pt thin films, this method was used here to study the overall representative behaviour of the films locally, and, moreover, allowed to perform multiple measurements on the same sample. Results from this method are further denoted as local technique, local scale or EMT.The set-up consists of an optical microscope, into which the electrochemical cell is mounted, and the sample functioning as working electrode (WE) is placed on a conductive sample holder, all placed inside a Faraday cage (Fig. 2 ). The samples’ surfaces were electrically connected to the sample holder via copper tape. The electrochemical microcell is filled with the electrolyte, a Pt wire is used as CE and an Ag|AgCl electrode as RE. The electrolyte is connected to the WE through a glass microcapillary with a tip diameter of 50–200 µm. The tip is covered with a silicone gasket to protect the capillary and facilitate the contact with the WE [24]. Using the objectives of the optical microscope, the area to be measured was checked before each measurement in order to confirm that it was homogeneous and free of defects or surface pollution.The EMT was used in the following studies: • CVs • in NaOH from –0.5 V to 0.6 V and 1.5 V vs. Ag|AgCl • in H2SO4 from –0.5 V to 0.5 V vs. Ag|AgCl • EIS in H2SO4 CVs • in NaOH from –0.5 V to 0.6 V and 1.5 V vs. Ag|AgCl • in H2SO4 from –0.5 V to 0.5 V vs. Ag|AgCl in NaOH from –0.5 V to 0.6 V and 1.5 V vs. Ag|AgClin H2SO4 from –0.5 V to 0.5 V vs. Ag|AgClEIS in H2SO4 EIS in NaOH did not yield any valid data since the currents were below the detection limit. EIS was conducted in a frequency range from 100 kHz to 3 mHz, after an equilibration time of 5 min at open circuit potential (OCP), with an amplitude of 10 mV.The thin films were analysed before and after electrochemical measurements by scanning electron microscopy (SEM) coupled with energy-dispersive X-ray spectroscopy (EDX) on a Zeiss Merlin electron microscope to study the effects of exposure to acidic and alkaline media on the microstructure and composition. Imaging was performed using the InLens detector with an acceleration voltage of 1–2 kV, while an acceleration voltage of 20 kV was used for EDX.For quantification of dissolution or leaching of the mesoporous Pt-Ni thin films in sulfuric acid, the surfaces of Ni95Pt5, Ni92Pt8, and Ni84Pt16 films were immersed in 0.5 M H2SO4 at OCP for 10 min and the solution was then analysed by inductively coupled plasma mass spectrometry (ICP-MS) using an Agilent 7500ce spectrometer to determine the amount of dissolved Ni and Pt.The TEM cross-sections of the mesoporous Ni92Pt8 (Fig. 3 a) and dense Ni91Pt9 (Fig. 3b) films are shown for comparison. While the latter shows its fully dense appearance, a homogeneous distribution of pores can be observed in the mesoporous counterpart. Film thickness lies between 200 nm and 300 nm, which holds for all compositions.At high resolution, the nanocrystallinity of the films is revealed (Fig. 3c) and further confirmed by the selected area electron diffraction (SAED) pattern, where the existence of arbitrary crystal orientations of the single-phase fcc Ni-Pt alloy is shown (Fig. 3d).The performance of the Ni-rich Ni-Pt thin films at HER in 0.5 M H2SO4 (Eq. (1)) shows that hydrogen production is highly reproducible, with very little changes up to 200 sweeps. Representatively for all compositions, the Ni84Pt16 alloy film shows the HER during LSV, reaching current densities of up to 150 mA/cm2 at –0.3 V vs. reversible hydrogen electrode (RHE, Fig. 4 a). A certain deviation between individual sweeps is always obtained due to the temporary blocking of the surface by hydrogen bubbles being formed [17]. Comparable activity at HER in the same media has been reported for mesoporous Pt-rich Fe-Pt films [25]. Thus, the Ni-Pt films reported here are able to achieve the same activity at HER with a significantly lower Pt content. (1) 2 H + + 2 e − ⟶ Ni-Pt H 2 Long-term electrolysis experiments indicate that the recorded potential changes over time. Specifically, at constant operation at –10 mA/cm2, it is observed that an increasingly cathodic potential is needed to maintain the HER current (Fig. 4b). However, irrespective of composition or the presence of porosity in the alloy, this potential stabilises over time and indicates that no further degradation in performance is expected. The higher the Pt content in the films, the lower is the resulting change in potential. An increase in the overpotential required for HER and oxygen evolution reaction (OER) at constant current density in macroporous electrodeposited Ni has also been observed in alkaline media [26]. Contrarily, the constant operation of Ni foam at HER in 0.5 M H2SO4 at a fixed potential for 15 h revealed a shift of the current density towards more negative values, which was related to a surface roughening of the Ni foam [27].The microstructure of the films after 200 sweeps of HER appears unaltered compared to the as-deposited state (Fig. 5 a, b). Upon closer examination, the size of the mesopores appears slightly bigger after HER experiments, suggesting the removal of material such as oxides, hydroxides or other contaminants from the surface. A comparison of EDX spectra before and after HER did not reveal significant differences in the Ni/Pt ratios.After the 24 h long electrolysis, cracking was observed in dense and mesoporous Ni-Pt films (Fig. 5c). In all cases, a slight decrease in Ni content after the electrolysis was observed, in addition to particles containing nickel and sulfur formed on the surface (possibly nickel sulfate), visible as white particles in Fig. 5c. Both these effects are likely to be the origin of the potential increase observed during electrolysis.Although the Ni-Pt films are targeted for application in acidic media, application in alkaline media is a possible alternative. Since alkaline media is far less aggressive, it allows for non-destructive electrochemical characterisation, and for validation of the set-up and measurement parameters before the behaviour of the Ni-Pt films in acidic media is studied. As aforementioned, a period of inoperation of electrolysers may transiently cause the application of oxidising potentials to the cathode and it is therefore important to evaluate the electrochemical response of the catalyst in the anodic range [21].In the CVs obtained from both global scale and local measurements in NaOH, a redox reaction is observed within the potential window from –0.5 V and 0.6 V (Fig. 6 a). The oxidation and reduction peaks correspond with the redox reaction between Ni(OH)2 and NiOOH (Eq. (2)) [28]. (2) Ni ( OH ) 2 + OH − ⇌ NiOOH + H 2 O + e − Both methods reveal the same redox reaction at similar current densities, confirming that—due to the homogeneity of the Ni-Pt thin films—the EMT technique is able to capture the films’ global electrochemical properties. Using the local technique, the oxidation and reduction peaks are wider with respect to the global technique, and the peaks are more separated. This observation is made for all film compositions, and applies especially at higher scanning speeds. This is probably due to the very small volume of the electrochemical microcell, where the availability of species such as OH − is reduced and the kinetics are thus slowed and limited by the diffusion of those species through the capillary as a very constrained path. This effect is more significant at high scanning speeds, where the depletion of species available at the electrode is even faster.At low scan rates with the local technique (i.e. 10 mV/s and 20 mV/s), progressive widening of the peaks makes it possible to discriminate between two oxidation peaks which were not resolved on the global scale (Fig. 6b). The split into two oxidation peaks can be explained by both the oxidation of α- and β-Ni(OH)2 at different potentials [29] as well as the formation of NiOOH with different crystallographic structures [30]. Indeed, a shoulder to the right of the main peak was already observed using EMT at 100 mV/s (Fig. 6a).The behaviour of the Ni-Pt films was studied in the potential range between −0.5 V and 1.5 V vs Ag|AgCl. When the anodic potential was as high as 1.5 V, a large oxidation peak centred around 1 V was observed for all samples (Fig. 7 ). After this large oxidation peak was recorded, no further major oxidation or reduction currents were observed within the entire potential window in the subsequent cycles. However, microscopic analyses confirmed that the Ni-Pt film was still intact and no dissolution had taken place. Therefore, the large oxidation peak observed in the first cycle can be attributed to a passivation of the film’s surface which could not be reversed within the applied potential window and may be related to an irreversible oxidation of Pt  [31,32]. In order to avoid the effect of passivation on the redox reaction Ni(OH)2 ⇌ NiOOH, the anodic limit was set to 0.6 V, resulting in a stabilisation of the redox reaction (Fig. 8 ).The characteristics of the redox reaction (Eq. (2)) resolved by the CVs show dependencies on the film composition, porosity, and the scan rate. Although the redox reaction is determined to involve the formation of Ni hydroxides/oxyhydroxides, the reaction is enhanced by the incorporation of higher amounts of Pt in the films (Fig. 9 ). This may be related to the fact that Pt as an excellent electrocatalyst is able to enhance electrochemical reactions on neighbouring Ni atoms due to synergistic effects. In this way, the reaction enhances while increasing the Pt content while the amount of Ni atoms available on the surface decreases. As a result, there should be a certain Pt content for which the maximum effect is obtained.In absence of porosity, the measured currents are significantly lower. This difference is not as high as might be expected from the difference in surface area between mesoporous and dense films, but correlates well with the fact that the ECSA determined in a previous study did not reveal significant differences between the different film morphologies [17].For the mesoporous Ni-Pt films with Ni contents of 84% and lower (i.e. 61%, 76% and 84% Ni), a square root dependency is appreciated between the peak oxidation (and reduction) current density and the scan rate (Fig. 10 ). The peak current densities were taken from the peaks corresponding to Eq. (2) at approx. 0.4 V (oxidation peak) and 0.3 V (reduction peak). In those cases where the relationship is linear, the redox reaction is diffusion-controlled by the diffusion of OH − in solution (cf. Eq. (2)). The highest activity is observed for 76% Ni, the films containing 84% and 61% Ni follow with similar activities. This observation consolidates the assumption that Pt acts as an electrocatalyst for the reaction, and thus there is a certain Pt content at which the electrochemical activity of the surface towards the observed redox reaction between Ni(OH)2 and NiOOH is the most active.For higher Ni contents, the trend is not linear. Above a certain scan rate of about 80–100 mV/s, the current stagnates, indicating that the kinetics of the redox reaction becomes limited by another factor. The peak reduction currents are generally lower than the corresponding oxidation currents due to the generally wider reduction peaks (cf. Fig. 9). Nevertheless, the trends observed on the reduction part of the reaction are equal to those on the oxidation.The observed characteristics of the Ni-Pt films may be exploited in application as an electrochemical supercapacitor. Although the current densities are not comparable to those reported for Ni-based electrochemical supercapacitors [33,34], an appropriate anodic oxidation treatment may achieve superior performance, taking advantage of the enhancement provided by the addition of Pt.Due to the high Ni contents in all films, a dissolution of Ni takes place in sulfuric acid when anodic potentials are applied. In the CVs of the dense films, this dissolution is predominant in the first anodic sweep (Fig. 11 ). A similar observation had been made during LSV experiments of electrodeposited Ni-P [21]. The observed oxidation is most likely not exclusively the dissolution of Ni (Eq. (3)), which is attributed to the highest oxidation peak for all compositions, but also that of the underlying Cu, which has its standard potential in this potential range and may thus be assigned to the smaller oxidation peak which appears as a shoulder at 0.48 V for 93% Ni and at 0.66 V for 98% Ni [35]. The third, most anodic oxidation peak can be referred to the oxidation or dissolution of Pt [35]. At the very beginning of the anodic sweep, negative currents related to proton reduction (i.e. HER) are clearly observed, suggesting that hydrogen adatoms form on the surface and are subsequently oxidised (Eq. (4)), thus contributing to the oxidation current as well, especially at very low anodic potentials. (3) Ni ⟶ Ni 2 + + 2 e − (4) H ad ⟶ H + + e − Interestingly, the oxidation waves shift to more anodic potentials with increasing Ni content of the films. Hence, the alloying of higher amounts of Pt with Ni increases the surface activity for electrochemical reactions, as seen before, but also accelerates the dissolution of the material in acidic media under anodic polarisation. For the mesoporous films, the dissolution processes were faster and immediately exposed the Cu seed layer which was dissolved simultaneously. This prevented a detailed analysis of the anodic processes taking place in sulfuric acid. In any case, an anodic polarisation of the Ni-Pt films in sulfuric acid leads to dissolution and failure of the films and must be avoided in application. Fig. 12 shows the Nyquist plots for both mesoporous and dense Ni-Pt films after EIS at OCP in acidic media. In general, the impedance increases with the Pt content.The fitting for the EIS spectra was done by simulating an electrical equivalent circuit with the solution resistance Rs, in series with a parallel circuit of the charge-transfer resistance Rct and the double-layer capacitance Cdl, whose behaviour is modelled by a constant phase element (CPE), and represents a simplified Randles circuit (Fig. 13 ) [36]. This model was used for the fitting of the spectra acquired in 0.5 M H2SO4 at OCP of all mesoporous and dense Ni-Pt films.The results of the fitting show that, expectably, the solution resistance is similar in all cases, except for Ni98Pt2 (Table 2 ). The determined charge-transfer resistance is generally higher for the mesoporous films, indicating that those are more resistant to corrosion than the dense films. This is a counter-intuitive result, suggesting that an increase in the surface area does not lower the corrosion resistance of the Ni-Pt thin films. The highest charge-transfer resistances is found for mesoporous Ni76Pt24 and Ni61Pt39. The double-layer capacitance, which depends mainly on the surface area and the surface composition, indicates an increase with the Pt content and is generally higher for mesoporous the films, although this difference is relatively low and may be related to the minor differences in ECSA observed [17].The Bode plots show that the general behaviour of the mesoporous Ni-Pt films is very similar, the main differences lying in the magnitude of both impedance and phase (Fig. 14 ).At high frequencies, the impedance follows a composition dependence, increasing with Ni content. The frequency dependence of the phase angle shows a shift of the maximum phase angle towards lower frequencies with the Pt content, indicating an increase in capacitance.Metikoš-Huković et al. compared sputter-deposited nanocrystalline Ni with electrodeposited Ni by EIS during HER in alkaline media, finding that the electrodeposited Ni films exhibited higher Faradaic resistance and significantly lower double-layer capacitance in the order of 1 µF/cm2 [37]. Krstajić et al. investigated the impedance of Ni at HER in NaOH and found that the impedance decreased when the HER potential was made more negative, reaching impedance values much lower than those reported here, and thus indicate that the impedance at potentials where HER takes place is significantly lower than at corrosion potential. The observed Cdl was in the order of 100 µF/cm2 [38]. Perez et al. conducted a similar study when investigating the oxygen reduction reaction (ORR) on Pt both in NaOH and H2SO4. A minimum of the impedance at a certain ORR potential was observed, again showing that impedance is lower when the material is employed in its function as a catalyst [39]. Juskowiak-Brenska et al. found that for electrodeposited Ni coatings the charge-transfer resistance in acidic media decreased significantly when the coating thickness was decreased, yielding 20 Ω cm2 for thickness of 1 µm [40]. In comparison, the mesoporous Ni-Pt films presented here show superior Rct considering their thickness of 200–300 nm.SEM/EDX analyses after EIS show that the Ni/Pt ratio decreased, i.e. a leaching of Ni has taken place in all Ni-Pt films, and indicate that the electrochemical process observed in EIS is dominated by the dissolution of Ni. A good stability of the Ni-Pt films in absence of external polarisation can therefore not be guaranteed, i.e. the Ni-Pt films may only be able to serve reliably at HER in 0.5 M H2SO4 as long as it is cathodically polarised, i.e. under cathodic protection.Indeed, a parallel experiment which consisted in an incubation of the mesoporous Ni-Pt films in 0.5 M H2SO4 resulted in the leaching of Ni, which was less pronounced when the Pt content was higher. For Ni95Pt5, the leaching of Ni was 23 ± 3% in mass with respect to the total mass of the Ni-Pt film, while this value amounted to 17 ± 3% for both Ni92Pt8 and Ni84Pt16. Alia et al. found that Ni-Pt nanowires did not present any leaching of Ni when the Pt content was above 75% in mass, corresponding to about 47 at% [41].The investigated electrodeposited Ni-Pt is a multifunctional alloy which may serve both in alkaline and acidic environments, provided that a reducing potential is applied when working in acidic conditions.Due to the observed reversible redox reaction Ni(OH)2 ⇌ NiOOH in NaOH, the Ni-Pt thin films with Ni contents of 84% and lower may be used in the charge and discharge of batteries as well as supercapacitors, where electrical charge can be stored in the form of chemical bonds, and very high currents can be retrieved [33,34,42].On the other hand, it was observed that the HER in 0.5 M H2SO4 is stable and reproducible up to 200 LSV cycles, however, anodic polarisation as well as an absence of polarisation at OCP will lead to leaching of Ni into the sulfuric acid. In long-term HER operation at –10 mA/cm2, an increase in potential is recorded, although stabilising over time. The higher the Pt content of the alloy, the lower was the resulting increase in potential. Interestingly, the mesoporous Ni-Pt films seem more resistant to corrosion, the corrosion resistance increasing roughly with the Pt content. The films showing the highest resistance to corrosion in sulfuric acid, and at the same time a very high activity at the redox reaction Ni(OH)2 ⇌ NiOOH, are mesoporous Ni76Pt24 and Ni61Pt39.Since in an application such as a PEMFC, the leaching of Ni will most likely lead to a degradation in the fuel cell performance, two approaches may be used to overcome this phenomenon: using a higher Pt-content alloy to minimise the dealloying effect, or, secondly, a dual-layer structure of the catalyst with a Pt-Ni alloy catalyst layer stacked onto a Pt catalyst layer [43]. Konrad Eiler: Conceptualization, Validation, Formal analysis, Investigation, Data curation, Writing - original draft, Writing - review & editing, Visualization. Halina Krawiec: Conceptualization, Methodology, Resources, Writing - review & editing, Supervision, Funding acquisition. Iryna Kozina: Conceptualization, Validation, Resources. Jordi Sort: Resources, Writing - review & editing, Supervision, Funding acquisition. Eva Pellicer: Conceptualization, Methodology, Validation, Resources, Supervision, Project administration, Funding acquisition.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This work has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 764977, the Generalitat de Catalunya under project 2017-SGR-292 and the Spanish Ministerio de Economía, Industria y Competitividad (MAT 2017-86357-C3-1-R and associated FEDER Project).

Reliability and long-term performance are the key features of modern energy storage and conversion devices. The long-term stability depends entirely on the electrochemical and corrosion properties of device components. Single-phase mesoporous Ni-rich Ni-Pt thin films have shown to be a promising electrocatalyst for hydrogen evolution reaction (HER) and therefore with potential application in fuel cells, water electrolysers or similar devices. The HER activity of the mesoporous Ni-Pt films is reliable and stable in 0.5 M H2SO4 up to 200 linear sweep voltammetry cycles, however leaching of Ni occurs in absence of cathodic polarisation. Long-term electrolysis measurements at a HER current of –10 mA/cm2 reveal an increase in potential over time, which is minimised when the Pt content is increased. In 1 M NaOH, the material is stable up to an applied anodic limit of 1.5 V vs. Ag|AgCl although surface passivation takes place at 1.0 V. If the anodic limit does not exceed 0.6 V vs. Ag|AgCl, a fully reversible redox reaction is observed by cyclic voltammetry, with redox charges increasing with Pt content and scan speed. In addition, significantly higher current densities are recorded for mesoporous films compared to dense counterparts. This charge/discharge behaviour of the redox reaction indicates that the mesoporous Ni-Pt films may as well be used as an electrochemical supercapacitor. As a HER catalyst, the material is safely applicable in alkaline media.

In recent years, the contradiction between environment and human development has intensified with the acceleration of the pace of human development, thus causing various problems [1–3]. Especially in the manufacturing industry, such as steel mill, metallurgic industry, electron plating, printing, leather and pharmaceuticals, various toxic organic compounds and heavy metal ions are discharged from these industries. They have caused serious environmental pollution and destroyed the aquatic environment [4–6]. Colored dyes in natural water bodies, on the one hand, affect the photosynthesis of aquatic organisms and, on the other hand, destroy human senses of scenery. But it has more serious potential threats. Toxic organic compounds have the characteristic of a difficult self-decomposition in natural environments [7,8]. For example, 4-nitrophenol will accumulate in water ecological food chains and, eventually, enter the human body. 4-Nitrophenol and its derivatives can also cause damage to the human central nervous system, liver, kidney and blood as well [9,10]. Not only organic pollutants, but also hexavalent chromium ions in water cause serious harm to the environment and the human body. When hexavalent chromium accumulates in the human body, it will cause anemia, nephritis, neuritis and other diseases. After long-term contact, hexavalent chromium will cause lung cancer and nasopharyngeal carcinoma [11,12]. Therefore, removing toxic organic compounds and heavy metal chromium ions from water is an inevitable and important task for researchers.In the past decades, researchers have studied the reduction/degradation of toxic organic compounds and heavy metals by various methods such as photocatalytic degradation, desorption, membrane flocculation and filtration [13,14]. Among them, bimetallic catalysts have been investigated by researchers due to their high efficiency and unique characteristics. In Han's experiment, the Cu–Fe bimetal was used and tested for RhB [15]. Wen et al. also prepared iron and cerium bimetal oxides and used them to oxidize arsenite [16]. Moghadam et al. also prepared Fe/Zn bimetal nanoparticles to treat petroleum wastewater [17]. Šuligoj et al. prepared silica-supported Cu–Mn and tested it for dye degradation [18]. Our research group also synthesized Cu based catalysts for which a strong catalytic reduction performance was observed towards different pollutants [19–23]. Based on the above consideration, the noble MoSrOS bimetal catalyst was synthesized by a simple method.In this study, a novel wool-coiled molybdenum-based bimetallic sulfur oxide MoSrOS catalyst with a varying amount of Sr was synthesized. The catalysts characterizations were performed by using XRD, XPS, SEM, FTIR, DRS, EIS, and S BET. The catalytic reduction efficiencies were also tested with MO, 4-NP, MB, RhB, and Cr(VI) pollutants. It is expected that the wool-coiled molybdenum-based bimetallic oxysulfide MoSrOS catalyst could be used in wastewater treatment.Under magnetic stirring, 20 mmol of ammonium molybdate ((NH4)6Mo7O24·4H2O) and 10 mmol of strontium nitrate (Sr(NO3)2) were added into 800 mL distilled water. Then, after 20 min, 40 mmol thioacetamide (CH3CSNH2) was dropped into the mixture solution and reacted for 30 min. The mixture solution was heated to 90 °C and reacted for 2 h. In order to determine the effect of strontium nitrate on the catalyst activity, 4.0, 10.0, 20.0, and 40 mmol of strontium nitrate were added to the preparation process. The final solutions are abbreviated as MoSrOS-1, MoSrOS-2, MoSrOS-3 and MoSrOS-4 respectively. The resulting samples were washed, and finally dried by a rotary evaporator.The MoSrOS catalysts were characterized by a Rigaku X-ray diffractometer. The PHI5700 photoelectron spectrometer was used for an XPS analysis. Ultraviolet–visible diffuse reflectance and absorption spectra were performed by the ultraviolet–visible spectrophotometer (TU1901). The field emission scanning electron microscopy (HITACHI SU-8010 microscope) was used for a morphology analysis. The N2 adsorption–desorption was done by an ASAP 2020 porosity and specific surface area analyzer. The electrochemical impedance (EIS) measurement was checked by a SP-300 Biologic Science together with a three electrodes system of platinum plate, Ag/AgCl/KCl, and glassy carbon electrodes in the 0.1M KCl electrolyte solution.MoSrOS catalysts catalytic reduction of pollution was performed as follows. Firstly, 15 mg of sodium borohydride was added in 20 ppm MB aqueous solution. Subsequently, 10 mg of the catalyst was added into the MB aqueous solution. By specified intervals in time, 2 mL mixture was taken and analyzed by the ultraviolet spectrophotometer. Other pollutants were tested according to the same procedure and time intervals adjusted, depending on the activity. Fig. 1 a shows the survey XPS spectrum of MoSrOS-2. The Mo, Sr, S, O, and C elements were obverted in the spectrum, the C1s peak belongs to a trace amount of foreign carbon. Fig. 1b indicates the high resolution Mo 3d XPS spectrum of MoSrOS-2. The peaks for Mo 3d5/2 and Mo 3d3/2 with binding energies of 232.6 and 235.8 eV, respectively, indicate the presence of a Mo6+ state in the MoSrOS-2 solution [24]. The peaks for Mo 3d5/2 and Mo 3d3/2 at 230.3 and 233.5 eV, respectively, are attributed to Mo4+ [24]. Fig. 1c also illustrates the XPS spectra of Sr 3d in the MoSrOS-2 solution. The peaks for Sr 3d5/2 and Sr 3d3/2 located at 133.1 eV and 134.9 eV respectively, with a separation of 1.8 eV, show the presence of Sr2+ in the MoSrOS-2 solution [25]. Moreover, Fig. 1d indicates the S 2p XPS spectra in MoSrOS-2. The peaks at 161.6 and 162.8 eV corresponded to S 2p3/2 and S 2p1/2 orbitals [26], respectively, and belonged to S2−. Fig. 1e illustrates the O1s XPS spectrum in the MoSrOS-2 sample. The peaks shown at 529.8, 530.5, and 531.4 eV corresponded to O Lattice , O Vacancy , and hydroxy oxygen, respectively [27,28]. Fig. 2 demonstrates the XRD diffraction patterns for MoSrOS and the standard SrMoO4 (PDF #85–0809). The XRD peaks for MoSrOS correspond to the structure of tetragonal SrMoO4. The peaks indicated at 27.680°, 29.713°, 33.190°, 38.014°, 45.140°, 47.645°, 51.493° and 55.994° were related to the (112), (004), (200), (211), (204), (220), (116), and (312) crystal planes, respectively. It is observed that the diffraction peak positions of MoSrOS-1, MoSrOS-2, MoSrOS-3 and MoSrOS-4 were similar, which indicated that the strontium nitrate contents did not affect the peak position in the XRD. The peak intensities of the MoSrOS samples decreased with increasing strontium nitrate contents. The average crystal sizes of MoSrOS prepared with different strontium nitrate content were calculated by the Scherrer formula. The average crystal sizes of MoSrOS-1, MoSrOS-2, MoSrOS-3 and MoSrOS-4 are 43.3, 47.9, 46.4 and 39.1 nm, respectively.The morphology of the samples specified from SEM images is indicated in Fig. 3 . The MoSrOS-3 catalyst granules looked as regular spherical-like knitting wool balls (Fig. 3a). As we checked from Fig. 3b, the ball were composed of weeny nanorods with several microns in size. The nanorods are straight and uniform. Fig. 3c–f also indicate the EDX elemental mapping of Mo, O, S, and Sr and an uniform distribution of the elements was observed. Fig. 4 illustrates the FTIR spectra of MoSrOS, prepared with different amounts of strontium nitrate. The broad absorption band situated at 3460 cm−1 and 1637 cm−1 are typical characteristics of O–H stretching vibrations and bending vibration due to the absorbed water on the sample surface, respectively [29]. The peaks located at 866 and 813 cm−1 are due to stretching vibrations of isolated molybdenum-oxygen tetrahedral molybdate ions [30]. The 1390 cm−1 peak is due to S–O stretching vibrations in the MoSrOS catalysts [31]. Moreover, the 941 cm−1 peak is due to the Sr–O stretching vibration [25]. It was observed that the peak intensities at 1390 cm−1 and 813 cm−1 decreased with the increased amount of strontium nitrate. While the peaks at 941 cm−1 increased with the increased amount of strontium nitrate.The nitrogen adsorption and desorption isotherm of the MoSrOS-2 catalyst is shown in Fig. S1a. It was obviously observed that the curves were consistent with the type IV isotherm of a hysteresis loop [32]. The BJH pore size distribution curve is shown in Fig. S1b. It was observed that the pore size distributions of MoSrOS-2 are between 10 nm and 80 nm. The S BET, average pore diameter, and total pore volume are listed in the Table S1. The S BET, average pore diameter, and total pore volume of catalysts were 17.1~20.2 m2/g, 4.18~4.67 nm, and 0.018~0.025 cm3/g, respectively. These three parameters also decrease with the increase of Sr content.The DRS and the absorption spectra of different Sr content catalysts are shown in Fig. 5 a,b. As it is clear from Fig. 5b, the samples have a very long wavelength absorption range including the ultraviolet and infrared regions and show that there were more absorption states or defects in the sample band. The classic Tauc approach, ( α h v ) 1 n = k ( h v − E g ) , (where α = absorbance coefficient, h = Planck constant, k = absorption constant for a direct transition, hν = absorption energy, and E g = band gap) was used to calculate the bandgap of the samples [33,34]. The molybdates of the scheelite type could have a tetragonal structure for which electronic transitions are allowed [35]. The charges present in the valence band (maximum energy level) will be transferred to the minimum energy level after the absorption process [36]. Hence, n = 1/2 was adopted in the equation and α was replaced by the absorption constant k. The band gaps of MoSrOS-1, MoSrOS-2, MoSrOS-3, MoSrOS-4, were obtained as 2.97, 2.39, 2.30, and 2.28 eV, respectively (Fig. 5c). Fig. 5d shows the EIS of MoSrOS prepared with varying amounts of Sr for which a Randles fitting was performed in order to estimate the electron transfer resistance in the electrochemical analysis. The CPE, R1, and R2 symbols were the double layer capacitance and the electrolyte and electron transfer resistances, respectively. Based on the Randles fitting, the electron transfer resistance values of MoSrOS-1, MoSrOS-2, MoSrOS-3, and MoSrOS-4 were 1604 Ω, 470 Ω, 1006 Ω and 2215 Ω, respectively. The MoSrOS-2 sample had the lowest electron transfer resistance, indicating that for this sample the most efficient electron transfer could be achieved. Based on the EIS measurement, MoSrOS-2 was expected to show the highest activity for the reduction reaction. Fig. 6 a,b show the catalytic efficiencies after adding only sodium borohydride and/or the catalyst on 4-NP. The results show that adding only one of them has no significant effect on the 4-NP reduction. After NaBH4 was added, the 4-NP absorption peak at 317 nm shifted to 400 nm, which is indicative for the 4-nitrophenolate ions formation in the alkaline solution [37] by which the color of the solution was changed from light yellow to dark yellow. Fig. 6c also shows the reduction of 4-NP when sodium borohydride and MoSrOS-2 were added simultaneously. As shown from Fig. 6c, 4-NP was completely reduced within 20 min. Simultaneously, a peak representing 4-AP appears at 300 nm, which indicates that 4-NP was reduced to 4-AP. Fig. 6d indicates the reduction activities of all the MoSrOS-based catalysts on 4-NP. It can be seen that MoSrOS-2 and NaBH4 have the best reduction efficiency on 4-NP. According to the kinetic analysis, it was found that the pseudo first-order kinetic rate constant (k) of the 4-NP reduction is shown as follows: MoSrOS-2 (k = 0.25494 min−1) > MoSrOS-3 (k = 0.02688 min−1) > MoSrOS-4 (k = 0.01668 min−1) > MoSrOS-1 (k = 0.00294 min−1). Table S2 shows the performance of our MoSrOS catalyst and its comparisons with data reported in literature on the catalytic reduction of 4-NP. From Table S2, Ni-PVAm/SBA catalyst can perform well, with a short reaction time and a high rate constant, but a catalyst of 120 mg has to be used, which amount is about 12 times higher than we used for MoSrOS. Our MoSrOS at 10 mg can not only save the cost but also improve the separation problem. While comparing with the g-C3N4/CuS catalyst, our kinetic rate constant of MoSrOS-2 catalyst is almost 3.6 times higher than that of the g-C3N4/CuS catalyst at the same dosage. Overall, the advantages of the MoSrOS catalyst are evident in the dark reduction of 4-NP. Fig. 7 a shows the MB reduction with MoSrOS-2 in the presence of NaBH4. The MB (100 mL, 20 ppm) was reduced completely within 6 min in the presence of MoSrOS-2 catalyst and NaBH4. Fig. 7b shows a control experiment in which only NaBH4 was added. When only NaBH4 was added into the MB solution, the curve of MB degradation had almost no change. Similarly, the degradation curve of the MB solution in the presence of the MoSrOS-2 catalyst did not change as well (Fig. 7c). This indicates that MoSrOS or NaBH4 alone had little effect on the MB reduction. Fig. 7d shows the MB reduction efficiencies of the catalysts with varying Sr contents in MoSrOS with NaBH4. It can be seen from the pattern that the reduction efficiencies of catalysts with different Sr content were different and MoSrOS-2 had the best reduction performance on MB. It was found that the pseudo first order kinetic rate constant (k) of the MB reduction were: MoSrOS-2 (0.3654 min−1) > MoSrOS-3 (0.0529 min−1) > MoSrOS-1 (0.0070 min−1) > MoSrOS-4 (0.0031 min−1).The MoSrOS catalysts were also used to evaluate their catalytic reduction activities for RhB and MO, as shown in Figs. S2 and S3, respectively. It can be seen that 100 mL (20 ppm) RhB and 100 mL (50 ppm) of MO aqueous solutions were completely reduced within 20 min, respectively. Figs. S2(b,c) S3(b, c) were the control experiments with only MoSrOS-2 and with only NaBH4 for the reduction of RhB and MO, respectively. It can be seen that RhB and MO didn't reduce by with the catalyst only or by only NaBH4. Figs. S2d and S3d also indicate the performance of the MoSrOS catalysts prepared with different Sr contents towards the reduction of RhB and MO. It can be seen by the reduction effect that MoSrOS-2 was the best catalyst for both RhB and MO dyes. According to the kinetic analysis, it was found that the pseudo first-order kinetic rate constant (k) of the 4-NP reduction is of the following order: MoSrOS-2 (k = 0.25494 min−1) > MoSrOS-3 (k = 0.02688 min−1) > MoSrOS-4 (k = 0.01668 min−1) > MoSrOS-1 (k = 0.00294 min−1), while the k constant of the MO reduction is in the order: MoSrOS-2 (0.1118 min−1) > MoSrOS-3 (0.0187 min−1) > MoSrOS-4 (0.0181 min−1) > MoSrOS-1 (0.0078 min−1). Fig. 8 a,b show the catalytic effect of adding only sodium borohydride or catalyst into the potassium dichromate aqueous solution. The results also show that adding only one of them has no effect on the Cr(VI) reduction. Fig. 8c indicates the reduction of a potassium dichromate solution with the simultaneous addition of MoSrOS-2 and sodium borohydride. As shown in Fig. 8c, the 20 ppm potassium dichromate aqueous solution was completely reduced within 30 min. Fig. 8d also shows the reduction of potassium dichromate in the presence of sodium borohydride by catalysts with different amounts of strontium. The catalytic efficiency with an appropriate molybdenum strontium ratio is far stronger. The kinetic pseudo first-order rate constant (k) of the Cr(VI) reduction were: MoSrOS-2 (0.0856 min−1) > MoSrOS-3 (0.0545 min−1) > MoSrOS-4 (0.0236 min−1) > MoSrOS-1 (0.0145 min−1). In conclusion, MoSrOS-2, MoSrOS-3, MoSrOS-1 and MoSrOS-4 are the catalysts with the best pollutant reduction ability. It can be seen from the reduction rate constants of 4-NP that the reaction rate constants of the catalyst MoSrOS-2 are 9.48, 15.28 and 86.71 times that of the catalyst MoSrOS-3, MoSrOS-1 and MoSrOS-4, respectively. It also has excellent reduction performance for other pollutants.To check the MoSrOS stability, the MoSrOS-2 catalyst was run six times towards the 4-NP reduction and the results are shown in Fig. S4a. At the 6th run, the MoSrOS-2 still maintained to reduce 93.1% 4-NP. After the 6th run, XRD and XPS characterizations of the MoSrOS-2 sample were performed. Fig. S4b shows the XRD diffraction patterns of MoSrOS-2 after 4-NP reduction. It was obvious that the diffraction peak positions of MoSrOS-2 after reduction of 4-NP are similar with those of MoSrOS-2 before the reduction of 4-NP. Fig. S4c shows the Mo3d XPS spectrum of MoSrOS-2 after the 4-NP reduction. The peak positions at 232.6 eV and 235.7 eV are due to Mo6+3d5/2 and Mo6+3d3/2 orbits, respectively [24]. The peaks at 230.3 eV and 233.4 eV are due to Mo4+3d5/2 and Mo4+3d5/2, respectively [24]. Fig. S4d shows the Sr 3d XPS spectrum in the MoSrOS-2 sample after the 4-NP reduction. The peaks at 133.1 eV and 134.9 eV correspond to Sr2+3d5/2 and Sr2+3d3/2, respectively [25]. Fig. S4e indicates the S2p XPS spectrum of MoSrOS-2. The peaks located at 161.6 eV and 162.8 eV originated from S2− 2p3/2 and S2− 2p1/2, respectively [26]. Fig. S4f also illustrates the O1s XPS spectrum in the MoSrOS-2 sample after the 4-NP reduction. The peaks shown at 529.8 eV, 530.5 eV and 531.5 eV correspond to O Lattice , O Vacancy , and hydroxy oxygen, respectively. The result further indicates the MoSrOS-2 catalyst stability after the reduction reaction. For a comparative purpose, other MoSrOS samples were also tested for their stability, as shown in Fig. S5. This table indicates the importance of composition control in achieving a good catalyst.After the exploration of the one-pot synthesis of forming MoSrOS, this compound can be identified as the S-doped and Mo4+-existing SrMoO4 phase. Rare earth-doped SrMoO4 was studied for its photoluminescence [38]. SrMoO4/MoS2 has been studied for its photoreduction of Cr(VI) [39]. The single SrMoO4 phase of MoSrOS performs with an excellent catalytic reduction as a few reports show. The S doping in the SrMoO4 lattice can distort the MoSrOS structure and increase the catalyst activity. The existence of Mo4+ in MoSrOS, which can be viewed as the Mo6+ ion attached with two electrons, enhances its electron transport ability. With this improved transport property, MoSrOS can initiate the catalytic reduction with NaBH4, as described below. Fig. 9 shows the proposed reduction mechanism of pollutants in the presence of the MoSrOS catalyst and NaBH4. Firstly, NaBH4 is hydrolyzed to generate borohydride ions and sodium ions immediately in water. Then, borohydride ions react with water to discharge the hydride ion on the catalyst surface and borate ions are simultaneously released [9,40,41]. Afterwards, pollutants adsorbed on the catalyst particles react with hydrogen ions to convert 4-NP into 4-AP [42]. When electrons and hydrogen ions are transferred from borohydride ions to pollutions, Mo4+ transfers two electrons to become Mo6+ in helping this reaction [43]. The catalyst's functions provide an interaction site for orientation-controlled electrophilic nucleophilic reactions. Mo4+ and Mo6+ in the MoSrOS structure provide the electron hopping Mo4+ → Mo6+, whereas NaBH4 is the hydrogen ions and electrons supply source. Mo4+ and Mo6+ in MoSrOS are also expected to generate the anion vacancy.A series of wool-coiled MoSrOS catalysts were synthesized via a simple hydrothermal method. In this experiment, the contents of strontium nitrate were changed to control the size, shape and physical properties of the catalysts. The catalysts were tested towards the reduction of organic and inorganic pollutants. The reduction test proved that the MoSrOS catalyst with a suitable amount of strontium nitrate has a great reduction activity for MB, RhB, MO, 4-NP, and Cr(VI) with NaBH4 as a reducing agent. In this experiment, 100 mL (20 ppm) of MB, 100 mL (50 ppm) of RhB, 100 mL (50 ppm) of MO, and 100 mL (50 ppm) of Cr(VI) were completely reduced within 6, 20, 40, and 30 min, respectively. The results suggest that MoSrOS based catalysts have great potential in practical applications.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This work was supported by the National Natural Science Foundation of China under the grant No. 31000269, the China Postdoctoral Science Foundation under the grant No. 2018M632562, the Strait Postdoctoral Science Foundation under the grant No. 1323H0005, and Innovation Foundation of Fujian Agriculture and Forestry University under the grant No. CXZX2020129B.The following is the Supplementary data to this article: Multimedia component 1 Multimedia component 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.jsamd.2021.07.008.

Wool-coiled MoSrOS bimetal oxysulfide catalysts were synthesized by a simple method. The catalysts were characterized by different instruments. The MoSrOS reduction activities were also investigated by the reduction of Cr(VI), Rhodamine-B (RhB), Methyl orange (MO), Methylene blue (MB) and 4-nitrophenol (4-NP). The results showed that 100 mL (20 ppm) of 4-NP was completely reduced into 4-aminophenol (4-AP) in 20 min. Moreover, 100 mL (20 ppm) of MB, 100 mL (50 ppm) of RhB, 100 mL (50 ppm) of MO and 100 mL (50 ppm) of Cr(VI) were completely reduced within 6, 20, 40, and 30 min, respectively, by MoSrOS-2 bimetal oxysulfide catalyst. Hence, the wool-coiled MoSrOS bimetal oxysulfide catalysts could be used in the detoxification of toxic organic and inorganic pollutants.

concentration, mol/m3 mass fraction, ppm or wt%activation energy of i reaction in l reactor, kJ/molpre-exponential factor of rate constant for reaction i in l reactor, h−1·(mg/kg)-n space time, hreaction rate constant for reaction i in l reactorheat of reaction i in l reactor, kJgas-law constant, J/(mol·K)the reaction temperature, Kreaction pressure, MPaorder of desulfurization for the 4 lumps, respectivelyorder of desulfurization concerning H2/oil volume ratio for the HDS, respectivelyinhibition factor of H2S for the HDS, respectivelythe superficial velocity of the gas, m/sthe superficial velocity of the liquid, m/sgas–liquid interfacial, cm−1 liquid–solid interfacial area, cm−1 partial pressures, MPagas–liquid mass-transfer coefficient mass-transfer coefficient, cm/sliquid–solid mass-transfer coefficient mass-transfer coefficient, cm/sthe liquid-phase concentrations of H2S at the catalyst surface, mol/cm3 the liquid-phase concentrations of sulfur composition, mol/cm3 sum of square errorsthe number of experimentsthe output variablesthe values calculated by the modelthe experimental dataaxial position along the catalyst bed, cmreaction rate of sulfur components, mol/(kg·h)the apparent rate constanttotal sulfur concentration, mol/m3 CalculatedExperimentalTriA-aromaticDi-aromaticMono-aromaticCycloalkaneParaffinthe kind of componentthe number of reactors, the 1st reactor or the 2nd reactorSulfur compoundsthe first lump in desulfurization at the 1st reactorthe second lump in desulfurization at the 1st reactorthe third lump in desulfurization at the 1st reactorthe third lump in desulfurization at the 2nd reactorthe third lump in desulfurization at the 2nd reactorthe concentration of TriA-aromatic at the l reactorthe concentration of TriA-aromatic at the l reactorthe concentration of Di-aromatic at the l reactorthe concentration of Cycloalkane at the l reactorthe concentration of Paraffin at the l reactorWith the environmental legislation increasing stringent[1], it is imperative to produce ultra-low sulfur content (＜10 ppm) and low polycyclic aromatics hydrocarbons content (PAHs) (＜3 wt%) diesel oil [2,3]. Hydrotreating (HDT) is fundamental to removing sulfur, PAHs, and other impurities [4]. As consequence, improved hydrotreating processes have to be developed.The high and low temperature dual reaction zone RTS process was developed by Sinopec Research Institute of Petroleum Processing Co., Ltd (RIPP) [5]. The RTS technology was characterized by the removal of most sulfides, PAHs, and almost all nitrogen compounds in the first reactor at high temperature and low space velocity. The second reactor was primarily responsible for aromatics hydrogenation and ultra-deep HDS under the condition of low temperature and high space velocity [6,7].For a new diesel hydrotreating unit, the construction and confirmation of kinetic models are necessary for optimizing the operation parameters [8]. The multi-lump kinetic models established in the published journals can describe deep HDS well with sulfur content reduced to 300 ∼ 500 ppm[9–11]. However, The HDS kinetics performance will change after sulfur content lower than 10 ppm and the kinetic behaviors of the reactions had not received considerable attentions in the present literatures.Hydrogen sulphide (H2S), which is produced as a by-product of sulphur removal, was reported to significantly inhibit hydrogenation of sulfur compounds by many scholars [12–14]. The existing researches were carried out by changing the concentration of H2S in the inlet of the reactor to determine the effect of H2S on HDS activity. However, the actual adsorption concentration of H2S in the catalytic active center was not considered in their kinetic model. Therefore, exploring the concentration distribution of H2S on the catalyst surface is crucial for optimizing the HDS kinetic model.Most of the aromatic hydrosaturation (AHS) kinetic models account for one sum reaction and very few were developed to accommodate three lumps of reacting aromatic compounds including mono-, di-, and tri-aromatics [15,16]. To improve the accuracy of kinetic model, Wu et al [17] proposed L-H kinetic models to describe the process of AHS, the aromatics were lumped into four groups including tri-aromatics, di-aromatics, mono-aromatics and non-aromatics. However, with the deep saturation of aromatics, the ring opening of cycloalkanes that occur in hydrotreating to form paraffin cannot be ignored[18]. As a result, a more accurate kinetic model for AHS in the RTS process which have scant attention in previous research in the above condition is deserved to be established to fit the real reaction in the industry.In this work, the main focus is on developing optimal kinetic models for simulating the reactions of HDS, HDN, and AHS of diesel from a refinery in the RTS process, specifically at sulfur content and PAHs content below 10 ppm and 3 wt%, respectively. For HDS, a hydrodesulfurization (HDS) kinetic model was developed based on a fundamental and comprehensive understanding of the different HDS reactivities of various sulfur compounds in middle distillates and the inhibition effects of the coexisting H2S compounds. The calculation of the H2S concentration on the catalyst surface was based on the three-film theory [19]. For deep AHS, the hydrocarbons in diesel are classified into five lumps to make up for the incompleteness of the four-lumped kinetic model. Finally, the developed models were used to predict sulfur and hydrocarbon concentrations for the determination of reaction conditions.A new generation NiMo/γ-Al2O3 catalyst (RS-2100) with strengthened metal-support interaction was used to replace the noble metal catalyst, by inventing the assembly technology of high‐performance active phase and the stabilization technology of well‐dispersion active phase [6]. The diameter (dp), length (dL ), surface area (Sp ), pore volume (Vp ), bulk density (ρb ) and average pore diameter (r) of the NiMo/γ-Al2O3 catalyst were 1.6 mm, 2.0 mm, 176.7 m2/g, 0.28 cm3/g, 0.826 g/cm3, and 6.3 nm, respectively.. Here, Mo was the catalytic active component, Ni was the promotor and γ-Al2O3 was the support. The oxidized catalyst had the following composition: NiO: 4.4 wt%, MoO: 26.6 wt%, P2O5: 7.1 wt%; Al2O3: 61.9 wt%. Table 1  presents the feedstock parameters of the 1st and 2nd reactors in the RTS process, respectively. Feedstock II was the product in the first reactor of the two-stage process by Feedstock I at the following conditions: the temperature = 340 °C, LHSV = 1.5 h−1, H2/oil volume ratio = 300 and pressure = 6.4 MPa. Feedstock I is composed of 80 % straight-run diesel and 20 % light catalytic cycle oil, which are provided by Tsingtao Refining & Chemical Company.The pilot-scale experiments were carried out in a continuous TBR system. A schematic diagram of the apparatus was depicted in Fig. 1 . The whole reaction system consists of the feed section, the reactor section, the product separation section and the collection section. In the reactor unit, the reactor is designed as a tube with a length of 160 cm and an inside diameter of 2.4 cm. The length of the reactor is subdivided into three sections. The first part, having a length of 60 cm, was packed with SiC particles which were used to heat the feedstock and to provide a uniform distribution of gas, liquid and hydrogen saturation of the feed. The second section with a length of 50 cm contained a packing of 60 g catalyst and SiC particles. The last section was packed with SiC particles of nearly the same size as the catalyst in the other reactor sections. At the center line of the reactor, there is a thermo-well containing three thermocouples used to control the axial temperature profile within the reactor. The greatest deviation from the desired temperature value was about 1 °C.In the initial stage of the diesel hydrotreating process, the catalysts were presulfurized by 2nd atmospheric side-stream solution containing 2 % CS2 at 320 °C for 10 h [20]. Following the presulfurization, the diesel feedstock was switched to the kinetic experiments, where the running time on stream of each experiment was 48 h to keep a stable catalyst activity. Experimental group 1 using feedstock I was designed to collect accurate hydrotreating kinetic data for the 1st reactor of the RTS process at a pressure of 6.4 MPa, temperature, LHSV, and H2/oil volume ratio were in the range of 340–360 °C, 1.5–4.5 h−1, 300–800 (NPT, v/v) ratio, respectively. Experimental group 2 was conducted by Feedstock II to obtain reliable hydrotreating kinetic data for the 2nd reactor of the RTS process at a pressure of 6.4 MPa, temperature, LHSV, and H2/oil volume ratio were varied from 330 to 350 °C, 2.25–4.5 h−1, and 300–800 (NPT, v/v), respectively. Fig. 2 describes the general flow pattern of the RTS process.The qualitative and quantitative analyses of sulfur compounds were accomplished with the aid of the Agilent 7890B HP gas chromatographic-sulfur chemiluminescence detector (GC-SCD), using an HP-5 (30 m × 0.32 mm × 0.25 μm) capillary GC column. The total sulfur, nitrogen, and aromatic concentrations were measured by the ASTM D-5453, ASTM D-4629, and ASTM D-2425 methods, respectively.To model the RTS hydrotreating process, sulfur compounds were divided into three lumps according to their reactivity. Hydrocarbons were divided into five categories according to chemical structure. In particular, nitrogen compounds could be rapidly eliminated in the 1st reactor of the RTS process, so that the HDN model could be neglected in the 2nd reactor.The model equations could be established with the following assumptions. (1) The pilot reactors were operated isothermally and isobarically. (2) No catalyst deactivation happened during the hydrogenation reaction. (3) On account of the high H2/oil volume ratio, the fluctuation of hydrogen partial pressure was negligible. (4) Vaporization and condensation could be neglected. (5) The catalyst surface was completely and uniformly wetted. The pilot reactors were operated isothermally and isobarically.No catalyst deactivation happened during the hydrogenation reaction.On account of the high H2/oil volume ratio, the fluctuation of hydrogen partial pressure was negligible.Vaporization and condensation could be neglected.The catalyst surface was completely and uniformly wetted.To estimate a kinetic model that accurately describes the ultra-deep HDS for the RTS process in the sulfur concentration range from more than 10000 ppm to 10 ppm, a clear understanding of the types of sulfur compounds present in diesel feed and hydrotreated product oils and their reactivity is very important [21]. Fig. 3 illustrated sulfur concentration and sulfur-type variation along the catalyst bed in the RTS process. To monitor the fluctuation of sulfur species, the results indicate that diesel feeds contain a large number of individual sulfur compounds which can be classified into three lumps. The GC-SCD chromatograph was first separated into three regions according to the retention time as follows: region 1 with a retention time ranging from 42.0 to 54.5 min, region 2 from 54.5 to 63.4 min, and region 3 from 63.4 to 75.0 min. Lump 1 contained all the sulfur compounds in region 1 except 4-MDBT, which represented the majority of the benzothiophene-type compounds. Lump 2 represented the major dibenzothiophenes without any alkyl substituent at the 4- or 6-position in region 2. Group 3 included 4,6-DMDBT, 2,4,6-TMDBT, and all sulfur compounds in region 3. Accordingly, the different numbers and distributions of groups in diesel would determine their different HDS kinetic behaviors, and the simplified reaction network of lumps is shown in Fig. 4 .Most of the works have studied the inhibiting effect of the hydrogen sulfide on the hydrodesulfurization rates of the S-compounds [22]. However, the actual adsorption concentration of H2S in the catalytic active center was not considered in their kinetic model. Therefore, exploring the concentration distribution of H2S on the catalyst surface is crucial for optimizing the kinetic model. The Langmuir–Hinshelwood reaction equation of each lump for the RTS process can be described as follows: (1) dw S 1 , 1 dt = k 1 , 1 w S 1 , 1 n 1 H 2 O i l a 1 1 + γ 1 w H 2 S # (2) dw S 2 , 1 dt = k 2 , 1 w S 2 , 1 n 2 H 2 O i l a 2 1 + γ 2 w H 2 S # (3) dw S 3 , 1 dt = k 3 , 1 w S 3 , 1 n 3 H 2 O i l a 3 1 + γ 3 w H 2 S # (4) dw S 2 , 2 dt = k 2 , 2 w S 2 , 2 n 4 H 2 O i l a 4 1 + γ 4 w H 2 S # (5) dw S 3 , 2 dt = k 3 , 2 w S 3 , 2 n 5 H 2 O i l a 5 1 + γ 5 w H 2 S # H2S was produced at the sulfided NiMo sites of the bifunctional catalyst, then transported into the liquid phase, and finally, diverted from the liquid phase into the gas phase via mass transfer [23–25]. Therefore, the concentration of H2S did not always increase with the reactor length on the catalyst surface. For reliable estimation and scale-up of pilot plant data, a three-phase reactor model was necessary. The calculation of the H2S concentration on the catalyst surface was based on the three-film theory as shown in Fig. 5 . The differential mass-balance Eqs. (6) - (11) had been obtained for the concentration of H2S at different phases as follows [14,26,27].Gas to liquid interface mass transfer equation: (6) u G RT d P H 2 S G dz + k H 2 S L a L P H 2 S G H - c H 2 S L = 0 # Liquid to gas interface mass transfer equation: (7) u L d P H 2 S G dz - k H 2 S L a L P H 2 S G H - c H 2 S L + k H 2 S S a S c H 2 S L - c H 2 S S = 0 # Solid to liquid interface mass transfer equations: (8) 1 u L k H 2 S S a S c H 2 S L - c H 2 S S = d c H 2 S dz # liquid to solid interface mass transfer equations: (9) k H 2 S S a S c H 2 S L - c H 2 S S = - r H 2 S # (10) r H 2 S = k app c S L 1.5 # Assuming the organic sulfur compound had the same molecular weight as the whole sample, its concentration could be estimated by using the weight fraction wS . (11) c S = ρ M w S # The first reactor of RTS was rapidly denitrifying in a high-temperature environment. The nitrogen molecules were considered as a single lump, and the HDN process was described with the pseudo-first-order kinetic power-law model [17]. The kinetic equation was shown in Eq. (12). (12) - d w N dt = k 4 , 1 w N # The network diagram of the aromatic hydrosaturation reaction in diesel fuel was shown in Eq. (13), which indicated the aromatic hydrosaturation reaction processes ring-by-ring reversibly [28]. To get a better understanding of reaction rules during diesel hydrogenation, the aromatics are classified into five lumps namely tri-aromatics (TriA), di-aromatics (DiA), mono-aromatics (MA), cycloalkanes (CA), and paraffin (PA). Kinetic models for the AHS reaction in a gas/oil system mainly assume that hydrogenation and dehydrogenation reactions occur according to the Langmuir–Hinshelwood mechanisms and the HDA reaction is represented as a first-order reversible reaction [29,30]. In Eqs. (14)-(18) the forward reaction and the backward reaction are assigned in first order for the aromatics to obtain the rate equations as follows: (13) (14) - d w T r i A , l dt = k 5 , l w T r i A , l - k 6 , l w D i A , l # (15) - d w D i A , l dt = - k 5 , l w T r i A , l + k 6 , l w D i A , l + k 7 , l w D i A , l - k 8 , l w M A , l # (16) - d w M A , l dt = k 9 , l w M A , l - k 10 , l w C A , l - k 7 , l w D i A , l + k 8 , l w M A , l # (17) - d w C A , l dt = - k 9 , l w M A , l + k 10 , l w C A , l + k 11 , l w C A , l # (18) - d w P A , l dt = - k 11 , l w C A , l # The constants of reaction rate and the adsorption equilibrium in equations satisfied the Arrhenius equation and Van’t Hoff equation [31], such as Eq. (19) and Eq. (20): (19) k i , l = A i , l exp - Ea i , l RT i = 1 - 11 ; l = 1 , 2 # (20) d l n k i , l dT = - Δ H i , l RT # The model parameters were optimized with a multivariate function Levenberg − Marquardt, and the ordinary differential equations were solved by a variable step-size adaptive Runge-Kutta method [32]. A program utilizing self-coded MATLAB language and the weighted least absolute error terms in Eq. (21) were used to estimate the kinetic parameters in the proposed lump model. (21) SSE = ∑ i = 1 NE ( Γ i , l Cal - Γ i , l Ex ) 2 Fig. 6 shows the distribution of H2S in the multi-phases along with the reactors in the RTS process. The H2S concentration in the liquid phase and solid phase increased initially due to the high reaction rate at the beginning. As the HDS reaction was processed, the H2S concentration in the liquid phase gradually decreased along with the mass transfer gradient prevailing rather than the chemical reaction. After entering the second reactor of the RTS process, the concentration of H2S in the gas–liquid-solid three-phase reached equilibrium and the content of H2S was 25 ppm on the solid surface at this reaction condition. The H2S concentration distribution along the reactor axis in the HDS process is in a reasonable range, which is similar to the reported values for several distillate cuts [8,27,33].To reflect the effect of H2S concentration on the HDS reaction, the H2S concentration distribution equation on the catalyst surface along the axial direction of the reactor was obtained by the BiHill simulating method [34]. Based on that the H2S concentration distribution equation in Eq. (22) was used to estimate the kinetic parameters of HDS in the proposed lump model. (22) w H 2 S = 1841.20 Ã · ( 1 + ( 0.25 z ) 3.49 ) Ã · ( 1 + ( z 1.06 ) 2.02 ) where the relationship between the axial position along the reactor and the residence time is shown in Eq. (23). (23) z = 42 t 0.667 # The distribution equation of H2S adsorption concentration in the first reactor and the equilibrium adsorption value in the second reactor on the catalyst surface are brought into Eqs. (1)-(5), respectively, to obtain Eqs (24)-(28). (24) d w S 1 , 1 dt = k 1 , 1 w S 1 , 1 n 1 ( H 2 Oil ) a 1 1 + γ 1 1841.20 Ã · ( 1 + ( 0.25 42 t 0.667 ) 3.49 ) Ã · ( 1 + ( 42 t 0.667 1.06 ) 2.02 ) (25) d w S 2 , 1 dt = k 2 , 1 w S 2 , 1 n 2 ( H 2 Oil ) a 2 1 + γ 2 1841.20 Ã · ( 1 + ( 0.25 42 t 0.667 ) 3.49 ) Ã · ( 1 + ( 42 t 0.667 1.06 ) 2.02 ) (26) dw S 3 , 1 dt = k 3 , 1 w S 3 , 1 n 3 H 2 O i l a 3 1 + 1841.20 Ã · 1 + 0.25 42 t 0.667 3.49 Ã · 1 + 42 t 0.667 1.06 2.02 (27) dw S 2 , 2 dt = k 2 , 2 w S 2 , 2 n 4 H 2 O i l a 4 1 + 25 γ 4 (28) dw S 3 , 2 dt = k 3 , 2 w S 3 , 2 n 5 H 2 O i l a 5 1 + 25 γ 5 The obtained kinetic parameters for the first and second reactors are presented in Table 2 . It is observed that the orders of reaction vary between 1.0 and 2.0. This is a common observation for lumped kinetics, as lumping large spectral compounds with a broad reactivity distribution can give any apparent reaction order [35]; It has been reported that the hydrotreating reaction of diesel follow half to second order kinetics [20,32]. The order of the inhibition effect of H2S was Lump1 ＞ Lump2＞Lump3, from those results we conclude that the inhibition effect of H2S monotonically decreases with the increasing number of substituents bonded to the thiophene ring. It can be inferred from the activation energy values that MA have stable resonance structures, which makes them the most difficult group of aromatic hydrocarbons to hydrogenate compared to DiA and TriA [29,30]. Although the activation energy and pre-exponential factor are dependent on the feedstock and catalyst utilization, our estimates for HDS, HDN, and AHS are within an acceptable range and are comparable to reported values [36]. However, mutual effects between sulfur compounds, nitrogen compounds and aromatic occur simultaneously in the HDT process were not considered in the proposed kinetic model. Further studies are necessary to include the inhibition of sulfur, nitrogen and aromatics under more severe conditions in the HDT kinetic models.As shown in Fig. 7 , the satisfactory linearity between the logarithm of the reaction rate constants and the reciprocal of temperature in HDS, HDN and AHS indicated that the rate constants fitted the Arrhenius equation well. The temperature dependence of the inhibition factor of nitrogen compounds described in Fig. 7(a) proved that the HDN reaction strongly depends on temperature, which was in accordance with the van’t Hoff equation perfectly [37].The derived kinetic parameters were incorporated into the kinetic equation and the RTS process under experimental circumstances could be simulated. Figs. 8 and 9 illustrate the excellent agreement between experimental and simulation results. The relative error between the predicted and experimental values is less than 10.7 % in the RTS process of HDS, HDN, and AHS, indicating that the kinetic model significantly predicts the effect of HDS, HDN, and AHS.The change in product concentration with the space time for sulfur, hydrocarbon, and nitrogen compounds at different reaction temperatures are presented in Figs. 10-13 . The simulated results proved that kinetic models were suitable for diesel HDS, HDN, and AHS in the RTS process, even if the sulfur and PAHs concentrations are lower than 10 ppm and 3 wt%, respectively. However, it should be noted that additional parameters such as internal diffusion of catalyst, axial and radial dispersion, energy and mass balance, and so on need further investigations when the set of rate laws is applied to larger scale reactors. Fig. 10 showed the effect of the reaction temperature acting on the amount of residual S and N in the 1st reactor. It indicates that N is removed significantly when the temperature reaches 350 °C. The total N content is almost removed at the residence time of 0.4 h. However, the residence time for deep desulfurization is longer than 0.67 h, indicating that it will perhaps be an obvious choice to add an extra reactor to the 1st reactor. Moreover, different lumps of S compounds showed significant variations for HDS. Lump 1 showed the highest reaction rate and improved rapidly with increasing temperature. The reaction rate of lump 2 was low below 350 °C. Catalytic activity improved rapidly at a temperature over 350 °C, resulting in higher conversion. The trend of lump 3 was similar to that of lump 2; the only difference was that conversion increased rapidly when the temperature reached 360 °C. Fig. 11 showed the effect of the reaction temperature acting on the amount of residual Sulfur in the 2nd reactor. It indicates that the relative content of these virtually refractory compounds in total sulfur increases over space time, the reaction rate gets slower and slower. Therefore, it is critical to developing catalysts with a higher activity that can convert these refractory compounds. Figs. 12-13 summarize pilot plant data demonstrating how the aromatic species change as a function of the residence time in the RTS process. For AHS, the two-ring aromatic is converted to the mono-aromatic relatively quickly as shown by a steep decline in PAHs concentration as a function of residence time below 0.2 h. At longer residence times, which represent space velocities of about 1 h−1 or lower, there is very little change due to equilibrium constraints. For mono-ringed aromatic saturation, there is a steady increase in conversion as the residence time is increased, and eventually, the mono-ringed concentration begins to decrease indicating that mono-ring saturation starts to increase as the residence time is increased. These data show that PAHs saturation occurs fairly readily under typical hydrotreating conditions, but the saturation of mono rings aromatics is much more difficult and is aided by higher residence time or improved catalyst kinetic ability.Since the conversion of the mono-aromatic compounds provides the most significant boost in product volume, hydrotreaters with very short residence time will have difficulty achieving higher volume swell due to the much slower rate of saturating the final aromatic ring. These units will require a higher temperature to drive the reaction's kinetic saturation portion. This can have some negative effects on expected cycle time due to the higher start of run temperature and the increased fouling rate associated with higher temperature. Therefore, considering the various factors, the optimal operating conditions in the 1st and 2nd reactors are 350 °C and 340 °C, respectively, at a constant pressure of 6.4 MPa and H2/oil volume ratio of 300 v/v.The hydrogenation efficiency of RTS and conventional HDT process on Feedstock I were tested at different LHSV under the reaction condition of pressure of 6.4 Mpa, H2/oil volume ratio of 300 v/v. The results are summarized in Table 3 . It is clear that the volume space velocity of RTS process is twice that of traditional HDT process at same HDS efficiency and the product color is close to water white in the RTS process. The test results show that RTS technology is a better choice than conventional HDT technology for ultra-deep hydrodesulfurization.The RTS process has been applied in SINOPEC Changling Refining & Chemical Company. The feedstock was composed of 75% SRGO and 25% LCO. The contents of sulfur and PAHs in feedstock and the refined oils during the long-term operation of the device are shown in Fig. 14 and Fig. 15 , respectively. It can be seen that the sulfur contents and PAHs contents of refined oil were less than 10 ppm and 5 wt% during the entire operation period of 1000 days. Therefore, the RTS process with high and low temperature dual reaction zone processes a high HDT efficiency and stability.A three-lumping L-H kinetics model, based on the molecular structures and the retention times of the sulfur compounds in GC-SCD chromatographs, is developed to accurately describes the trend of sulfur content decreasing from more than 10000 ppm to less than 10 ppm. The actual adsorption concentration of H2S in the catalytic active center is calculated by the three-film theory and the inhibiting effect of H2S on the hydrodesulfurization rates of S-compounds was studied. The results show that the inhibition effect of H2S monotonically decreases with the increasing number of substituents bonded to the thiophene ring.Based on proper division method, a new five lumping model for the AHS of diesel is established. The model includes tri-aromatics, di-aromatics, mono-aromatics, cycloalkanes, and paraffin as lumps, which can describe the trend of PAHs concentration decreasing from 19.7 wt% to less than 3.0 wt%.In addition, comparisons between the experimental data and predictions using the lumping kinetic models showed agree well, with average deviation lower than 10.7% at different operating conditions. Therefore, these models can serve as an effective guide for diesel hydrotreatment, and the calculated results indicate that the optimal operating conditions in the 1st and 2nd reactors are 350 °C and 340 °C, respectively, at a constant pressure of 6.4 MPa and H2/oil volume ratio of 300 v/v.We thank the financial support of China Petrochemical Corporation (Sinopec Group, 120051-1) for financial support.

The high and low temperature dual reaction zone RTS (removing trace sulfur) technology is a novel hydrotreating process but lack of in-depth understanding of its kinetics. Three-lump and five-lump kinetic models were developed based on the diesel hydrogenation experimental data which were carried out in the RTS process under various operating conditions to predict the concentrations of ultra-level sulfur and aromatics in hydrotreated oil samples, respectively. Moreover, the inhibiting effect of the hydrogen sulfide (H2S) on the hydrodesulfurization rates of S-compounds has been studied by performing calculations with the actual adsorption concentration of H2S in the catalytic active center. The proposed models were able to reproduce the RTS process with good adjustment and accuracy, and relative deviations below 10.7 % at sulfur content below 10 ppm and polycyclic aromatics hydrocarbons content below 3 wt%. Therefore, these models can serve as an effective guide for diesel hydrotreatment, and the calculated results indicate that the optimal operating conditions in the 1st and 2nd reactors are 350 °C and 340 °C, respectively, at a constant pressure of 6.4 MPa and H2/oil volume ratio of 300 v/v.

As an important intermediate in the manufacture of bulk commodity, such as nylon 6 and nylon 66, the production of cyclohexanone on a commercial scale is very intriguing[1]. Hydrogenation of phenol is thought to be a better alternative compared to the oxidation route of cyclohexane which requires harsh reaction conditions with one-way conversion less than 10% and generates complex byproducts difficult to separate[2–5]. Generally, hydrogenation of phenol involves either a “one-step” or “two-step” process[6]. The “one-step” process, i.e. hydrogenation of phenol straight forward to cyclohexanone, is considered to be a greener route compared with the “two-step” one as it avoids the endothermic back-dehydrogenation step towards cyclohexanol, which consumes additional energy[7]. The ultimate challenge for “one-step” process, which would limit the possibility for industrial application, lies in holding a high selectivity (>95%) when reaching the full conversion for an efficient catalyst[8–10].Supported noble metal catalysts have been verified as very efficient for hydrogenation[8,11–15]. However, the selectivity to a certain substrate such as phenol varies to a great extent among different noble metals[8,11,16–19]. Though the selectivity can be affected by reaction conditions [15,21,23–26] and supports of different properties[9,15,27–29], the general trend can still be informed by studies currently available, which reflects the inner factors controlled by the active components (i.e. noble metals) in nature[8,11,15,16,19,20,30,31]. Generally, the selectivity to cyclohexanone follows the order: Pd > Pt > Rh, Ru and Ni for the most studied active metal components[14,32]. Due to the superior performance, Pd catalysts are the most studied catalysts. They usually give a selectivity in the range of 80–100% at full conversion ranging different supports. Pt catalysts, yet, always afford a moderate selectivity in the range of about 50–85% without modification of the second metal like Cr[16,33,34]. In contrast, Rh, Ru and Ni catalysts often offer a much lower selectivity of only 5–60%[17,35,36]. Therein identifying the fatal factors determining the inner selectivity of noble metal catalysts to phenol hydrogenation, which would guide the direction for catalyst screening[37], is the key to efficient rational catalyst design.A deep investigation to the reaction mechanism is certainly helpful in that the difference along the reaction pathways will give valuable information causing the distinction of such apparent catalytic behaviors as selectivity, activity and deactivation. Previously, it is generally agreed that phenol is hydrogenated to an unsaturated cyclohexenol firstly. Then, quick tautomerization of cyclohexenol leads to the desired product cyclohexanone, which may undergo further hydrogenation to cyclohexanol[11,38,39]. However, it is noteworthy that tautomerization of cyclohexenol most likely occurs under the catalysis of noble metals [40,41] in that the barrier for this process will be much lower than that in the gas phase or under the water mediated condition [14,42] without the catalysis of metals. On the other hand, surface chemistry investigations revealed that the OH bond (in phenol) scission could readily occur on various metals upon phenol adsorption[43–47], which may greatly influence the reaction pathways[40,48,49]. In fact, this OH bond scission will cause tautomerization of phenol before the hydrogenation process[40,50]. To what extent the OH bond scission would affect the reaction pathways and further the product distribution under certain conditions is to be assessed considering that the selectivity to cyclohexanone varies a lot over sorts of noble metals. Whether the factors determining the selectivity remain the same among various noble metals is also to be addressed since recent work suggest that the strong interaction between alkali metal cation and the carbonyl group in cyclohexanone (Mn+-OC-) would also suppress the further hydrogenation of cyclohexanone[51,52].In this work, first-principles studies on the reaction pathways for phenol hydrogenation on noble metals (Pt, Pd and Ru) are performed with OH bond scission under consideration. To support our results, experiments for aqueous hydrogenation of phenol are conducted accordingly on Pt/SiO2, Pd/SiO2 and Ru/SiO2 catalysts. Pt, Pd and R are chosen because of their typical discrepancy of selectivity to cyclohexanone. SiO2 is chosen as the support since it is normally considered as inert in most reactions [22,27,50,53]. The conversion of phenol is deliberately controlled at a low value to capture the initial selectivity evolution at the very beginning. The solvation effects haven’t been included in this theoretical calculation since it has been discussed in detail elsewhere and summed up into four points [8,42,54–60]: different solubility of hydrogen in reaction solution, competitive adsorption between solvent molecules and reactants/products on the active sites of the catalyst, inducing agglomeration of supported metal nano-particles, and non-covalent interactions between solvent molecules and or reactants/products with the solvent. Recently, our work has also revealed the promotion of tautomerization by water [14]. Our main results show that no matter whether the OH bond scission occurs before or during the aromatic hydrogenation process, different reaction pathways always result in the formation of cyclohexanone at first under mild conditions. Cyclohexanol is produced by the over hydrogenation of cyclohexanone. The remarkable low selectivity to cyclohexanone on Ru is ascribed to stabilization for metastable adsorption of cyclohexanone, which enhances the chance for sequential hydrogenation, and the co-catalysis by H2O, which promotes the hydrogenation of the CO bond in cyclohexanone to a great deal.Na2PdCl4 (Pd, 40%), H2PtCl6•xH2O (Pt, ≥37.5%), RuCl3•xH2O (Ru, 40%), fumed SiO2 were used as received from Aladdin Chemistry Co with a specific surface area of 600 m2/g. Active carbon (HPC) was home-made by a “bread leavening” method [61].Catalysts were prepared with the classical impregnation method. In a typical process, Na2PdCl4 precursor was dissolved in deionized water and impregnated on fumed SiO2 to produce a catalyst of 1 wt% Pd over fumed SiO2. After the impregnation, the catalysts were dried at 50 °C overnight and then reduced in a 30 mL/min H2 flow at 80 °C for 1 h with a heating rate of 5 °C/min. The obtained catalyst was denoted as Pd/SiO2. As for Pt and Ru, the reduction temperatures were settled as 80 and 160 °C, respectively.X-ray diffraction (XRD) data were collected on an Ultima TV X-ray diffractometer with Cu Kα radiation (1.54 Å). Transmission electronic microscopy (TEM) measurements were taken on a Hitachi HT-7700 microscope instrument at 100 kV. The detailed results and discussion for the as-prepared catalyst were shown in Figs. S1 and S2 in the supporting information.For atmospheric reaction, in a typical process, 0.585 mmol phenol, 25 mg Pd/SiO2 and 5 mL deionized water were put into a three-neck flask. The reaction was carried out at a temperature of 65 °C with magnetic stirring at a speed of 1000 rpm. Before reaction, a balloon filled with hydrogen was connected to the flask to replace the air. For catalytic reaction operated beyond atmospheric pressure, the hydrogenation was carried out in a 50 mL stainless steel high-pressure batch reactor. Firstly, certain amounts of substrate, catalyst, and solvent were put into the autoclave. Then, the reactor was purged three times with pure H2 to remove residual air. After that, the reactor was charged with H2 of required pressure and the reaction mixture was stirred at 65 °C with magnetic stirring at a speed of 1000 rpm. After reaction, the reactor was cooled to room temperature with water bath and then the remaining H2 was vented. The contents of products and substrate were determined by GC-FID and the products were identified by GC-MS.The calculations are performed by using periodic, spin-polarized DFT as implemented in Vienna ab initio program package (VASP) [62,63]. The electron-ion interactions are described by the projector augmented wave (PAW) method proposed by Blöchl [64] and implemented by Kresse [65]. RPBE functional [66] is used as exchange-correlation functional approximation. A plane wave basis set with an energy cutoff of 400 eV is used. A p (5 × 5) surpercell containing a four-layer slab with 100 atoms was modeled as catalyst and (111) plane is considered as the active surface. For Pd and Ru, only gamma point is used for the Brillouin zone sampling when energy barrier is calculated, while a (2 × 2 × 1) k-point grid is used for Pt and Ir owing to the deep d band. For phenol and cyclohexanone adsorption, a (2 × 2 × 1) k-point grid is used on all the catalysts. The results of adsorption energy for k-point test are listed in Table S1 And (3 × 3 × 1) k-point grids were used for the Brillouin zone sampling for bader charge analysis [67]. The periodic condition is employed along the x and y direction. The vacuum space along the z direction was set to be 13 Å. The upper two layers are allowed to relax during the structure optimization, while the bottom two layers of atoms are fixed. The relaxation is stopped when the force residue on the atom is smaller than 0.02 eV/Å. The transition states are calculated by using the climbing image nudged elastic band (CI-NEB) method [68].The adsorption energy for molecule chemisorption is defined respectively as: Eb = Etot – Eslab – Emol where Etot is the total energy after a molecule adsorption on catalysts; Eslab is the energy of the clean catalyst alone; Emol is the energy of the molecule in the gas phase.The conversion and cyclohexanone selectivity are plotted versus time during the hydrogenation of phenol at 65 °C as shown in Fig. 1 . To minimize the impact of over hydrogenation, mild conditions are chosen and low conversions (<5%) are controlled deliberately. Only cyclohexanone is formed when the conversion is below 1% on Pt/SiO2. Then the selectivity decreases when extending the time, indicating that cyclohexanol may well be obtained by the sequential hydrogenation of cyclohexanone since the main product can well be expected to be cyclohexanol on Pt catalysts given the full conversion of phenol [16,19]. The selectivity remains 100% under the reaction time on Pd/SiO2, which is consistent with many other reports showing good selectivity. However, severe over-hydrogenation occurs on Ru/SiO2 since the beginning, with the selectivity of cyclohexanone expected to be lower than 50% given the conversion higher than 5%. Despite the distinct selectivity, one thing in common is that the path to cyclohexanone is prior even if the path direct to cyclohexanol exists since the main.Product is cyclohexanone at low conversions. Similar results can also be found for noble metals supported on active carbon as seen in Fig. S3. Note that the alkali metal cation induced by the Pd precursor might modify the selectivity to cyclohexanone. The Pd/SiO2 catalysts made with PdCl2 precursor is also tested. The results show that the selectivity to cyclohexanone also keeps to 100% at low conversions of phenol (see Fig. S4 in the supporting information), indicating that the selectivity regulating ability occurs only when the alkali metal cations reach a certain concentration [52].To unravel the hydrogenation mechanism and factors resulting in the selectivity distinction over different metal catalysts, the reaction pathways were conducted in first principle. According to Yoon et al. the first hydrogen (H) addition on Ni (111) and Pt (111) has the highest energy barrier [42]. And Li et al. suggested that the dissociation of phenol group (OH) before phenyl ring hydrogenation would lead to a different product distribution or selectivity [40]. Therein the first hydrogenation step seems to be rather important and is investigated on Pt, Pd, and Ru firstly. Both direct hydrogenation and dissociative hydrogenation (hydrogenation after OH dissociation, i.e. the tautomerization of phenol) are considered.For phenol adsorption, the most stable geometries on Pt and Pd are the same (see top view in Fig. 2 and side view in Fig. S5), which is consistent with previous reports [40], with their adsorption energies being −0.60 and −0.49 eV respectively. However, this is not the case on Ru, with C1, C3, and C5 bonded to the corresponding Ru atoms. The adsorption energy on Ru (−0.71 eV) is much bigger than that on Pt and.Pd, indicating a stronger adsorption.Then the first hydrogenation steps were calculated (see Fig. 3 ). On Pt, the dissociation barrier of OH along the dissociative hydrogenation path is only 0.50 eV, which is in consistent with reported work [22,40,49], while the sequential phenoxy hydrogenation barrier reaches 0.96 eV. Hence the overall barrier along the dissociative hydrogenation path is generally determined by phenoxy hydrogenation. The energy barrier for the direct hydrogenation path is 1.06 eV, much similar to that for the sequential phenoxy hydrogenation. Together with the similar reaction energy for each step of the two paths (about 0.22 eV), it is reasonable to conclude that both the direct and dissociative hydrogenation paths may exist in the reaction under mild conditions. In fact, numerous phenol adsorption studies on Pt (111) [45], Pd (110) [47], and other metal surfaces [43,44,46] suggested that phenol can dissociate to phenoxy at rather low temperatures. When reaction occurs on Pd, the situation is rather similar with those on Pt, except that the overall energy barrier is determined by OH scission (0.69 eV) along the dissociative path, making the trend along this path relatively smaller than that on Pt.However, the situation on Ru is different. OH scission on Ru is rather exothermic [22,69], together with a lowest barrier of only 0.41 eV compared with that on Pt and Pd. Meanwhile, the carbon atom in aromatic ring to be hydrogenated is different on Ru (C3) compared to that on Pt and Pd (C1) along the dissociative path. Moreover, the hydrogenation energy barrier along the dissociative hydrogenation path is still lower than the one along the direct hydrogenation path. So the dominant reaction may well undergo the dissociative hydrogenation path on Ru [22].To figure out the OH scission tendency on different metals, i.e. the favor of the dissociative hydrogenation path before the aromatic hydrogenation process, OH scission on the transition metal catalysts which are commonly used as hydrogenation or hydrodeoxygenation catalysts is calculated. A well-defined BEP relationship is shown in Fig. 4 (a) except for Pt, with a slope of 0.30 and intercept of 0.64. The small slope value (<0.5) indicates an early transition state during the OH scission and a less thermodynamic driving process [70–73], which is consistent with the experimental facts at low temperatures mentioned above. In general, OH dissociates exothermically on Fe, Co and Ru, while endothermically on Ni, Pt, Pd and other noble metals, with most and least favorable scission on Fe and Ni respectively. The BEP relationship suggests that more oxophilic catalysts, such as Fe, Co and Ru, tend to dissociate OH more easily and in turn the reaction more likely follows the dissociative path from the beginning, and in reverse. The reason giving rise to the deviation of Pt contributes to the strong repulsion of ketonic group (CO) after OH scission [40,48]. As shown in the DIS state in Figure.3, CO of phenoxy titled away from the Pt surface, while the O atom bind to the surface for the other metals (see Fig. 4 (c) and (d)). The repulsion increases the dissociation energy of phenol on Pt, making the datum of Pt deviating to the right from the normal line.To figure out the inner reason for the deviation, the electronic structure analysis was applied. As seen in Fig. 5 , the density of states (DOS) of C atom in the ketonic group on metal surfaces along the normal axis, like Pd, Ir and Ru, locate deeper than that on Pt. The strong interaction between the ketonic group (through the aromatic ring) and the metal surface makes sure the anti-bond between the O atom of CO and metal.Surface being lifted above the Fermi level (see violet dashed lines). One would suspect that whether the anaerobism of Pt leads to the repulsion of CO since Pt is noblest among the studied subjects. This possibility can be excluded as we compare the DOS’s for stable and meta-stable adsorption of phenoxy on Pt respectively, as shown in Fig. 5 d and e. At meta-stable state, the O atom in CO binds to the Pt surface and the main bonding DOS of the carbon atom in CO locates between −8 and −4.5 eV. However, the binding breaks when phenoxy adjusts to a stable state and the location of the main bonding DOS of the C atom shifts up forward by 0.5 eV correspondingly. It clearly illustrates the local energy sacrifice of the O–Pt binding so as to achieve a global optimization of phenoxy, while it is not the case on Pd and Ir as shown in Fig. S6.The similarity of the first hydrogenation step on Pt and Pd cannot tell the difference on the selectivity of cyclohexanone. Then further hydrogenation steps are investigated. Note that OH scission may happen in the aromatic hydrogenation process [49]. The OH scission possibility during the aromatic hydrogenation process is valued (see Fig. 4b). Since the OH group titled more away from the surface when the first two H atoms are added to the phenyl ring (see Fig. 4 e and f), the scission barriers are expected to be rather high. Hence the OH dissociation energies were calculated after 0, 3, 4 and 5 H atoms were added to the phenyl ring (see Fig. 4b). Results suggest that the OH scission becomes thermodynamically favorable after the phenyl ring was hydrogenated by 4 H atoms. Combining the BEP relationship, the scission barriers would be lower than 0.65 eV on Pd, so further hydrogenation calculations started after 4 H atoms addition before OH scission. Both direct and dissociative paths were considered on Pt and Pd, while only dissociative path was calculated on Ru since the dissociative path dominates the hydrogenation of phenol on Ru catalyst (see Fig. 4a).The intermediate with 4 H atoms being hydrogenated are denoted as 4H. For 4H, OH remains on Pt and Pd, forming adsorbed cyclohexenol, while it has dissociated at the beginning on Ru. The energy profiles for 4H sequential hydrogenation on these catalysts are shown in Fig. 6 . For 4H hydrogenation on Pt along the dissociative hydrogenation path, OH dissociates at first, overcoming 0.46 eV in the formation of IM3. Chemisorbed cyclohexanone forms when one H was added to IM3 with an energy barrier of 0.77 eV. Then it takes only 0.17 eV for cyclohexanone to desorb. In view of the over hydrogenation fact on Pt, the sequential hydrogenation of.Chemisorbed cyclohexanone is considered. However, the hydrogenation of C1 in cyclohexanone is calculated to be rather difficult, within a barrier up to 1.67 eV, much higher than the first H addition step. Thereafter cyclohexanol is not likely to be formed along this path. Along the direct hydrogenation path, a barrier of only 0.15 eV is required in the fifth hydrogenation step ending with IM1. When IM1 reacts forward continually, it has to desorb first, forming suspended and meta-stable IM2. The path splits into two branches then. One branch leads to the OH scission in IM2, forming the desired cyclohexanone without any barrier. Another is that IM2 overcomes a small energy barrier of 0.30 eV undergoing sequential hydrogenation of C1 to form cyclohexanol. Apparently, the first branch (forming molecular cyclohexanone) dominates along the direct path. The second branch unseals given that cyclohexanone assembled on the surface to a concentration enough to make the equilibrium of the first branch reversed. Comparing with the two paths, the direct path encounters lower barriers and seems more favorable from a kinetic point of view. However, the intrinsic selectivity to cyclohexanone is determined by none of them since both the two paths prefer to form cyclohexanone firstly in principle, while cyclohexanol is formed subsequently from molecular cyclohexanone hydrogenation. Note that all the calculation is performed at a low hydrogen surface coverage without considering hydrogen partial pressure, aiming to simulate the experiment conditions. When increasing the hydrogen partial pressure, OH scission in IM2 is expected to be inhibited or even reversed owing to the high hydrogen coverage or increased reduction potential. As a result, hydrogenation of IM2 or resulted cyclohexanone to cyclohexanol becomes favorable. This helps explain the reason why phenol hydrogenation on Pt gives a high selectivity to cyclohexanone at both low hydrogen pressure and conversion (see Fig. 1a), while the selectivity decreases when hydrogen pressure rises [20,74–76]. Note that 4H and/or IM2 would also desorb and undergo the tautomerization by water to form cyclohexanone [14]. Since the process on the metal surface is more energy favorable, it is not included this time.In the case of the hydrogenation process on Pd, the general trend is similar with that on Pt except for some concrete values. It is essential to claim that the direct path is still not responsible to the formation of cyclohexanol since it is easier to form cyclohexanone through OH scission with a barrier of only 0.11 eV than to form cyclohexanol through sequential hydrogenation with a barrier up to 0.58 eV (this value increase to 0.67 eV when same k-points sampling was used on Pt). The desorption energy of cyclohexanone is 0.20 eV, similar with that on Pt. Compared with Pt, the selectivity difference to cyclohexanone may be well caused by one single reason, namely, the ease of over-hydrogenation. The lower barrier on Pt (0.30 eV) makes cyclohexanone easier to be over hydrogenated than that on Pd (0.58 or even 0.67 eV), thus a lower selectivity to cyclohexanone is well expected. To confirm the hypothesis, cyclohexanone was hydrogenated under the same conditions as phenol (see Fig. 7 a). Results show that cyclohexanone is easy to be converted to cyclohexanol on Pt, while it is not the case on Pd, which testifies the hypothesis above.As discussed above, only dissociative path is considered for phenol hydrogenation on Ru in that phenol prefers to undergo OH scission before the aromatic hydrogenation.Hence the final state is actually IM3 after 4 H atoms are added, which naturally serves as the initial state in Fig. 6c for Ru. The energy barrier (0.70 eV) for the fifth H addition is similar with the case on Pt, forming chemisorbed cyclohexanone. Then the desired product formed after overcoming a desorption energy as large as 0.46 eV. For the over hydrogenation of cyclohexanone to cyclohexanol, two possible paths are studied as discussed above on Pt and Pd. The first path starts from desorbed cyclohexanone. The first H is added at the O atom of CO in cyclohexanone, absorbing 0.84 eV energy. Then the second is added at the C atom of CO (C1), with an energy barrier of 0.44 eV and a great heat release. The second path starts from chemisorbed cyclohexanone. Note that there is a metastable chemisorption state (IM6, endothermic by 0.40 eV) of cyclohexanone on Ru which is not found on Pt and Pd, with both the C and O atoms of CO binding to the surface. Though this state makes no contribution to cyclohexanone over hydrogenation at this moment (in gas phase), it does be responsible when solvent of H2O is considered as you will see below. The origin of this meta-stable state can be attributed to the radicalization of C1 on Ru compared with that on Pt and Pd as seen in Fig. 7b for chemisorbed ketone (IM4). Bader charge analysis reveals that C1 on Ru gains the most electrons. Interestingly, only a barrier of 0.71 eV and an energy gain of 0.10 eV are required when the first H is added to C1 and the second to the O atom of CO with a barrier of 0.76 eV, more favorable than the first path, which is opposite to that on Pt and Pd. Specifically, It is easier to add H atom to the O atom of CO first on Pt and Pd, while it is preferential to C1 first on Ru. The big difference is expected to originate from this radicalization.Comparing the over-hydrogenation situation with that on Pt and Pd, the selectivity to cyclohexanone on Ru is expected to be rather high. However, the experimental results in Fig. 1 are incompatible to this expectation. In fact, the lowest selectivity on Ru was achieved. This inconsistence may well be caused by the promotion of the solvent (H2O), since H2O is proved to be a very efficient co-catalyst for ketone hydrogenation on Ru catalyst, while it is not on Pt and Pd [77–79]. Then the effect of solvent is considered, i.e. a chemisorbed H2O near cyclohexanone was implemented concretely to model the effect of H2O as reported [77,80]. Note that the solvation effect (dielectric constant) of H2O is out of the scope in this work in that it has been investigated in detail elsewhere and its impacts to noble metals are considered to similar [42,54]. Fig. 8 shows the energy profiles along the preferred over-hydrogenation path of cyclohexanone in water. The co-adsorption of cyclohexanone and H2O (IM6’) gives a lower energy (stabilization energy) of 0.25 eV due to the hydrogen bond stabilization between the O of CO and the H of H2O, compared to their separate adsorption, which suggests that H2O induced co-adsorption will increase the coverage of cyclohexanone on the surface of Ru and accordingly the possibility of further hydrogenation of cyclohexanone. Consequently, the following hydrogenation steps are more likely to advance in the presence of H2O nearby [77]. In contrast to the situation of no solvent being considered in Fig. 6. The barrier of H addition to the C1 decreases by 0.15 eV, and reaction energy changes to be rather exothermic. The barrier of the following H addition decreases by surprising 0.56 eV when using the H atom of co-adsorbed H2O as H source rather than H2, forming chemisorbed cyclohexanol. At last, the resulted OH is hydrogenated by the dissociated H atom of H2 with a barrier of 0.94 eV. It is obvious that H2O as a co-catalyst can greatly accelerate the over hydrogenation process of cyclohexanone on Ru. This can be testified by changing the solvent of H2O to ethanol for cyclohexanone hydrogenation as seen in Fig. 7a. When.Hydrogenating in ethanol, the conversion of cyclohexanone on Ru/SiO2 is only 9%, lower than that on Pt/SiO2 (12%). Instead, nearly full conversion is achieved on Ru/SiO2 in H2O, while it is only 68% on Pt/SiO2. The obvious increase of the activity for cyclohexanone hydrogenation and the shifts in activity order verify the co-catalyzing effect of H2O. Note that the regeneration of H2O is the rate-determining step for over hydrogenation in H2O. The high barrier of this step seems to be paradoxical to the promotion effect from H2O. In fact, water molecules exist as clusters or slides on the surface in experimental conditions [81–84]. The positions of H affect the regeneration activity of H2O. Two typical positions of H are then explored in a water cluster as examples [84], as seen in Fig. 8. The barriers and energy profiles for water regeneration both decrease to a certain extent when H is located at the edge and in the center of the water cluster respectively. This indicates the regeneration of water may be much easier in real conditions. In fact, the regeneration of water requires only 0.28 eV when the solvation effect is considered [85].For Pt and Pd, this path is not feasible in that the chemisorbed cyclohexanone desorbs from the metal surface and stays with chemisorbed H2O by hydrogen bond interaction (see Fig. S7). Hence the path with C0 being preferentially hydrogenated is interrupted. For another path started from IM1, though the energy barrier for the formation of cyclohexanone rises from 0.16 to 0.36 eV on Pt, the meta-stable intermediate IM2 vanished, which indicates the inhibition of IM2 to cyclohexanol. In contrast, the transformation from IM1 to cyclohexanone becomes barrierless on Pd (see Fig. S7). To sum up, phenol is more likely to be hydrogenated to cyclohexanol on Ru, while cyclohexanone is preferred on Pt and Pd in the presence of H2O.Then how to make an integrated description to predict the trend of cyclohexanone selectivity for phenol hydrogenation based on the intrinsic character of noble metal active sites, eliminating the co-catalyzing effect of the solvent? Generally, two factors must be taken into account, namely, the intrinsic activity to cyclohexanone hydrogenation and the competitive chemisorption between phenol and cyclohexanone which affects the proportion of the active sites available for cyclohexanone hydrogenation.After carefully examining the structures along the favorable hydrogenation pathways of cyclohexanone, we suspect that the binding strength of H atom (Eb(H)) on metal surface may be vital to the activity. As seen in Fig. 6, the hydrogenation of IM2 (Pd and Pt) or IM5 to cyclohexanol actually involves the hydrogenation of a radical. In this process, the radical is not binding to the surface of the catalysts. In other words, the catalysts will not affect the radical directly. Since the radical is same to all the catalysts, then the barriers for hydrogenation are only determined by the activity of the H atoms on the metal surfaces. The stronger binding of H atom on the metal surface, the harder release of H atom from the metal surface, the less activity for the hydrogenation of the radical. Then the H atom binding energy against the corresponding energy barrier of cyclohexanol formation on various noble metal catalysts is plotted as shown in Fig. 9 a. As expected, a clear linear relationship was obtained, indicating weak binding of H atom on the surface is favorable for cyclohexanone hydrogenation. Since Eb(H) is charged by the d band center (see Fig. 9b), which has been proved by previous work [86], the intrinsic activity for cyclohexanone hydrogenation is then actually controlled by the d band center of the noble metal catalyst as shown in Fig. 9c. According to the Sabatier principle, the activity is roughly proportional to the binding energy of the reactant or intermediate for each branch. Then the activity should be proportional to 1/Ea owing to the linear relationship between Ea and Eb(H).As for the second aspect, we simply assume the proportion of the active sites available for cyclohexanone hydrogenation is proportional to the binding energy ratio (Eb (one/pl)) of cyclohexanone to phenol. Combining with the first aspect, we propose the value of Eb (one/pl)/Ea should be a reasonable rough descriptor to predict the trend of cyclohexanone selectivity for phenol hydrogenation. To testify our hypothesis, Eb (ketone/phenol)/Ea is correlated to the yield ratio of cyclohexanol to cyclohexanone (Rationol/one) when ethanol was used as solvent. As shown in Fig. 9d, Rationol/one increases along with Eb (one/pl)/Ea, meaning Eb (ketone/phenol)/Ea can be used as a reasonable rough descriptor for rational catalyst design about phenol selective hydrogenation. And with this relationship, we can predict that the value of Eb (one/pl)/Ea should be no higher than 0.6, which is the value for Pd, if high cyclohexanone selectivity were expected to be achieved on this catalyst. It is noteworthy that we failed to correlate Eb (one/pl) to a single electronic structure descriptor intending to build a full-electronic-structure descriptor in that the d band center failed to describe the binding strength trend of phenol and cyclohexanone with multiple adsorption sites (see Fig. S8).Both experimental and theoretical studies suggest that phenol is intrinsically inclined to be hydrogenated to cyclohexanone at first among the studied noble catalysts, independent of the direct or dissociative paths. Cyclohexanol is mainly produced by over hydrogenation. Generally, two main reasons are responsible for the over hydrogenation for most of the studied noble metal catalysts, i.e. the discrepancy of the over hydrogenation barrier and the competitive chemisorption between phenol and cyclohexanone, based on which a quantitative descriptor, Eb (one/pl)/Ea, is firstly proposed to theoretically evaluate and predict the selectivity to cyclohexanone for rational catalyst design. For a special case, H2O can serve as an efficient co-catalyst for phenol over hydrogenation on Ru, which results in the extremely low cyclohexanone selectivity in aqueous solution.The authors declare no competing financial interest.We sincerely appreciate the fruitful discussion with Prof. Qingfeng Ge and Dr. Quanxi Zhu.This work was supported by Financial support from the National Natural Science Foundation of China (21908189, 21872121), the National Key R&D Program of China (2016YFA0202900), the Key Program supported by the Natural Science Foundation of Zhejiang Province, China (LZ18B060002), and the Key R&D Project of Zhejiang Province (2020C01133).The following is the Supplementary data to this article: Multimedia component 1 Multimedia component 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.nanoms.2020.11.002.

Selective hydrogenation of phenol to cyclohexanone is intriguing in chemical industry. Though a few catalysts with promising performances have been developed in recent years, the basic principle for catalyst design is still missing owing to the unclear catalytic mechanism. This work tries to unravel the mechanism of phenol hydrogenation and the reasons causing the selectivity discrepancy on noble metal catalysts under mild conditions. Results show that different reaction pathways always firstly converge to the formation of cyclohexanone under mild conditions. The selectivity discrepancy mainly depends on the activity for cyclohexanone sequential hydrogenation, in which two factors are found to be responsible, i.e. the hydrogenation energy barrier and the competitive chemisorption between phenol and cyclohexanone, if the specific co-catalyzing effect of H2O on Ru is not considered. Based on the above results, a quantitative descriptor, Eb(one/pl)/Ea, in which Ea can be further correlated to the d band center of the noble metal catalyst, is proposed by the first time to roughly evaluate and predict the selectivity to cyclohexanone for catalyst screening.

Data will be made available on request.To cope with the climate crisis we are facing today due to greenhouse gases from fossil fuel use, the transition to renewable energy is essential. Hydrogen is a key factor in increasing the share of renewable energy because it is an efficient, clean energy carrier: it can be produced from renewable energy and later be used as a fuel without greenhouse gas emissions. As hydrogen is a gas at room temperature, finding a means to more efficiently store hydrogen in a solid-state hydrogen storage material with enhanced accessibility remains a challenge.TiFe-based metal hydrides are one type of promising solid-state hydrogen storage materials [1,2]. They have a reasonable hydrogen storage capacity of 1.9 wt% when a dihydride (TiFeH∼2) is formed [3], and are cost effective for scalable deployment. However, initial hydrogenation of TiFe-based alloys—the so-called activation process—is difficult [4–6]. In part, this is thought to be due to the formation of native protective surface oxides or hydroxides by reacting with air, which prohibit successive reaction with hydrogen. A number of studies have been conducted to mitigate such difficulty and have found that the activation must typically be achieved through repeated heat treatment and/or hydrogen absorption/desorption cycling that induces surface cracking due to volume changes. For example, Reilly et al. activated TiFe by heating to 673 K in vacuum and cooling to room temperature in a hydrogen atmosphere; this cycle was repeated until hydrogenation occurred [3,7]. Besides time- and energy-intensive thermal treatment processes, research efforts have been made primarily along two lines [2] to achieve activation at room temperature: mechanical treatment and alloying. Mechanical treatment is usually done by high-pressure torsion [8,9], ball-milling [10,11] and rolling [12]. These mechanical treatments produce nanostructured TiFe that remains active even after prolonged exposure to air, probably due to enhanced hydrogen diffusion both at the surface and inside the bulk. In the alloying strategy, Mn was first used to improve the activation properties [13]. Other elements such as Cr, Zr, Ni, etc. are also known to be effective for room-temperature activation [14–16].Recently, Kobayashi et al. reported that nano-sized TiFe powders were difficult to activate because the oxide layer is more stable than that of bulk [17], which contrasts well with some previous reports showing comparatively easy activation of TiFe powders composed of nano-sized crystallites [10]. Their studies highlight that activation is closely linked to the characteristic of surface oxide layer, and accordingly, identifying the surface catalysts that may promote activation has been an issue. Many studies have argued that Fe clusters serve as catalysts. Fe clusters on the activated surface were first observed by Bläsius et al. [7], and Schlapbach et al. further observed the formation of Fe and TiO2 after the heating process for activation [18]. Selective oxidation of Ti at the surface supported the idea of metallic Fe catalyzing the dissociation of hydrogen molecules on the surface [19–21]. Surface characterization utilizing X-ray photoelectron spectroscopy (XPS) also revealed the presence of metallic Fe after vacuum heat treatment [5], after temperature cycling between 223 and 573 K under 3 MPa of hydrogen [22], or after repeated absorption/desorption at 673 K [23]. Nevertheless, there have also been reports that other types of compounds instead contribute to activation. Hiebl et al. [24] claimed that Ti2FeO x may help absorb hydrogen, and Mintz et al. [25] also found it during heat treatment of TiFe. Schober pointed out that the oxidation of TiFe mainly produces TiFe2 instead of Fe clusters and argued that the catalyst for activation should be oxides such as FeTiO x , Ti n O2 n -1, etc. [26]. Therefore, identifying how Ti and Fe are distributed in the oxide layer is crucial to understanding initial activation mechanisms and promoting the utilization of TiFe-based alloys as hydrogen storage materials.In addition to the attempts to activate pure TiFe, efforts have been devoted to elucidating the role of alloying elements in modifying the surface oxide layer since alloying elements critically affect the activation kinetics. Seiler et al. observed more Fe particles on the surface of a Ti(Fe, Mn) alloy than pure TiFe by means of Auger electron spectroscopy, XPS and magnetization measurements. During 25 hydrogenation (5 MPa) and dehydrogenation (0.1 MPa) cycles at room temperature, Fe segregation was found to be much more pronounced on the surface of Ti(Fe, Mn) than on TiFe [27]. This fact was used to explain the relatively easier activation of Mn-doped TiFe. Likewise, alloying elements such as Zr, Cr, Mn, and V, have led to different activation kinetics [16,28,29]. According to Park et al., TiFe alloys in which Fe is partially substituted with Mn or Cr were activated without heat treatment, whereas TiFe alloyed with Ni, Co, or Cu were not activated under the same conditions [30]. Part of the reason was the existence of an easy-to-activate secondary phase formed by Mn or Cr [31–33]; however, other studies reported that even single-phase TiFe can exhibit improved activation kinetics depending on the composition [29,34]. Kim et al. [35] attributed the varying performance of alloying elements to their oxidizing power compared to that of Fe. Based on the hypothesis that preferential sites for hydrogen chemisorption on TiFe surfaces are Fe clusters formed during heat treatment, the retarded activation of alloys mixed with Cu and Ni powders was understood in terms of the less stable Cu and Ni oxides versus Fe oxides. Conversely, mixing with Al, Si, Mn, and Mg powders led to easier activation because they are stronger oxide formers than Fe and they sacrificially form oxides, protecting Fe clusters from oxygen contamination.The effect of alloying elements is not limited to modification of the surface oxide layer. It was well established, both by experiments and calculations, that the thermodynamics of hydride formation is also affected by the alloying elements [28,36–38]. Interestingly, the dissolution of hydrogen in pure TiFe is endothermic and the solubility of hydrogen at room temperature is very low [37,39]. This is an uncommon characteristic in interstitial metal hydrides and indicates that there is an energy penalty for TiFeH formation prior to further hydrogenation, which can make the initial stages of hydrogenation very difficult. Most alloying elements sitting on the Fe sublattice tend to stabilize the monohydride, TiFeH—in other words, the formation enthalpy of TiFeH from TiFe and H2 becomes more exothermic upon alloying [40]. This effectively reduces or eliminates the thermodynamic penalty, thereby accelerating the activation process.Systematic study of the activation kinetics of TiFe alloys alongside analysis of the oxide layer and the thermodynamics of the hydride formation will help to further understand the activation mechanism. Herein, we characterized TiFe mixed with alloying elements with different electron affinities. In each alloy, 10 at% Fe was replaced by M (M = V, Cr, Co and Ni), giving a general formula TiFe1− x M x (x = 0.1). The alloying elements V and Cr have higher oxide stability than Fe, whereas Co and Ni have lower oxide stability; pure TiFe was also investigated as a reference. Activation experiments were conducted for each alloy, and the oxidation states of the elements in the oxide layer were analyzed by XPS. The spatial distribution of the atoms in the oxide layer was also observed on an atomic scale using atom probe tomography (APT). These surface analysis results were combined with density functional theory calculations, whereby the energetics of hydride nucleation were investigated and correlated with the activation characteristics.As starting materials, Ti (99.995 % purity, RND Korea), Fe (99.9 % purity, RND Korea), V (99.99 % purity, KRT), Cr (99.95 % purity, RND Korea), Co (99.95 % purity, KRT) and Ni (99.99 % purity, KRT) were used. Alloys were arc-melted in an Ar atmosphere. All samples weighed approximately 20 g (±0.005 g). Ti getter was used to minimize the absorption of oxygen during melting. Alloy buttons were turned over five times to ensure compositional homogeneity. The weight loss of the samples after the arc-melting was <1 %. Heat treatment was performed after arc-melting to eliminate the secondary phases. The samples were annealed at 1273 K for three weeks in vacuum-sealed quartz tubes and then water quenched right after annealing.A home-made Sieverts-type apparatus was used to carry out the first-time hydrogenation, i.e., activation. For the activation process, a stainless steel reactor (2 cm3) was charged with approximately 300 mg of the samples (ground to a particle size of 2–3 mm in air). The reactor was evacuated to a rough vacuum of 0.1 Pa for 60 min at 323 K (or 473 K). After evacuation, 5 MPa H2 (99.9999 % purity) was applied to the sample at 323 K.X-ray diffraction (XRD, Bruker D8 Advance X-ray diffractometer, Cu Kα radiation, λ = 1.5418 Å) was employed for the phase analysis. About 1 g of sample was hand-crushed, sieved under 100 μm and loaded. The 2θ range for the diffraction studies was from 20 to 115°, with a step size of 0.03° and a duration time of 15 s step−1. The phase analysis was carried out using the Rietveld refinement method (TOPAS software ver. 5, Bruker AXS GmbH).The microstructure of the alloys was characterized by scanning electron microscope (SEM, INSPECT F, FEI Company). The chemical composition of the annealed samples was analyzed using energy dispersive X-ray spectroscopy (EDS). The specimens were mounted and mechanically polished with 1 μm diamond compound for the SEM observation.The oxidation states of the elements at the surface layer were characterized by XPS (NEXSA, Thermo Fisher Scientific) depth profiling. The X-ray source is a microfocus monochromatic Al Kα (1486.6 eV) with a spot size of 100 μm × 100 μm. The specimens were mechanically polished and handled in air before the XPS measurement. For depth profiling, sputter etching of the alloy surface was performed using Ar+ at 1 kV under the condition optimized to the sputter rate of 4.2 nm min−1 for SiO2. The depth profile passing energy was 58.70 eV and the detection limit was 0.5 at%.Needle-shaped specimens for APT analysis were prepared using a dual-beam focused ion beam (FIB, Nova NanoLab 600, FEI Company) with the site-specific “lift-out” method. The specimens were analyzed in a local electrode atom probe (LEAP 4000 X HR, AMETEK) by applying a 10 ps and 50 pJ of ultraviolet laser pulses (λ = 355 nm) with a pulse repetition rate of 200 kHz. The detection rate is 3 ions per 100 pulses on average. The base temperature is 54 K, and the ion flight path is 382 mm. The detection efficiency is limited to 38 % due to the open area of the microchannel plate. The APT data were processed using the commercial software package (IVAS 3.8.8, CAMECA). To visualize the distribution of the elements, raw files containing the number of counts of each element in the 1 nm3 cubic grid (3D grid-voxel) were used.To simulate the energy change upon TiFeH nucleation in pure and alloyed TiFe, density functional theory (DFT) calculations were performed using the Vienna Ab initio Simulation Package (VASP) [41]. The exchange–correlation energies were evaluated within the Perdew–Burke–Ernzerhof generalized gradient approximation [42], and a projector-augmented wave (PAW) potential [43] was used. The PAW potentials employed are H, Ti_pv, V_pv, Cr_pv, Fe, Co, and Ni in the VASP pseudopotential library.A 4 × 4 × 4 supercell of TiFe unit cell containing 128 metal atoms and a 6 × 6 × 6 supercell containing 432 metal atoms were employed to compare the nuclei-size dependent total energy of the systems. For the 4 × 4 × 4 supercell, 3 × 3 × 3 and 4 × 4 × 4 k-point grids were used for the structure optimization and the final total energy calculation, respectively. For the 6 × 6 × 6 supercell, a 2 × 2 × 2 k-point grid was used for all processes. The Methfessel–Paxton smearing scheme [44] with a smearing energy of 0.2 eV was applied with a plane-wave cutoff energy of 500 eV. Structure optimization was performed until the energy of the system converged within < 10−4 eV, and we confirmed that the same trends for TiFe1− x M x alloys were reproduced when computed using tighter energy and force convergence criteria of < 10−5 eV and < 0.005 eV/Å, respectively, in the 4 × 4 × 4 supercell. Our calculations were done without spin polarization and Hubbard U correction, which best matches with the enthalpy of reaction from TiFe + ½ H2 to TiFeH as measured by Guo et al. [45]. As a reference state for hydrogen, an isolated H2 molecule was simulated in a cubic cell of 30 Å3 in size. The optimized H–H distance was 0.750 Å.When the composition of TiFe alloy is Ti-rich as in the case of TiFe1− x M x , the DFT study by Ko et al. [37] predicted that all M elements (M = V, Cr, Co and Ni) prefer to occupy the Fe site. Therefore, we consider configurations in which only the Fe atoms are replaced by M atoms. Four and eight M atoms were introduced in the 4 × 4 × 4 supercell, corresponding to x = 0.0625 and 0.125, respectively. The sublattice composed of the M atoms takes either a face-centered cubic (4 M atoms) or a diamond (8 M atoms) structure where the M atoms are equally spaced by 8.3 and 5.1 Å, respectively, before relaxation. On the other hand, 24 M atoms were introduced in the 6 × 6 × 6 supercell (x = 0.111), where they form multiple {111} planes. The distance between the M atoms is 7.3 Å on {111} planes, and the M-substituted {111} planes are spaced by 5.1 Å along the [111] directions. The initial positions of the M atoms are summarized in Table S1. Within the supercell, hydrogen atoms are arranged to have a local TiFeH structure, and the cell parameters and atomic positions were optimized. Zero-point energy was not included because we are interested only in the relative energy change before and after alloying in order to extract trends.The motivation of this study is to understand the mechanistic effects on initial hydrogenation of Fe substitution with other elements. To exclude the effect of secondary phases such as TiFe2− y M y , which strongly changes the activation kinetics [15,34,46], it is important to make a single-phase alloy. For this purpose, each sample was heat treated at 1273 K for three weeks. XRD and SEM analyses were performed to identify the phase composition. Fig. 1 shows the XRD results for the five samples. Diffraction peaks from the typical secondary phases such as TiFe2− y M y or Ti4Fe2O z were very small, and the proportion of the major phase TiFe0.9M0.1 was higher than 96 wt% in all samples. The phase composition and the lattice parameters of TiFe0.9M0.1 are summarized in Table S2. In accordance with the XRD analysis result, SEM images in Fig. S1 shows that regions having different contrast, i.e., secondary phases, are very minor. The overall sample composition in the entire area shown in Fig. S1 was obtained by EDS and listed in Table 1 , confirming that the compositions are close to the target compositions and that M alloying elements preferentially replace Fe atoms.The initial hydrogenation of these samples was tested at a hydrogen pressure of 5 MPa at 323 K. Two sets of samples were prepared by vacuum-treating at 323 or 473 K for 1 h before applying hydrogen. The time variation of the absorbed hydrogen of these samples is presented in Fig. 2 . The performance of vacuum-treated samples at 323 K is presented in Fig. 2a, indicating that samples with M = V and Cr are activated within a few hours, whereas those with M = Fe, Co, and Ni remain inactivated even after 400 h of hydrogen exposure. This result is similar to the study by Kim et al. [35] showing that mixing with metal powders that form more stable oxides than iron oxide promoted activation. One important difference is that we directly alloyed TiFe with other metal elements, while the metal powders were simply mixed with TiFe in their study; nevertheless, the trend appears the same. However, since the activation condition in Fig. 2a could not distinguish among the kinetics of the M = Fe, Co, and Ni samples, we increased the temperature of the vacuum treatment from 323 to 473 K and performed another set of activation experiments. High-temperature vacuum treatment generally promotes activation [7] by removing surface impurities, especially oxygen [47]. Fig. 2b shows that the heat treatment at higher temperature did not significantly change the kinetics for M = Fe, but it did improve the kinetics for M = Co and Ni (M = V and Cr retained the faster activation kinetics observed for the lower-temperature treatment). The result is similar to the report by Lee et al. [40] showing that single-phase TiFe1− x M x (M = Co, Ni, and Al) is more easily activated than pure TiFe. We note that slight fluctuations in the absorbed hydrogen, including occasional drops below zero, are artifacts of temperature fluctuations as only the temperature of the reactor containing the alloy was controlled, while other parts were exposed to the varying temperature of the room.To probe the effect of alloying on the chemical composition of the native surface oxide layer, we first identified the oxidation states of its constituent elements by XPS analysis. The samples were polished under ambient conditions prior to the XPS experiment, and the XPS depth-profiling was conducted without pre-cleaning so as not to lose the outermost oxide layer. Fig. 3 illustrates the XPS depth profile results. To better visualize the depth dependence of oxidation states, 36 profiles were plotted in a single graph with color-coded intensities. Depth in the y-axis value, i.e., the distance from the surface, was estimated based on the sputtering rate of SiO2 under the same condition. The outermost surface is located at y = 0 nm, and the y-value decreases toward the oxide/alloy interface. Among the five alloys, M = V, Fe, and Ni were chosen as representative cases for analysis and the results are presented in Fig. 3a, b, and c, respectively. In the outermost oxide layer, all metal elements exist in oxidized states as follows: Ti 2p3/2 peak at 459 eV from Ti(IV) oxide, Fe 2p3/2 peak at 710 eV from Fe(III) oxide, and V 2p3/2 peak at 515–516 eV from V(IV) or V(III) oxide [48–50]. In contrast, no distinct Ni(II) peak is observed in the oxide layer based on the Ni binding energy (Fig. 3c). It seems that only a minor amount of nickel oxide, if any, is formed. In terms of the penetration depth and peak height, the degree of oxidation is more pronounced in the order of Ti(IV) > V(IV or III) > Fe(III) > Ni(0), which is consistent with the oxygen affinity of the metallic elements [51]. Moving towards the bulk alloy (negative y-values in Fig. 3), the peaks shift to lower binding energies, indicating the transition to a metallic state. The Ti 2p3/2 peak at 454.2 eV and Fe 2p3/2 peak at 707.1 eV can be assigned to Ti(0) and Fe(0), respectively [52]. The alloying elements also become metallic: V 2p3/2 peak at 512.6 eV and Ni 2p3/2 peak at 853.1 eV correspond to V(0) and Ni(0) state, respectively [48–50] (the next strongest peaks on the graphs correspond to 2p1/2 peaks of the respective metal element). One conclusion from the XPS analysis is that the oxidation of V is more prominent than Fe, which may indicate sacrificial oxidation of V and protection of Fe from oxidation, as discussed previously in the context of Mn-substituted TiFe [22,27,53]. On the other hand, Ni is apparently more inert to oxidation than Fe despite the better activation kinetics of TiFe0.9Ni0.1 over pure TiFe, implying the elemental oxyphilic tendency alone, as derived from the XPS results, cannot fully explain the roles of the alloying elements in the TiFe activation mechanism.Beyond its overall chemical composition, the atomistic distribution of elements in the oxide layer can also be varied upon introduction of the alloying species. To visualize this distribution within the surface oxide, atom probe tomography was conducted. For the APT analysis, each alloy was mechanically polished under ambient conditions as was done for the XPS analysis. The surface was coated with Ni (approximately 100 nm thickness) using an e-beam evaporator to protect the surface during FIB milling, and finally a needle-shaped specimen was prepared by FIB-SEM as shown in Fig. S2. Fig. 4 displays the APT reconstruction of the five samples, resembling the shape of the tips in the lower panel of Fig. S2. In Fig. 4, the green spheres at the top are Ni atoms from the Ni-cap coating on the specimen, and the blue ones are TiO or TiO2 species; the maps indicate that the oxide layer is mainly composed of Ti oxides. Since the reconstruction images are drawn at the same scale for five samples, the thickness of the oxide layer can be directly compared, based on the z-values approximately perpendicular to the surface. Fig. 5 shows a proximity histogram measuring the depth-dependent atomic percent of elements, computed by taking the planar average of the three-dimensional (3D) information from Fig. 4. Here, zero distance refers to the location of the isoconcentration surface of 30 at% Ni, with negative values penetrating into the bulk interior; note that this distance definition is different from the z-value in Fig. 4. The difference among the samples is quite noticeable. First, in pure TiFe (Fig. 5c), the Ti:Fe ratio is almost 1:1 both in the oxide layer and in the bulk matrix (distance below ∼  −10 nm). Only the increased oxygen concentration distinguishes the oxide layer. On the contrary, the relative concentration between Ti and Fe deviates for each of the alloyed specimens. In the case of M = Co and Ni (Fig. 5d and e), Ti-oxide with an overall stoichiometry of TiO∼1 (a mixture of TiO2, TiO, and Ti) is formed at the outermost part of the oxide layer (distance between −3 and 0 nm). Below this Ti-oxide layer towards the bulk interior, an Fe-rich region with significantly lower oxygen concentration appears, which indicates partial oxidation of Fe. Notably, the co-existence or segregation of Co, which is more inert element than Fe, is suppressed in the outermost Ti-oxide layer. Ni appears to behave similarly to Co, although the overlap with the Ni coating hinders unambiguous interpretation of the histogram profile. Nevertheless, the Ni concentration slightly increased right below the Ti-oxide layer (3–5 nm from the surface), which is consistently found in the XPS result in Fig. 3c indicating that Ni remained in the matrix not participating in the oxide formation.Different tendencies are seen for M = V and Cr, for which the formation of Ti-oxide layer is much less pronounced. For M = V, an Fe-rich region stretches from the outermost surface to distance =  −5 nm, and the concentration of Ti and O is quite low even in the Ti-oxide layer. For M = Cr, although the Ti-oxide layer develops as in the M = Co and Ni samples, the concentrations of Ti and O are lower than those for M = Co and Ni, and the layer is quite thin. If we define an oxygen threshold of 20 at% or more as the boundary for the Ti-oxide layer, its thickness follows the order Fe (9 nm) > Ni (3.1 nm) > Co (2.6 nm) > Cr (1 nm) > V (0 nm). This trend agrees very well with the activation behavior in Fig. 3: the thinner the Ti-oxide layer, the easier the activation.In addition to the depth-profiles in Fig. 5 integrated over the xy-plane, we use the APT 3D reconstruction data in Fig. 4 to investigate the atomic distribution within the xy-plane. To determine appropriate z-values for cross-section image analysis, we first plotted in Fig. 6 a the integrated oxygen content at each z-value (normalized to its maximum). In each case, a z-value that lies slightly closer to the surface than the layer with maximum oxygen content was selected to draw a representative contour map for the outermost Ti-oxide layer. The contour maps in Fig. 6b and 6c show the concentrations of Ti and Fe, respectively. It is clear that the concentration of Ti in the oxide layer of the M = Co and Ni samples is higher than that of Fe. Moreover, Ti and Fe are segregated within the xy-plane, indicating phase separation between Ti and Fe within the oxide. Conversely, for M = V, the concentration of Fe is significantly higher than that of Ti, although separation of Ti-containing and Fe-containing phases is again observed. The segregation of Ti and Fe is not obvious in pure TiFe or for M = Cr. These characteristics are summarized in Table 2 . Contour maps for some other z-values are provided in Figs. S3–S7 and show similar trends.The APT analysis reveals that thinner Ti-oxide and higher Fe concentration in the oxide layer promotes activation, as expected given the known protective nature of Ti oxides compared to Fe oxides. From the surface oxide characterization, we could successfully uncover how the alloying elements could accelerate the activation process. These alloying elements induce selective oxidation and segregation of Ti and Fe within the oxide layer [53], and the redistribution of the major elements suppresses the formation of the thick passivating Ti oxide found in pure TiFe: the overall effect is enhanced activation kinetics and the degree to which this kinetics is promoted depends on the oxygen affinity of the alloying elements. Although the detailed surface analyses provide meaningful information on the characteristics of the surface oxides, they are not complete enough to fully explain the observed activation behavior of the alloys in this study. Therefore, we proceeded further to study the hydride nucleation as described in the following section.In addition to the consequences for the passivating surface oxide layer, we looked into another potential result of alloying elements on the activation kinetics of TiFe—namely, the nucleation of TiFeH from the TiFe matrix. A peculiar characteristic of hydrogen dissolution in TiFe is that the reaction to TiFeH initially proceeds endothermically, with the solubility increasing with temperature prior to eventual exothermic formation of the fully hydrided product [39]. It contrasts with other interstitial metal hydrides such as LaNi5, for which hydrogen dissolution is strictly exothermic [54]. Hence, in addition to the surface oxide composition, the alloying element may also affect the hydrogen dissolution energy, which in turn governs the kinetics of nucleation. Because this possibility is difficult to probe experimentally, we invoked DFT calculations to estimate the energy change for the following hydride nuclei formation reaction: (1) Ti n Fe n − p M p  + q/2 H2 ↔ Ti n Fe n − p M p H q , ΔEnucl where M stands for the alloying elements. To mimic experimentally relevant compositions, we chose three representative cases for study. Specifically, we used n = 64 and 216 for the 4 × 4 × 4 and 6 × 6 × 6 supercells, respectively. The number of alloying atoms per supercell, p, was 4 or 8 in the 4 × 4 × 4 supercell and 24 in the 6 × 6 × 6 supercell, corresponding to x = 0.0625, 0.125, and 0.111, respectively, in TiFe1− x M x (compared with x = 0.1 experimentally).The hydride nucleus formation energy, ΔEnucl , is plotted as a function of the number of hydrogen atoms (q) in Fig. 7 . Because hydrogen incorporates interstitially in TiFe without significant lattice rearrangement, H atoms were simply inserted into the would-be H lattice sites of TiFeH around the central Fe or M atoms within each TiFe simulation cell. To minimize spurious interactions across periodic supercell images, the maximum number of H atoms in each case was set to 24, which corresponds to a ∼ 2 × 2 × 2 TiFe supercell for the dimension of the TiFeH nuclei, as illustrated by the shaded areas in the insets of Fig. 7a-c. As expected, the computed hydrogen absorption reaction in pure TiFe is endothermic, which is evident from the positive ΔEnucl regardless of the number of inserted H atoms. For the 4 × 4 × 4 supercell in Fig. 7a and b, ΔEnucl reaches a maximum at q = 16 corresponding to a radius for TiFeH nuclei of ∼ 3.8 Å and an apparent barrier for homogeneous nucleation of ∼ 1.5 eV. However, determining the exact barrier and critical nucleus size is challenging within DFT due to long-ranged strain relaxation effects, which demand large supercell sizes. As a step in this direction, we expanded the size of TiFe matrix to 6 × 6 × 6, inserted the same number of H atoms, and examined the energy evolution in Fig. 7c. The energy of system monotonically increases up to 24H atoms, placing a lower bound on both the critical nucleus radius (which should be greater than ∼ 4.8 Å corresponding to the limit of 24H atoms explored) and on the nucleation barrier (which should be greater than ∼ 2.0 eV).Overall, we can conclude that a sizeable barrier must be overcome to nucleate TiFeH within the TiFe matrix. Although endothermic TiFeH formation is well documented, the energetics of the initial process has not been fully appreciated in relation to the nucleation barrier and the difficulty of activation in TiFe. Nevertheless, we caution against overinterpretation of our computed values, as the model considers only homogeneous nucleation within constrained, perfect single-crystal TiFe, and even for this simple case, exact quantities of critical nucleation parameters are difficult to discern. Instead, we are primarily concerned with how trends change with addition of alloying elements. Fig. 7 further demonstrates that the alloying elements strongly affect the energetics of hydrogen dissolution and hydride nucleus formation. At the lowest concentration, x = 0.0625 in Fig. 7a, which is closest to pure TiFe, all alloying elements M exhibit an energy barrier when 16H atoms are inserted in the 4 × 4 × 4 simulation cell. However, V and Cr dramatically decrease the nucleation barrier from 1.5 eV (pure TiFe) to 0.04 and 0.4 eV (for V and Cr, respectively); eventually, these barriers disappear altogether at the higher alloying concentration of x = 0.125 (Fig. 7b). Co and Ni also help decrease the nucleation barriers, but not as powerfully as V and Cr, as demonstrated by the retention of barriers at x = 0.125. Trends are similar for the larger 6 × 6 × 6 supercell (Fig. 7c), which can more effectively capture the large elastic strain fields. In this case, Co and Ni show overall increasing energy profiles up to 24H atoms for x = 0.111, but hydrogen incorporation remains consistently more favorable than for pure TiFe. Meanwhile, alloying with V or Cr removes the nucleation barriers and converts endothermic hydrogen absorption to an exothermic reaction, in agreement with the results in the smaller supercell for x = 0.125. Therefore, we conclude that the chemical effect brought by the alloying elements shows the sequential trend of V > Cr > Ni > Co > Fe, matching well with the experimentally measured time to initiate activation.While our simulations more directly studied the energetics of hydride nucleus formation, previous DFT studies proposed that the bulk TiFeH formation energy can also inform changes in nucleation behavior upon alloying [36,37]. The stability of hydride can affect activation kinetics by adjusting thermodynamic driving force. For instance, Mn addition resulted in more stable hydride and allowed hydrogenation under lower pressure [8]. The bulk hydride formation energy is defined according to the following reaction: (2) TiFe1− x M x + ½ H2 ↔ TiFe1− x M x H. ΔEform Fig. 8 a shows that the formation energy follows a volcano plot with the atomic numbers (Z) of the alloying elements (V (Z = 23), Cr (Z = 24), Fe (Z = 26), Co (Z = 27) and Ni (Z = 28)), with the highest value centered at Fe. The largely symmetric nature of the plot demonstrates that bulk thermodynamics alone cannot explain the observed activation trends (V > Cr > Ni > Co > Fe) upon alloying. Values of ΔEnucl for the largest simulated Ti216Fe192M24H24 nuclei in Fig. 7c are co-plotted as solid triangles in Fig. 8a. Similar to ΔEform , pure TiFe shows the least favorable ΔEnucl ; however, it is interesting to see that incorporation of electron donors with smaller electronegativity (V and Cr) are far more effective at promoting hydride nucleation compared to the electron acceptors with larger electronegativity (Co and Ni), even rendering its formation energetically favorable.In order to identify the origin of the chemical trends among the M elements, we proceeded to perform further structural and electronic analysis. First of all, based on the optimized cell volume of TiFe1− x M x prior to hydrogen absorption, we found that the electron donors (V and Cr) expand the TiFe lattice more than the electron acceptors (Co and Ni). Fig. 8b shows that at x = 0.111, incorporation of V increases the molar volume of TiFe by 1.70 %, whereas Ni expands the lattice by only 0.66 %, in good agreement with the experimental volume expansion in Table S2. As the TiFeH phase requires larger molar volume than TiFe, the lattice expansion upon alloying may benefit nucleation kinetics, especially for V and Cr. Besides the volumetric effect, V and Cr atoms also have lower electronegativity than Fe, Co and Ni (Fig. S8), resulting in stronger reduction of the H atoms in the TiFeH nuclei. This is reflected in the higher average Bader charges in Fig. 8b. Both ground-state molar volume and electronegativity are important descriptors determining the hydrogen absorption properties in intermetallic alloys [56]. Elements with larger volume and lower electronegativity tend to form more stable hydrides, as demonstrated by V and Cr in the current investigation.In addition, it seems that electron-donating nature of V and Cr leads to the migration of nearest-neighbor octahedral H atoms to tetrahedral interstitial sites, as displayed in Fig. S9. Larger Bader charges on H imply a larger effective volume, and hence H atoms that are nearby V or Cr may prefer the more spacious tetrahedral sites. These differences in charge state and tetrahedral interstitial occupancy of H atoms can further impact the hydrogen absorption-induced volume expansion of TiFe. When 24H atoms were inserted to form TiFeH nuclei, the M−H bonds were found to be longer than Fe−H bonds for all M atoms (Fig. S10a), even though Co and Ni have similar metallic radii to Fe. On the other hand, Fe−H bonds within the TiFeH nuclei are shortened only when V or Cr atoms are nearby, implying that the H atoms at tetrahedral interstitial sites can help mitigate the elastic energy of TiFeH nuclei while facilitating stronger Fe−H bond formation. Indeed, the least degree of volume expansion upon TiFeH nucleus formation was observed for M = V and Cr (Fig. S10b).In summary of the calculation results, the reducing power of V and Cr and their accommodation of lattice expansion play a synergistic role in changing the local chemical environment and electronic state of H atoms in the TiFeH nuclei region and reducing the elastic energy penalty for the nucleation of TiFeH. These impacts are realized by the conversion of endothermic to exothermic hydrogen absorption. Although some lattice expansion is observed for M = Co and Ni, their electronegativity is insufficient to further reduce the inserted H atoms beyond pure TiFe, somewhat muting their impact. The penalties for hydride nucleation in alloys with M = Co and Ni are lower than in pure TiFe, but the endothermic nature of hydrogen absorption remains, explaining the asymmetric behavior of ΔEnucl with the atomic number of the alloying element observed in Fig. 8a.In this study, TiFe0.9M0.1 (M = V, Cr, Fe, Co and Ni) alloys were investigated from two different perspectives to explain the experimentally observed variation in activation kinetics: (i) elemental distribution and valence state inside surface oxide revealed by XPS and APT analyses and (ii) energetics of hydride nucleation predicted by DFT calculations. M = V and Cr alloys were easily activated at 323 K, whereas M = Co and Ni alloys were activated only after vacuum pretreatment at 473 K; pure TiFe was not activated at all. Overall, our findings point to a combined effect of passivating oxide modification and altered hydride nucleation energetics as key elements for promoting activation.Several specific factors were identified in our analysis. XPS indicated that the oxidation of V is more pronounced than that of Fe, which suggests sacrificial oxidation of V and protection of Fe from oxidation. APT analysis further suggested that thinner Ti oxide and higher Fe concentration in the surface oxide layer may promote activation. In addition, the introduction of a foreign element with a different oxygen affinity appears to induce selective oxidation and phase separation of Ti and Fe in the oxide layer. We propose that this redistribution of major elements inhibits the formation of the thick, continuous passivating oxide layers found on pure TiFe, thus accelerating the activation process.From a nucleation point of view, the reducing power and lattice expansion tendency for M = V and Cr act synergistically to alter the local chemical and strain environment and reduce the energy penalty associated with TiFeH nucleation. Their effect is realized by converting the endothermic hydrogen absorption into an exothermic absorption. Meanwhile, the effect of M = Co and Ni is limited to a more modest decrease in the nucleation penalty, while retaining endothermic absorption. The trend in hydride nucleus formation energetics follows that of the activation kinetics (V > Cr > Ni > Co > Fe), highlighting the additional importance of understanding nucleation to interpret the activation behavior of TiFe alloys. Hayoung Kim: Investigation, Formal analysis, Writing – original draft, Visualization. ShinYoung Kang: Methodology, Investigation, Formal analysis, Writing – original draft, Visualization. Ji Yeong Lee: Investigation. Tae Wook Heo: Resources, Project administration. Brandon C. Wood: Funding acquisition, Writing – review & editing. Jae-Hyeok Shim: Resources, Writing – review & editing. Young Whan Cho: Resources, Writing – review & editing. Do Hyang Kim: Supervision. Jin-Yoo Suh: Writing – original draft, Project administration, Funding acquisition. Young-Su Lee: Conceptualization, Methodology, Investigation, Writing – original draft, Funding acquisition.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This research was supported by the Korea Institute of Science and Technology [grant numbers 2E31851, 2E31858] and by the National Research Foundation of Korea [grant number NRF-2020M1A2A2080881]. Part of this study was performed under the auspices of the DOE by Lawrence Livermore National Laboratory under contract DE-AC52-07NA27344, with support from the Hydrogen Materials Advanced Research Consortium (HyMARC), established as part of the Energy Materials Network by the U.S. Department of Energy (DOE), Office of Energy Efficiency and Renewable Energy, Hydrogen and Fuel Cell Technologies Office, under Contract No. DE-AC52-07NA27344. Computational resources were sponsored by the Innovative and Novel Computational Impact on Theory and Experiment (INCITE) program and the DOE's Office of EERE, located at the Argonne National Laboratory and National Renewable Energy Laboratory.Supplementary data to this article can be found online at https://doi.org/10.1016/j.apsusc.2022.155443.The following are the Supplementary data to this article: Supplementary data 1

Despite the promise of TiFe-based alloys as low-cost solid-state hydrogen storage materials with mild operating conditions and reasonable hydrogen capacity, their initial hydrogenation process is difficult, hindering broad utilization. The effect of alloying element on the initial hydrogenation kinetics of TiFe alloys, TiFe0.9M0.1 (M = V, Cr, Fe, Co and Ni), was evaluated by analyzing changes to the passivating surface oxide layer that inhibits hydrogen permeation, as well as the ease of initial-stage hydrogen absorption into the underlying matrix. X-ray photoelectron spectroscopy and atom probe tomography revealed key variations in surface oxide compositions and thinning of the passivating oxide layer compared to pure TiFe, which suggests suppressed oxide growth by alloying-induced elemental redistribution. At the same time, density functional theory calculations predicted exothermic formation of hydride nuclei when alloying with V or Cr, as well as a reduced nucleation barrier when alloying with Co or Ni. Overall, these results are consistent with the observed experimental trend of the activation kinetics. We propose that improvements in activation kinetics of TiFe with alloying arises from the combined effect of reduced passivating oxide thickness and easier hydride nucleation, offering a starting point for alloy design strategies towards further improvement.

Compared with conventional fossil fuels, hydrogen fuel (Nikolaidis and Poullikkas, 2017; Luo et al., 2018; Holladay et al., 2009; Dawood et al., 2020) has attracted significant attention for its unique features, such as high flame propagation velocity, high heating value of combustion, and zero carbon emissions. However, the storage and transportation problems of hydrogen fuel impede its practical applications (Dincer, 2012; Kalinci et al., 2015; Granovskii et al., 2007, 2006; Acar and Dincer, 2014; Hu and Ruckenstein, 2003). New technology of energy carrier for the storage and transportation of hydrogen gradually becomes more urgent and essential. Ammonia (NH3), as an alternative energy carrier of hydrogen, has drawn extensive attention and research (Xue et al., 2019; Miura and Tezuka, 2014; Elishav et al., 2017) due to the advantages of ammonia mainly including easy liquefaction, low cost of storage and transportation, and no carbon emission.The methods of hydrogen production from ammonia decomposition can be classed as: catalyst (Xiao et al., 2011; Walkosz et al., 2020; Wang et al., 2019, 2017; Jolaoso et al., 2018; Podila et al., 2020, 2016, 2017; Zaman et al., 2018), plasma (Soucy and Boulos, 1995; Hayakawa et al., 2020; Akiyama et al., 2018; Qiu et al., 2004; Hsu and Graves, 2005, 2003; Zhao et al., 2014), and catalyst combining plasma (Zhao et al., 2013; Yi et al., 2019; Hayakawa et al., 2019; El-Shafie et al., 2020). Decomposition of ammonia by catalyst was a thermal decomposition method, which usually needs relatively large space for the sake of catalyst storage and heating elements. In addition, the start-up time of ammonia decomposition by thermal catalyst usually cannot satisfy the requirement of engine, which needs the instantaneous ammonia decomposition. Owing to high-energy electron and enough reactive species produced in a short time scale (less than micro-seconds), plasma technology was a promising alternative for instantaneous ammonia decomposition. Various types of plasma reactor have been designed and investigated, including thermal plasma, e.g. RF discharge (Soucy and Boulos, 1995), cold plasma, e.g. dielectric barrier discharge (DBD) (Hayakawa et al., 2020; Akiyama et al., 2018) and micro-hollow cathode discharge (Qiu et al., 2004; Hsu and Graves, 2005, 2003), and warm plasma, e.g. non-thermal arc plasma (Zhao et al., 2014, 2013). In thermal plasma, the thermal effect on the ammonia decomposition is dominant. For instance, Soucy and Boulos (1995) has developed RF thermal plasma reactor to produce hydrogen from ammonia, and achieved a high ammonia conversion rate of 98% under the condition of 13 kW discharge power and 28 SLM NH3 gas flow rate. However, relatively high heat loss by conduction leads to low energy efficiency, which would impede its practical application. In cold plasma, high-energy electron and reactive species play an important role in ammonia decomposition. Hayakawa et al. (2020) has designed a DBD reactor with a hydrogen separation membrane to produce hydrogen from ammonia. High pure hydrogen was produced, but hydrogen generation rate was only 20 ml/min because of low gas flow rate for the sake of increasing the residence time in discharge zone. Warm plasma can maintain moderate gas temperature (usually 1500 ∼ 4000 K) and enough reactive species, which would be more suitable for the ammonia decomposition. In general, non-thermal arc plasma (NTAP) was one of fundamental methods for warm-plasma generation. Zhao et al. (2014) has designed a NTAP reactor by two tube electrodes driven by alternating current. The increase of effective plasma volume and gas temperature results in high energy efficiency (EE) of hydrogen production (330.1 L/kW ⋅ h). However, the gas flow rate in the NTAP reactor was still in the order of milliliter, and the absolute hydrogen production rate is limited. Besides, the heat generated from NTAP needs to be utilized more efficiently, such as heating the catalyst. In previous study combining with catalyst, only DBD plasma source was investigated (Yi et al., 2019; Hayakawa et al., 2019). The catalyst was placed in the discharge zone of DBD plasma, which may lead to the spark discharge and damages to the plasma reactor. In hence, taken together, requirement of a new NTAP reactor combining with catalyst for hydrogen production from ammonia becomes urgent and essential.In the present work, a novel NTAP reactor combined with NiO/Al2O3 catalyst has been developed to dissociate ammonia instantaneously. The moderate gas temperature in this NTAP reactor is very suitable for the high performance of ammonia decomposition. Large processing capacity and no cooling system contributes to the higher energy utilization efficiency. To utilize the extra heat of NTAP and further increase the ammonia conversion rate, NiO/Al2O3 catalyst was added at the nozzle exit of NTAP reactor to avoid the interference between catalyst and plasma, as mentioned in above DBD plasma combining with catalyst. The effects of gas flow rate and discharge power on the gas temperature, electron density of arc plasma, and hydrogen production rate are investigated through optical emission spectrometer, thermal couple and hydrogen detector, respectively. Fig. 1 presented the schematic diagram of the experimental setup for hydrogen production by NTAP. The experimental apparatus mainly includes a NTAP reactor, NiO/Al2O3 catalyst, a high-frequency high-voltage power supply (driving frequency: 23.8 kHz), a feed gas system, a cooling condenser, hydrogen detector, and a diagnostic system. The NTAP reactor mainly consists of a rod-type high-voltage electrode, concentric ground electrode, and a swirl gas ring. The commercial NiO/Al2O3 catalyst (weight: 200 g, model: AD-946, company: JIANGXI HUIHUA TECHNOLOGHY CO., LTD) was placed in the region with a distance of 6 mm away from the plasma generator exit. The mass fraction of NiO in NiO/Al2O3 catalyst is about 15%. The crystal structure of NiO is hexagonal, and the crystal parameters of NiO is 2.9552 Å × 2.9552 Å × 7.2275 Å. The average crystalline size of NiO is 23.975 nm. The crystal structure of Al2O3 is hexagonal, and the crystal parameters of Al2O3 is 4.758 Å × 4.758 Å × 12.991 Å. The average crystalline size of Al2O3 is 77.82 nm. The NTAP reactor was powered by a high-frequency high-voltage power supply with adjustable output power ranging from 0 to 700 W. The pure ammonia (99.999%) was injected into the swirl gas ring to act as the discharge gas, and its gas flow rate was controlled by a mass flow controller (0 ∼ 50 SLM). The discharge current I a r c was obtained by measuring the current of a resistor ( R = 1 Ω ) in series with the NTAP reactor. The discharge voltage V a r c and current I a r c were measured by a four-channel oscilloscope (KEYSIGHT DSOX2024A) with a high-voltage probe (Tektronix P6021A) and normal voltage probe (KEYSIGHT N2862B), respectively. The discharge power can be obtained by discharge voltage V a r c and current I a r c as follows: (1) P = 1 T ∫ 0 T V a r c × I a r c d t , Where T represented the one cycle of discharge voltage waveform. The effective value of discharge voltage V R M S and I R M S were calculated by the following equations: (2) V R M S = 1 T ∫ 0 T V a r c 2 d t , (3) I R M S = 1 T ∫ 0 T I a r c 2 d t , The optical emission spectrum was obtained by a spectrometer (AvanSpec-2048). The gas temperature at the nozzle exit was measured by a thermal couple. Ammonia was decomposed by NTAP according to the following reaction: (4) 2 NH 3 ⟶ plasma N 2 + 3 H 2 , The products of ammonia decomposition were NH3, H 2 , and N 2 . The hydrogen content was analyzed by a hydrogen detector (SKY600-XYH2, Shanghai Xiyu Instrument Equipment Co., Ltd). The energy efficiency EE for hydrogen production can be defined as follows: (5) EE = v o l u m e o f H 2 p r o d u c e d [ L ] D i s c h a r g e p o w e r [ kW ⋅ h ] , Fig. 2 shows the typical discharge voltage and current evolution of NTAP reactor at a time span of 1000  μ s and 50  μ s. It can be seen that the shape for the envelope of discharge voltage was sawtooth-like with a time scale of hundreds of microseconds as shown in Fig. 2(a). The magnitude of discharge voltage waveform varied from about 1.8 kV to 3 kV. The detailed voltage and current waveforms were shown in Fig. 2(b). The waveform of discharge voltage has characteristics of the sinusoid-like shape with the amplitude of around 1.9 kV. The frequency of discharge voltage was about 23.8 kHz. The discharge current can be divided into two parts: the spike component and the sine-like component. The spike component and the amplitude of sine-like component of current reached to around 4 A and 430 mA, respectively. The current–voltage characteristics of NTAP was shown in Fig. 3. The voltage–current characteristics of NTAP presented negative slope. Both the variable region of V R M S and I R M S and the absolute value of the slope increased with the increasing of gas flow rate as shown in Fig. 3, which indicates that the NTAP reactor presented in this paper is more controllable for operating at high ammonia gas flow rate. Fig. 4 presents the typical optical emission spectrum of NTAP at the gas flow rate of 30 SLM and discharge power of 700 W. The emission spectrum of NTAP was dominated by the emission bands of NH*(A3 Π → X3 Σ − ), N 2 *(C3 Π u → B3 Π g ), H 2 *, NH2*( A ̃ 2 A 1 → X ̃ 2 B 1 ), and the atom spectrums of H α (656.3 nm), H β (486.1 nm), and Cr (422.7 nm, 425.3 nm, 427.4 nm, 428.9 nm, 588.9 nm, 589.5 nm). These reactive species, such as NH*, N 2 * and H 2 *, indicate that ammonia has been effectively dissociated by NTAP. These intermediate and final products was formed by following reaction process (Fateev et al., 2005; van den Oever et al., 2005; Yang et al., 2002; Yasui et al., 2003; Arakoni et al., 2007):Energy exchange: (6) e + NH 3 → 2 e + NH 3 + , (7) e + NH 3 + → NH 2 + H , (8) e + NH 3 + → NH + p H 2 + 2 − 2 p H ( p = 0 , 1 ) , (9) e + NH 3 + → N + q H 2 + 3 − 2 q H ( q = 0 , 1 ) , (10) e + NH 3 → NH 2 − + H , (11) e + NH 3 → NH 2 + H − , Binary processes: (12) NH 2 + N → N 2 + 2 H , (13) NH 2 + H → H 2 + NH , (14) NH 3 + H → H 2 + NH 2 , (15) NH 2 + NH → N 2 H 3 , (16) NH + H → H 2 + N , (17) NH + NH → N 2 + 2 H , (18) N 2 H 3 + H → N 2 + 2 H 2 , It can be seen that high-energy electron is crucial to exchange energy with atom and molecules, and form the intermediate products. Besides, in the binary processes, most of these reaction rate coefficients rely on the gas temperature, such as reaction (13) and (14) (Arakoni et al., 2007). Therefore, the parameters of electron, such as electron density, and gas temperature are the key factors that influence the performance of ammonia decomposition. Fig. 5 presents the gas temperature ( T g ) at the exit of NTAP reactor under different gas flow rate and discharge power. It can be seen that the gas temperature increased with the discharge power when the gas flow rate was kept a constant. The lowest value of gas temperature was 673.2 K at the discharge power of 210 W and gas flow rate of 30 SLM, and the highest value was 1116.2 K at the discharge power of 700 W and gas flow rate of 20 SLM. According to precious literature Lin et al. (2018), the gas temperature inside the nozzle channel was much higher than that at the exit of NTAP reactor. The thermal pyrolysis effect of NTAP on the ammonia decomposition may be remarkable inside the nozzle channel, because some key endothermic reactions of producing hydrogen becomes more significant with the increasing of gas temperature. For instance, the abstraction of H from NH2 and NH3 by H atoms can produce H 2 by following reactions as mentioned in above section (Arakoni et al., 2007): (19) NH 2 + H → H 2 + NH , k = 1 . 1 × 1 0 − 10 e − 4451 / T g (20) NH 3 + H → H 2 + NH 2 , k = 6 . 5 × 1 0 − 13 ( T g / 300 ) 2 . 76 e − 5135 / T g The rate coefficients for reaction (13) and (14) at 300 K are 3.96*10−17 cm 6 s−1 and 2.39*10−20 cm 6 s−1, respectively, while these values at 1000 K are 1.28*10−12 cm 6 s−1 and 1.06*10−13 cm 6 s−1, respectively. This phenomenon suggests that increasing the gas temperature is beneficial for the production of hydrogen. In hence, the gas temperature plays an important role in the ammonia decomposition by NTAP. In addition, the gas temperature in NTAP is higher than that of dielectric barrier discharge, so the effect of thermal decomposition of ammonia becomes more significant in NTAP. The electron density is an important parameter in the energy change and the formation of intermediate products. The electron density can be calculated from the Stark broadening of the H β Balmer line  (Kelleher et al., 1993; Wiese et al., 1975). The typical Voigt fit of the recorded H β line and the electron density as a function of discharge power and gas flow rate were shown in Fig. 6. It can be seen that the electron density was in the order of 1018 m−3, and it increased by enhancing the discharge power and decreasing the gas flow rate. Hence, there is enough number of high-energy electron in NTAP to participate in the energy change and the formation of intermediate products. The effects of gas flow rate, discharge power, and catalyst on the hydrogen production performance of NTAP were investigated and presented in Fig. 7. It can be seen that increasing the discharge power and decreasing gas flow rate contributed to the enhancement of hydrogen production. On the one hand, increasing gas flow rate results in the decreasing of residence time of ammonia in the discharge zone, which leads to reducing the chance of collision probability between ammonia molecule and high-energy electrons, and further the decreasing of hydrogen content. On the other hand, increasing gas flow rate contributes to the increasing of the volume of ammonia processed by NTAP in unit time. Obviously, in our experiment condition (gas flow rate of NH3: 10 slm ∼ 30 slm), increasing the gas flow rate is beneficial for the absolute value for the production of hydrogen. Enhancing the discharge power can result in the higher average energy obtained by each ammonia molecule in the discharge zone. Therefore, more ammonia molecules have enough energy to participate in the ammonia decomposition. In addition, the NTAP combining with catalyst would effectively enhance the hydrogen production. The highest hydrogen content in the production of ammonia decomposition reached about 34.8% as shown in Fig. 7(a). Although the highest hydrogen production rate was achieved at the gas flow rate of 20 SLM, the highest energy efficiency 1080.0 L/kW ⋅ h was obtained at the gas flow rate of 30 SLM as shown in Fig. 7(b). Owing to the catalyst heated by the plasma jet automatically, the length of plasma jet can influence the size of the space loading catalyst. The length of plasma jet can be elongated with the increasing of gas flow rate, which suggests that higher gas flow rate helps to more catalyst heated by plasma jet. Therefore, high gas flow rate contributes to the homogeneous heating of catalyst, and the higher absolute hydrogen production performance can be obtained by combination of plasma and catalyst. Besides, the hydrogen production rate is closely associated with the ammonia decomposition rate. Higher hydrogen production rate can be realized by increasing the discharge power, which means that higher ammonia decomposition rate can be obtained at high discharge power.The hydrogen production performances of NTAP in this paper compared with that of other plasma types were shown in Table 1. The largest absolute hydrogen production rate and highest energy efficiency were simultaneously obtained, which were difficult for other plasma sources due to the limitation of structure and plasma techniques. The high performance of ammonia decomposition may be related to the discharge features of NTAP. In comparison with that of cold plasma reactor, relative higher gas temperature and higher electron density of NTAP as mentioned in above section can generate more radical species facilitating chemical reaction rate. The plasma reactor of this paper has higher energy utilization efficiency due to no cooling systems, compared with that of thermal plasma reactor. In addition, compared with that of the NTAP reactor in literature Zhao et al. (2014), the higher performance of ammonia decomposition in this paper may be associated with the gas injection method of plasma reactor. The swirl flow in NTAP reactor of this paper is favorable for the energy exchange between ammonia molecules and plasma, and thus the ammonia decomposition, while the straight flow is utilized in literature Zhao et al. (2014).Besides, according to experiment, the hydrogen content of 5 ∼ 20% in hydrogen/ammonia combustion can efficiently enhance the combustion stability and flame propagation speed. Therefore, the shortcoming of ammonia fuel may be overcome by means of on-line hydrogen production from ammonia by NTAP. In addition, in order to evaluate the potential of ammonia decomposition by NTAP, the NH3/H2 mixture after ammonia decomposition was injected to the engine. In our engine (NH3 H2 fuel) under development, the ratio of the power of electric generator to the output power of engine that we expected is less than 2%. The rated output power of this engine is about 5 kW, which means that ideally the discharge power of plasma must be less than 100 W. However, in order to maintain the steady operation of engine under present conditions, the discharge power of plasma needs to be larger than 500 W, which suggests that the energy efficiency must be enhanced by five times at least. In summary, an instantaneous, high-efficiency and large-scale hydrogen production device from ammonia by NTAP combined with NiO/Al2O3 catalyst has been developed. The effects of gas flow rate, discharge power on the gas temperature, electron density, and the hydrogen production performance were investigated. The main experimental results and conclusions were listed as follows:With the increase of gas flow rate, the variable region and slope of discharge voltage and current were larger, which indicates that the NTAP reactor is more suitable for operating at high gas flow rate for the sake of easy output adjustment.Owing to the emission spectrum of NTAP dominated by NH*(A3 Π → X3 Σ − ), N 2 *(C3 Π u → B3 Π g ), H 2 *, and the atom spectrums of H α , H β , it can be observed that ammonia can be effectively dissociated by the NTAP. By measuring the gas temperature, the thermal decomposition effect of NTAP would be more remarkable compared with some other plasma sources, e.g. dielectric barrier discharge.The hydrogen content in the production of ammonia decomposition varied in a large region (5.3% ∼ 34.8%) by adjusting the gas flow rate, discharge power, and adding catalyst or not. In the case of NTAP, the highest energy efficiency was 783.4 L/kW ⋅ h at the discharge power of 700 W and gas flow rate of 30 SLM, while the highest energy efficiency of 1080.0 L/kW ⋅ h was obtained by means of NTAP combining with catalyst. Q.F. Lin: Investigation, Writing - original draft. Y.M. Jiang: Data curation. C.Z. Liu: Resources. L.W. Chen: Writing - review & editing, Conceptualization. W.J. Zhang: Validation. J. Ding: Project administration. J.G. Li: Supervision, Funding acquisition.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This work was supported by the National Natural Science Foundation of China (11575252, 11775270), and Institute of Energy of Hefei comprehensive National Science Center, China (19KZS206, 21KZS201).

Owing to the storage and transportation problems of hydrogen fuel, exploring new methods of the real-time hydrogen production from ammonia becomes attractive. In this paper, non-thermal arc plasma (NTAP) combining with NiO/Al2O3 catalyst is developed to produce hydrogen from ammonia with high efficiency and large scale. The effects of ammonia gas flow rate and discharge power on the gas temperature, electron density, the hydrogen production rate, and energy efficiency were investigated. Experimental results show that the optical emission spectrum of NTAP working with pure ammonia medium was dominated by the atom spectrum of H α , H β , and molecular spectrum of NH component. Under the optimum experimental condition of plasma discharge, the highest energy efficiency of hydrogen production reached 783.4 L/kW ⋅ h at NH3 gas flow rate of 30 SLM. When the catalyst was added, and heated by the NTAP simultaneously, the energy efficiency further increased to 1080.0 L/kW ⋅ h.

Climate change caused by the burning of fossil fuels is one of the major global problems. The development for clean and renewable energy resources to ensure the sustainable development of the world is in urgent need [1,2]. Proton exchange membrane fuel cells (PEMFCs) that can convert the chemical energy (from clean and renewable fuels such as hydrogen) directly into electrical power have attracted tremendous attentions [3]. In the fuel cell system, oxygen reduction reaction (ORR) proceeds at the cathode while hydrogen oxidation reaction (HOR) occurs at the anode. The catalysts are essential to lower the activation energy, thus making these reactions kinetically-activated [4,5]. Moreover, the kinetics of ORR is five-orders-of magnitude slower than that of HOR, which leads to more catalyst usage in the cathode. Therefore, developing highly efficient cathodic ORR electrocatalysts is highly desired for high-performance and low-cost fuel cells.ORR can be catalytically achieved through two different mechanisms: (i) four-electron (4e−) pathway to directly produce water in acid media or hydroxide anion in alkaline media; (ii) two-electron (2e−) pathway to create the intermediate compound (hydrogen peroxide) in acid media or peroxide anion in alkaline media [6]. Compared with 2e− pathway, the 4e− process is more attractive because it could ensure higher operating potentials and current efficiency in fuel cells. Currently, the carbon black supported platinum (Pt/C) catalysts are the state-of-the-art 4 e− catalysts for ORR in PEMFCs. Due to the scarcity of Pt, finding alternative catalysts with less cost is one of the central tasks for the large-scale commercial implementation of the fuel cell industry. Low-Pt catalysts, non-precious metal catalysts, and metal-free catalysts have been widely investigated as substitutes for the commercial Pt/C [7–9].Graphene is a 2D layer with sp2-bonded carbon atoms, which has attracted worldwide attention in both metal-free catalysts and supporting materials for ORR. Inspired by the successful development of graphene, a variety of studies of using other types of 2D materials as cost-effective ORR catalysts have started to appear, such as transition metal dichalcogenides (TMDs), phosphorene, hexagonal boron nitride (h-BN) and graphitic carbon nitride [10–12]. Besides, the 2D graphene can be split into smaller particles such as 0D quantum dots and 1D nanoribbons [13]. With this revelation, low-dimensional materials have recently emerged as new types of electrocatalysts [14–17].When the materials are smaller than 100 nm in at least one dimension, their electronic structure and surface properties change drastically, resulting in unique physical and chemical properties due to their quantum size effect and large surface area. In low-dimensional materials, the motion of electrons is restricted in zero, one, and two dimensions. As shown in Fig. 1 , the polts of the density of states (DOS) versus energy in low dimensional structures are distinct from each other. It's a staircase for 2D materials, a fence for 1D material, and a pack of discrete lines for 0D materials. In 0D materials, such as nanoparticles or quantum dots, the electrons are confined in all three directions (x, y, z-axis) and cannot move anywhere [15]. The 0D material shows abundant edges and low coordinated sites. In 1D materials, such as nanotubes, nanorods, and nanowires, the electrons are restricted in two directions and can only move in one direction. The 1D material displays a high length-to-width ratio with preferred facets. In 2D materials, such as thin nanofilms and nanosheets, only one direction is restricted for the movement of the electron. And they exhibit high surface area and edge effects. Dimensionality plays a vital role in depicting the fundamental properties of a material. Due to the quantum mechanical effects, the properties of these low-dimensional materials are significantly different from those of bulky materials.Low-dimensional materials with designated thickness, size, and morphology show many advantages for catalysis [15]. On the one hand, the high surface area could increase the number of active sites on the surface, which enhances the atom utilization. On the other hand, their unique electronic structure and low coordinated environment benefit the interactions between catalysts and reactants. According to the Sabatier principle, the adsorption energy of the reactants and intermediates on the catalyst surface should be neither too small nor too large. The interactions between catalysts and reactants are largely dependent on the electronic structure of the catalyst, affecting the catalytic activity and stability of the catalyst. Therefore, the controlled synthesis process of low-dimensional catalyst materials with desirable electronic structures is the key to the promotion of catalytic performance.The synthetic strategies of low-dimensional materials can be divided into two main categories: (i) top-down and (ii) bottom-up, as shown in Fig. 2 . Briefly, the top-down approach is to reduce the bulk material to the nanoscale by liquid-phase exfoliation or physical methods. For example, graphene can be obtained from graphite by breaking the van der Waals forces between the graphite layers using adhesive tape [18]. The general idea of the bottom-up strategy is using small building blocks made of atoms or molecules to synthesize nanomaterials. Chemical synthesis methods such as chemical vapor deposition (CVD) and hydro/solvothermal methods are typical examples [19]. In general, top-down approaches are simple and easy for scale-ups but lack the precise control of the morphology. The bottom-up method is good at synthesizing nanomaterials with precisely controlled size, shape, and morphology. The synthetic approaches of low-dimensional materials have been well summarized in many recent papers [20,21]. This review will focus on the relationship between the structure and the ORR activity of the low low-dimensional materials.2D materials have emerged as the promising catalysts for ORR since the 2D-graphene was explored experimentally in 2004 [23]. As shown in Fig. 3 , a wide range of 2D materials such as graphitic carbon nitride (g-C3N4), metal dichalcogenides (TMDs, e.g. MoS2, MoSe2), layered metal, layered metal-organic frameworks (MOFs), MXenes (transition metal carbides, nitrides, and carbon nitrides), layered transition metal oxides (TMDs, e.g. MnO2), and hexagonal boron nitride (h-BN) have also been successfully prepared and applied in electrochemistry in recent years. These 2D materials are of single or a few layers in thickness. This distinct structure along with the extraordinary physical and chemical properties such as mechanical flexibility and high specific surface area make 2D nanomaterials promising as the catalyst support in electrochemical energy conversion. Moreover, the electrocatalytic activity of 2D material can be tailored by inducing intramolecular charge transfer with heteroatom-doping or defect engineering. The unique electronic structure and increased active sites render 2D material appealing candidates for the ORR catalyst [11]. However, the active sites of these 2D materials differ significantly, which will be discussed in the following section.Graphene-based catalysts have attracted great attention for ORR [24–26]. Graphene exhibits many excellent chemical and physical properties as well as the unique graphitic basal plane structure that could supply numerous active sites and facilitate the electron transport. Unfortunately, the ORR activity of pristine graphene is very poor due to the lack of free electrons for the reaction and a limited number of active sites [27]. The electroneutrality of pristine graphene should be broken down to create charged sites that are favorable for O2 adsorption. Introducing dopants and generating defects on graphene could modulate the surface energy, the chemisorption energy of O2, as well as the local electronic properties and thus enhance the catalytic properties [28].Among various possible dopants, N-doped graphene is by far the most investigated material. The atomic radius of N is close to C. So it is relatively easy to incorporate N into the backbone of the graphene material. The carbon atoms that are adjacent to the more electronegative nitrogen atoms are supposed to be the active sites for ORR [30]. The interaction with oxygen molecules will be facilitated because of the favorably changed charge profile. As shown in Fig. 4 , there are three types of nitrogen dopants (pyridinic, pyrrolic, and graphitic). Each type of these nitrogen dopants influence the catalytic property differently and the exact role of the specific N specie in ORR is still under debate [31,32]. Pyridinic nitrogen is believed to be the most active site for ORR [33,34]. The lone electron pair of pyridinic N could donate to the π-bond of the carbon matrix, making pyridinic N atom itself electron-attractive and catalytically active. Satoshi Yasuda et al. [35] used temperature-induced surface polymerization of nitrogen-containing aromatic molecules to get two types of N-doped graphene: one is mainly composed of pyridinic nitrogen and the other is graphitic nitrogen. It was revealed that pyridinic nitrogen could catalyze the ORR via the 4e− process, whereas graphitic nitrogen reduced oxygen via a 2e− pathway. However, different conclusions were proposed by other researchers. [36,37], Luo et al. [38] employed a pyrolysis method using methane (CH4) and ammonia (NH3) as carbon and nitrogen sources, and the graphene doped with nearly 100% pyridinic N was obtained on the surface of a Cu substrate. The results showed that the pyridinic N doped graphene prefers a 2e− pathway for ORR, indicating that the pyridinic N couldn't effectively promote ORR by a 4e− dominated process. This result may play a vital role in tailoring the electronic properties for the improvement of the ORR performance of the pyridinic N. Ruoff et al. [37] prepared the N-doped graphene with controlled N doping species by selectively annealing graphene oxide (GO) with different N-containing precursors. Higher ORR activity and larger limiting current density could be obtained by graphitic N-dominated catalysts. In contrast, more positive onset potential of ORR could be found in the pyridinic N-dominated catalysts. However, the catalytic ability of doped graphene regarding the ORR depends on many complicated factors of the coordination environment including surface area, morphology, crystallographic structure, oxidation state, and chemical composition. It cannot be simply summarized over a few studies. The comprehensive and profound research for ORR is needed to discuss the real mechanism.Meanwhile, as shown in Fig. 5 , the dopping of heteroatoms in graphene lattice can cause a charge density redistribution because of their distinct electronegativity. In practice, the atomic radius of heteroatoms should be considered. Large atoms such as Br, I, and Se cannot be incorporated into the carbon lattice. Boron (B), sulfur (S), and phosphorus (P) have been widely used for the heteroatom-doped graphene [30,39,40]. The catalytic activities will be enhanced because of the spatial distortion and charge redistributions. Since the electronegativity of B atoms is less than that of carbon atoms, when they are introduced into graphene, the p-type conductivity can be induced. The B-doped graphene is believed to be a good catalyst for promoting the 4e− process [41]. As for S, the electronegativity of S is close to that of carbon. However, the size of the S atom is larger than that of the C atom. Spatial distortion of S-doped graphene is a key factor to enhance the ORR activity. The carbon atoms near doped sulfur atoms and the zigzag edges are the catalytic active sites in S-doped graphene [42]. In terms of P doping, the P3p and C2p orbital are hybridized by sp [3]-orbital configuration, which is shown as a pyramidal structure. These structures are easily oxidized to generate C–P–O bonds. The active sites of P-doped graphene are positively charged carbon atoms because the O atoms in the C–P–O structure can polarize the P and the adjacent C atoms. Graphene doped with two or more kinds of heteroatoms have also been proven to be interesting because of the increased number of dopant heteroatoms and the synergistic effects between the dopants. B–N co-doped, P–N co-doped, S–N co-doped graphenes are the most widely studied materials for ORR [43,44]. The intermediate adsorption on the surface of doped graphene can be enhanced by tailoring the electron-donor properties with controlled dopant types and their amounts. Elucidating the underlying ORR mechanism of the doped graphene will be one of the top priorities for future research.Transition-metal dichalcogenide (TMD or TMDC) monolayers are thin-layer materials and their general chemical formula is MX2. The M represents transition-metal atoms (e.g. Mo, W, etc.) whereas the X represents chalcogen atoms (e.g. S, Se, Te). In the MX2 catalysts, a single layer of transition metal atoms is situated between two layers of chalcogen atoms where a strong covalent bond between the M and X atoms is observed [45]. The 2D TMDs show adjustable bandgaps that are decided by the stacking order between layers and the different coordination models between M and X atoms, which will lead to different surface properties and electronic structures and thus varied catalytic activities will be observed [46].Among the known sixty TMDCs thus far, 2D molybdenum disulfide (MoS2) has attracted much attention for the ORR catalyst. Their active sites are the edges and corner atoms with low-coordination. In particular, Mo atoms passivated by S atoms that are found at the edges are determined to be the most active sites because of their sulfide rich and unsaturated coordinated environment as well as their dangling bonds [47]. However, the pristine MoS2 shows poor electron transport properties and limited active sites. Various attempts have been made to expose more active edge sites and improve conductivity. For example, heteroatom doping can be used to activate the atoms on the basal plane. Similar to the doping process of the graphene, B, N, P, and O were introduced to the MoS2 [48,49]. Moreover, metallic atoms such as Fe, Co, Ni are proved to be effective for improving the ORR activity as well [50,51]. In addition, conducting nanomaterials, such as graphene, are used as the support to improve the electron transport properties of the catalysts. As shown in Fig. 6 , the N–MoS2/C composite materials were obtained by thermal treatment with the mixture of ammonium molybdate, thiourea, melamine, and Pluronic F127 [52]. The ORR activity was enhanced as a result of the abundant active sites that were derived from rich-defects on the N-doped MoS2/carbon materials and the high electron conductivity.Transition-metal oxides (TMOs) consist of oxygen atoms with transition-metals from the d-block of the periodic table (such as Mn, Fe, Co, Ni, Ti, etc) [53]. The 2D layered structures have extra advantages such as high surface area and edge defects. Quasi-2D sheets of the TMOs are stacked together by van der Waals interactions. Transition metal atom possesses unpaired electrons and dangling bonds, leading to strong surface polarization of the TMOs [5,54]. The 2D TMOs also show multiple valence states, various metal oxidation states, and hence a tunable ORR activity. Since the thickness of the 2D TMOs is limited (always less than 2 nm), most of the low-coordinated transition-metal atoms are exposed on the surface and serve as the active sites [55,56].However, due to the poor conductivity, TMOs exhibit unsatisfactory ORR performance. Highly conductive support material with large surface areas such as graphene is used to form a 2D graphene/TMO heterostructure to address the conductivity issue [58]. There are interphase ligand effects and excellent interfacial interactions between TMOs and graphene due to oxygen-containing functional groups found on graphene. Thus, the use of graphene can improve the conductivity while the TMOs on the surface could maximize electrochemically accessible surface area. Therefore, this synergistic effect provides graphene/TMO heterostructure with a better ORR performance. As shown in Figs. 7 and 2 D manganese oxide nanosheets/graphene composite was synthesized by heating the graphene oxide with KMnO4 [57]. The nanopores were then introduced into the composites by heating with S powder. The improved ORR activity of the nanoporous MnO2 nanosheets was related to the active Mn3+/4+ sites, the large surface area, and oxygen vacancies, which reduced the kinetic barriers and facilitated the oxygen adsorption.Among different crystal structures of TMOs, the perovskite structures with a general formula ABO3 have attracted great attention recently because of their non-stoichiometric chemistry and variable crystal structures [60,61]. The 2D perovskite oxide consists of stacked layers with edge-sharing octahedra and shows a higher surface. As shown in Fig. 8 , three mechanisms are proposed to explain the active sites in 2D perovskite oxide for ORR. The oxygen can be absorbed on the catalyst surface by three configurations: (a) end-on adsorption, (b) side-on adsorption, and (c) bidentate adsorption. For both mechanisms in Fig. 8a and b, the metal sites are the active sites for ORR. However, the mechanisms in Fig. 8c show that the metal sites and oxygen vacancies can both serve as the active sites for ORR. In alkaline media, the OH group near the oxygen vacancies is displaced by one atom of O2 while the other atom of O2 fills the oxygen vacancy. In this case, the O–O bond is weakened, facilitating the further reaction to form OH−. The oxygen vacancies can also change the electronic structures of perovskite oxides, leading to the presence of redox couples and electron holes, benefiting the ORR by promoting charge transfer and increasing the electrical conductivity. Therefore, the ORR performance of perovskite oxides is largely related to oxygen vacancies.The 1D electrocatalysts are the ideal candidates with high catalytic activity for ORR, owing to their unique structural superiorities such as rapid electron and mass transport, large surface area, low vulnerability to aggregation, and dissolution. The catalytic activity of 1D electrocatalysts is not only dependent on their composition but also their geometry such as the number of walls, length, diameter, and chirality. Nanotubes, nanoribbons, and nanowires are the most common 1D electrocatalysts that have been widely used for boosting the ORR activity.Nanotubes are made of various materials that take the shape of tubes with their diameters measured in nanometers. Nanotubes are frequently used as ORR catalysts because of their inherent morphologic features such as large surface area and rapid mass transport as shown in Fig. 9 . Moreover, nanotubes show high ORR stability because they can inhibit carbon corrosion, Ostwald ripening, and aggregation under fuel cell operating conditions [62].Carbon nanotubes (CNTs) can be directly used as the metal-free ORR catalyst or as the catalyst support. However, the electrons in pristine CNTs are too inert for ORR. Surface engineering is essential to modulate the chemisorption energy of O2 and increase active sites thereby enhancing the ORR activity. Similar to graphene, the doping of heteroatoms is frequently used to change the surface properties. In particular, the N-doped CNTs are attracting much more attention [63,64]. This pioneering work was achieved by L. Dai's group in 2009 [6]. Vertically aligned nitrogen-containing carbon nanotubes were employed to be the ORR catalyst with improved catalytic activity and high stability compared to commercial Pt/C catalysts. The carbon atoms adjacent to N dopants show high positive charge density and thus serve as active sites. Moreover, N-doped CNTs are used to replace carbon black to build Pt nanoparticles supported by N-doped CNTs material [65,66]. Uniform Pt nanoparticles with an ultra-small size can be prepared on the N-doped CNTs and show improved ORR activity and high stability when compared with the carbon black and pure CNTs.To avoid the carbon-corrosion problem, self-supported Pt nanotubes were developed to address the activity and long-term durability of ORR [67,68]. These unique Pt nanotubes show preferential exposure of highly active crystal facets, easy electron transport, and inherent high stability. Also, the built of 1D Pt-based alloy and core/shell nanotubes structure has attracted remarkable attention [69,70]. Self-supported Pt-based nanotubes can be served as an alternative to carbon-supported materials and offer a broad materials library of noble metal structures.Nanoribbons are strips of 2D materials with a narrow width (less than 50 nm) in the plane. Graphene nanoribbons (GNRs) can be regarded as the unzipped CNTs. Compared with CNTs, GNRs show an open structure where the abundant edges are visible. The desirable catalytic activity towards ORR can be obtained by controlling the edge structures and chemical terminations [71]. However, the synthesis of the GNRs is not simple with opening the CNTs being one of the most promising approaches. As shown in Fig. 10 , the longitudinally unzipping of CNTs was used to get GNRs, and after an in-situ polymerization and thermal treatment process, nitrogen doping was achieved [71]. The TEM image showed the unzipped CNTs and the as-prepared GNRs demonstrate a thin elongated morphology. The N-doped GNRs showed an enhanced ORR activity with the dominant 4e- pathway. The active sites were the C atoms near the graphitic N and the pyridinic N at the edges.Compared to the CNTs and GNRs, the nanowires are relatively easier to be synthesized, especially for metal and metal oxide nanowires, such as MnO2, Co3O4 [62,72,73]. In particular, the building of the nanowire structure has been proved to be an efficient strategy to enhance the mass activity of Pt-based catalyst [74,75]. The morphology of 1D nanowires is beneficial to maintain high Pt-surface utilization and thus increases the electrochemical surface area (ECSA). As shown in Fig. 11 , because of the uncoordinated configuration on the surface, jagged Pt-based nanowires show the highest mass activity compared with other Pt nanostructures such as nanoparticles, nanoframes, etc. Moreover, the Pt-based nanowires display enhanced durability. The unique 1D nature of nanowires can be anchored on the support by multiple points. Therefore, Ostwald ripening and particle agglomeration are not prone to occur. Also, the Pt-based nanowire shows fewer defective sites and lattice boundaries vulnerable to Pt dissolution [76].The 0D materials are materials having all three dimensions on the nanometer scale. Various materials, such as nanoparticles and quantum dots, are the typical example of 0D materials. Usually, 0D electrocatalysts for ORR can be classified into metal nanoparticles, metal alloy nanoparticle, and metal complexes. Compared with bulk materials, the 0D nanostructures always show high surface free energy and low-coordinated atoms, and thus considerably improving their ORR activity [77–79]. However, 0D nanoparticles always suffer from aggregation. Therefore, the use of the appropriate support material is an efficient way to enhance their stability. In low dimensional materials, the van der Waals forces of the 1D and 2D materials allow them to strongly interact with 0D nanoparticles. Therefore, the 0D/1D or 0D/2D heterostructures are formed and show a synergistic effect on enhanced catalytic activity for ORR [80,81].When the size of the catalyst reaches the limitation (single atom) it can be considered as the 0D material as well. Single-atom catalysts (SACs) can maximize the atomic utilization, which indicates that ideally, every active single atom can serve as an active site [83–90]. Moreover, SACs show excellent activity and exclusive selectivity for ORR because of the unsaturated coordination environments and their unique electronic structure. As shown in Fig. 12 , a carbon‐supported defect‐anchored Pt SACs were synthesized and showed high ORR performance with high onset potential and the 4e− ORR process [82]. In an acidic H2/O2 fuel cell, the Pt SACs with carbon defeat showed super-high platinum utilization of 0.09 gPtkW−1. The individual Pt atom was anchored by the carbon defects and the Pt dispersion was improved. The Pt atoms anchored by four carbon atoms (Pt/C4) was supposed to be the active sites. Furthermore, the Pt alloy and the other metal SACs were also prepared [89,91,92]. The SACs catalyst should be strongly coupled to the support to enhance stability. Furthermore, the catalytical activity is highly related to the coordination enlivenment of the metal atoms [93]. So the support effect should be considered when using the SACs.M-N-C (M = Fe, Co, Cu, Ni, Mn, etc.) based ORR catalysts is another unique class of SACs. In 2009, breakthroughs were achieved by the Dodelet group at INRS that the activity of the Fe–N–C catalyst reached to the level of Pt/C catalyst [94]. Since then, M-N-C catalysts have attracted extensive research interests. When the active species of M-N-C catalysts are downsizing into single atoms level, it can maximum atom-utilization and thus enhance the intrinsic activity of catalysts [95,96,97]. As shown in Fig. 13 , similar to noble metal SACs, the catalytic performances of these atomically dispersed non-precious metal catalysts are also affected by their coordination environments and geometric configurations. To further elucidate the synthesis−structure−property correlations, more work should be done in the combination of precisely controlled synthetic process, in-situ structural characterizations, simulations, and intrinsic catalytic performance in the future.The different categories of low dimensional materials used in ORR have been reviewed. When the size of the materials is reduced to the nanometer scale (at least one of their dimensions), their electronic structure and their surface properties change drastically, resulting in unique physical and chemical properties due to the quantum size effect and large surface area. The electrocatalytic activity of low dimension materials is highly related to their electronic properties and nanostructures. The reaction mechanisms of low dimension materials are discussed by the combination of computational and experimental results. By reviewing the past progress, we expect to extend in-depth research on PEMFC catalysts of different dimensions. The low dimensional material can be used as both the catalysts and the supporting materials. Low-dimensional materials not only show a promising way to enhance the ORR activity but also provide new fundamental insight into ORR.Despite much progress, great efforts are still needed to further develop the low dimensional catalysts for ORR. For example, more theoretical and experimental studies are required to develop efficient strategies for alloying, hybridizing, doping (bi-doping and multi-doping), and combining different dimensional materials to further improve the thermodynamics and the kinetics of the catalysts. More systematical structure-property studies should be done to quantify the active sites and clarify the corresponding reaction mechanism of ORR. This calls for the employment of advanced characterization techniques, especially in-situ ones (such as XRD, XAS, IR, Raman), to investigate the ORR process based on low-dimensional catalysts in half cells and real fuel cell environments.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.We thank the support from the Doctor's Initial Funding of Guizhou Normal University, China (GZNUD[2019]25, GZNUD[2019]29), International Science and Technology Corporation Program of Guizhou Province, China (G[2013]7022), Youth Science and Technology Talent Development Project from Guizhou Provincial Department of Education, China (KY[2016]063), Fonds de Recherche du Québec-Nature et Technologies(FRQNT), the Natural Sciences and Engineering Research Council of Canada (NSERC), Institut National de la Recherche Scientifique (INRS), and Centre Québécois sur les Matériaux Fonctionnels (CQMF).

Developing highly efficient electrocatalysts to facilitate the sluggish cathodic oxygen reduction reaction (ORR) is a key challenge for high-performance fuel cells. Low-dimensional materials have attracted great attention recently because of their unique structure and properties. In this review, the application of zero-dimensional (0D), one‐dimensional (1D), and two‐dimensional (2D) materials in ORR are discussed and particular attention is given to the relationship between their structure and the ORR activity. Graphene-based materials, transition metal dichalcogenides, transition metal oxide, nanotubes, nanoribbons, nanowires, and single-atom ORR catalysts are introduced and classified by their geometric dimension.

Scientific consensus on the dangers of anthropogenic climate change has become near irrefutable in recent years, prompting a drive to reduce greenhouse gas emissions from a number of industries. Hydrogen gas is an important bulk chemical that is used primarily in the production of fertilizers and refining of petroleum crudes. It is also considered a strategic, clean alternative to fossil fuel for transport and grid electricity. The industrial production of H 2 as a bulk chemical is currently dominated by steam reforming (SR) of natural gas; a large scale, carbon intensive process that requires centralized production.Chemical looping steam reforming (CLSR) is a H 2 production technology designed to improve the provision of heat and reduce the burden of H 2 separation in the SR process, thereby reducing carbon dioxide emissions, improving efficiency and potentially allowing decentralized generation and distribution (Protasova and Snijkers, 2016; Adanez et al., 2012; Dueso et al., 2012). Decentralization of H 2 production via smaller scale generation processes can facilitate the use of widely available renewable biomass and waste feedstocks such as biogas, biodiesel, pyrolysis oils, ethanol and glycerol (Cheng et al., 2017). The CLSR process relies on the cyclical reduction and re-oxidation of a metal oxide, which is known as an Oxygen Carrier (OC). The OC is cyclically exposed to a reducing fuel/steam atmosphere and an oxidizing air atmosphere in one of two ways: (1) by alternating the fuel/steam mix and air feeds to a single fixed bed reactor or (2) by physically moving the OC between two fluidized bed reactors, each one fed with a continuous stream of fuel/steam or air. In the steam reforming half cycle, the OC is reduced to a catalytically active state by a hydrocarbon fuel/steam flow and serves as the catalyst for the steam reforming of the methane feed (SR, Eq. (1)) and the water gas shift reaction (WGS, Eq. (2)). (1) CH 4(g) + H 2 O (g) ⇆ 3H 2(g) + CO (g) Δ H 298K = 228  kJ mol − 1 (2) CO (g) + H 2 O (g) ⇆ CO 2(g) + H 2(g) Δ H 298K = − 33  kJ mol − 1 Carbon may be formed during this process causing deactivation of the catalyst. Carbon is produced by: thermal cracking (Eq. (3)); CO disproportionation (Eq. (4)); or reduction of CO (Eq. (5)). (3) CH 4(g) ⇆ 3H 2(g) + C (s) Δ H 298K = 75  kJ mol − 1 (4) 2 CO (g) ⇆ CO 2(g) + C (s) Δ H 298K = − 172  kJ mol − 1 (5) CO (g) + H 2(g) ⇆ H 2 O (g) + C (s) Δ H 298K = − 131  kJ mol − 1 Under the oxidation half-cycle the OC, and any carbon deposited during steam reforming, is oxidized by an air flow in highly exothermic reactions. These reactions provide heat throughout the reactor bed for the subsequent endothermic SR reaction (Protasova and Snijkers, 2016; Adanez et al., 2012).The choice of OC is therefore integral to the design of the CLSR process. An OC must present high reactivity for the reduction and oxidation reactions involved and maintain that reactivity over extended cycling. It must exhibit high oxygen transfer capacity and favourable thermodynamic properties. The OC should also be readily reduced by a number of fuel mixtures, and act as an effective catalyst for the SR reactions involved in the process. Resistance to attrition and agglomeration is important if the OC is to be used in a fluidized bed reactor, and the cost and complexity of synthesis of the OC must be kept to a minimum (Adanez et al., 2012; Dueso et al., 2012; Rydén and Ramos, 2012; Tang et al., 2015; Noorman et al., 2007; Gayán et al., 2009; Quddus et al., 2013). Moreover resistance to sintering induced degradation is important; low density fibrous mats with minimal numbers of contact points between the fibres may offer advantages in this context.Many carriers have been tested for use in chemical looping applications but supported transition metals have attracted the most interest (Adanez et al., 2012; Quddus et al., 2013). Metal oxides of iron, nickel, manganese, copper and cobalt have all been used, supported on alumina, zirconia, silica, titania, or bentonite (Adanez et al., 2012; Gayán et al., 2009; Ryu et al., 2001). Nickel when supported on alumina (Ni/ α -Al2O3 or Ni/ γ -Al2O3) is by far the most widely researched OC for syngas and H 2 production. This is thanks to strong catalytic activity for the SR reactions, acceptable oxygen transfer capacity and high redox reaction rates at the temperatures encountered in CLSR (Adanez et al., 2012; Tang et al., 2015; Noorman et al., 2007; Quddus et al., 2013; Zafar et al., 2005, 2006; Noorman et al., 2010; de Diego et al., 2008).The synthesis method used to produce an alumina supported OC must achieve close control of the distribution of the deposited metal oxide and minimize chemical reactions that degrade the catalytic activity (Dueso et al., 2012; Gayán et al., 2009, 2008; Mattisson et al., 2006). In the case of nickel oxygen carriers the formation of spinel nickel aluminate (NiAl2O4) reduces the amount of NiO (Gayán et al., 2009; Quddus et al., 2013; Mattisson et al., 2006; Dueso et al., 2010): NiO is the more reactive species in redox and in the SR and WGS reactions (Eqs. (1) and (2)) (Noorman et al., 2007; Mattisson et al., 2006; Dueso et al., 2010). It is thus imperative to devise a synthesis procedure that minimizes the formation of NiAl2O4 and therefore maximizes the proportion of Ni in the form of NiO.Metal support interactions dictate whether NiO or NiAl2O 4 is formed during synthesis; a strong chemical interaction will produce NiAl2O 4 whereas a weak interaction will produce NiO (Quddus et al., 2013; Zafar et al., 2006). Bolt et al. suggested that there are likely to be two major factors that affect these interactions (Bolt et al., 1998); the crystalline phase of alumina used and the temperature to which the OC’s are exposed. Bolt hypothesized that due to its “defect spinel structure” γ -Al2O 3 facilitates the entry of metal cations (e.g. Ni 2 + ) into its crystal lattice, and therefore γ -Al2O 3 would interact more strongly with the metal species and at a more rapid rate than α -Al2O3 during synthesis (Bolt et al., 1998). Additionally if exposed to temperatures above 900 °C, the γ -Al2O3 would undergo phase transformation to θ -Al2O3 and subsequently α -Al2O3 thereby promoting chemical reactions between metal and support (Bolt et al., 1998). These hypotheses were confirmed by a number of studies. Cheng et al. and Matisson et al. found that this strong interaction begins to occur when a Ni γ -Al2O3 catalyst is exposed to temperatures exceeding 600 °C and becomes increasingly prevalent above 800 °C (Mattisson et al., 2006; Cheng et al., 1996). In a series of investigations, Dueso et al. (2012) and Dueso et al. (2010). investigated the metal-support reactions between two nickel OC’s involving γ -Al2O3 and α -Al2O3 supports; the α -Al2O 3 OC produced a higher proportion of NiO than the γ -Al2O3 OC. This increased proportion of NiO when using α -Al2O3 was corroborated by the work by You et al. (2014).The addition of cobalt has been found to improve the performance of Ni/Al2O3 catalysts (Bolt et al., 1998; Hossain et al., 2007; Hossain and de Lasa, 2007). This has been attributed to preferential formation of CoAl2O4 over NiAl2O4 spinel due to faster kinetics of the CoO–Al2O3 reaction (Bolt et al., 1998). This results in a greater proportion of ‘free’ NiO, and therefore improved OC performance.These OC’s also offer advantages such as excellent redox cycling durability, strong suppression of carbon formation and stable SR performance attributed to the interactions between the two active metals (You et al., 2014; Hossain et al., 2007; Hossain and de Lasa, 2007; Jin et al., 1998).Although metal-support interactions have been widely reported, the effect of synthesis method upon homogeneity of metal dispersion is less well discussed. In many of the papers mentioned above, traditional incipient wetness impregnation techniques were used to deposit an active metal upon the support. In this technique, a metal salt solution is introduced to the support and the mixture then dried in order to bring about the deposition of the salt inside the pores of the support. A critical stage in this technique is drying; as the solvent evaporates, the salt is gradually concentrated and deposited. As a result a non-uniform deposition may be expected (Dueso et al., 2010; Zhao et al., 2000). Urea-decomposition is an alternative deposition route which offers superior control of the distribution of catalyst over a substrate. The method relies on the slow decomposition of urea above 90 °C in aqueous media which causes a uniform increase in pH throughout the solution (Tang et al., 2015; Gayán et al., 2009).In the present work low-density mats of Saffil fibre are investigated as a novel Ni/Co substrate material for CLSR applications. Urea decomposition and wet impregnation deposition routes have been examined. Saffil (catalytic grade) is a fibrous crystalline material consisting of γ -Al2O3 (95%) and SiO 2 (5%). The Saffil supported OC’s have been evaluated in terms of hydrogen yield and purity, and methane conversion in a fixed bed CLSR process over several redox cycles. The results demonstrate a significant difference in morphology of the deposited metal oxide coating and the corresponding steam reforming performance for OCs manufactured by controlled precipitation reactions as opposed to evaporative wet impregnation.The motivation for the work was to investigate the feasibility of alternative OCs in CLSR. If high activity in both steam reforming and oxygen transfer, retention of large specific surface area and thermal stability under chemical looping conditions can be demonstrated, the high void spaces in a loose assembly of Saffil supported OCs offers the potential to deliver sufficient residence times with low reactor load. Consequently new small and medium scale hydrogen production processes could be realized in future, catering to distributed feedstocks such as unconventional gases (e.g. shale wells) and biomass products (e.g. large farms, anaerobic digestion plants, biorefineries).The Saffil-supported OC’s were synthesized via three methods: Wet Impregnation (WI), Urea Decomposition Precipitation (DP) and Hydrothermal Synthesis (HT). Nickel and cobalt were used in varying quantities providing three different nickel to cobalt ratios. A total of nine OC’s were made (summarized in Table 1). A conventional pelletized 18 wt% NiO/ α -Al2O3 (18 wt% NiO, crystallite size 45 nm (Cheng et al., 2017)) was lightly ground into ∼ 200 μ m granules which could be accommodated in the bench top reactor. Starting reagents were Ni(NO3)2.6H2O (10.8756 g), and for mixed catalysts, Co(NO3)2.6H2O (0.2981 g or 0.9052 g) of purity 99.9% (Fischer Scientific). The cobalt loadings were either 0 wt%, 0.6 wt% or 1.8 wt%. The reagents were dissolved in 250 ml of distilled water, into which 10 g Saffil mat, cut into approximately 5 mm3 shreds, was introduced. The mixture was placed in a drying oven held at 100 °C for 6 h under an air atmosphere to impregnate the fibres with Ni/Co salts. The resultant material was then calcined at 600 °C for 4 h under an air atmosphere to decompose the Ni/Co nitrates to oxides.Experimental conditions for deposition of the catalyst on Saffil were based on those reported in the literature to be optimal for urea based homogeneous precipitation onto porous Al2O3 substrates (Zhao et al., 2000). The Ni and Co precursors in the same amounts as in WI were added to water, 250 ml, containing dissolved urea, CO(NH2)2, (Fischer Scientific) and Saffil to give a molar ratio of 1.7 Ni/Co:Urea. This mixture was placed in a beaker covered with a watch glass and placed in an oven; the temperature was raised from room-temperature at 10 °C min−1 to 95 °C for a dwell time of 24 h under an air atmosphere. This allowed for the degradation of urea to raise pH and precipitate Ni/Co hydroxides as per the reaction: (6) CO ( NH 2 ) 2 + 3H 2 O → 2OH − + 2NH 4 + + CO 2 ↑ [42] The mixture was then placed in a drying oven for 6 h under an air atmosphere at 100 °C to evaporate any remaining solution, and then calcined at 600 °C for 4 h.Starting reagents in the same proportions as for WI and DP but in smaller quantities (4.3502 g of Ni(NO3)2.6H2O; 0.1191 g or 0.3621 g of Co(NO3)2.6H2O) and a proportionate amount of Saffil were added to 100 ml water into which urea (CO(NH2)2,) was also added at a molar ratio of 1.7 Ni/Co:Urea. The mixture was transferred to a 125 ml Teflon lined autoclave (Parr Scientific) and placed in a furnace heated at 10 °C min−1 to 95 °C and held at this temperature for 24 h. The mixture was transferred from the hydrothermal reactor to a drying oven and heated at 100 °C for 6 h to evaporate any remaining solution, followed by calcination at 600 °C for 4 h.The morphology and microstructure of the OC’s were analysed using a Hitachi SU8230 high performance cold field emission scanning electron microscope (SEM). The samples were affixed to 10 mm aluminium stubs via carbon tape and all samples were carbon coated (thickness 2–3 nm) prior to SEM analysis. The materials heavily charged under the electron beam; therefore low voltage (1–2 kV) and current (0.9 μ A) settings were used to reduce this.The phase composition of the raw support and synthesized OC’s were analysed using powder X-ray diffraction (XRD). This was carried out using a Bruker D8 diffractometer with a Cu X-ray tube ( λ = 1 . 5406 Å). A step size of 0.025°, scan speed of 2 s per step and a range of 10–75° 2 θ were used. Background determination, peak identification and phase identification were conducted using XPert HighScore Plus analysis software. All samples were ground by pestle and mortar to form a powder prior to analysis.The nanostructures of the OC’s were analysed using an FEI Titan3 Themis (scanning) transmission electron microscope (TEM) operating at 300 kV and fitted with a high angle annular dark field (HAADF) STEM detector, a Super-X 4-detector energy dispersive X-ray (EDX) analysis system and a Gatan One-View CCD. The EDX analysis was performed using Bruker Espirit software (version 1.9.4). All samples were ground in ethanol and added by pipette onto 400 mesh holey carbon coated Cu TEM grids (Agar Scientific).Quantitative elemental analysis was conducted with the use of a Varian 240 s Atomic Adsorption Spectrophotometer (AAS). Samples were digested in 10 ml of 1:1 HCl and heated under reflux for 30 min. The solution was then diluted to 250 ml with distilled water. Nickel content was analysed using a wavelength of 352.4 nm and a slit width of 0.5 nm; cobalt was analysed at a 240.7 nm wavelength and a slit width of 0.2 nm.A Quantachrome Instruments NOVA 2200e was used to determine specific surface area (SSA) via the Brunauer–Emmett–Teller (BET) method. All samples were outgassed at 200 °C for three hours and analysed using N 2 as the adsorbate gas at a temperature in the region of 77 K.Chemical looping steam reforming experiments assessed the performance of the Saffil based OC’s in comparison to the conventional SR catalyst (granulated) in the CLSR process. The experiments consisted of a SR half-cycle and an oxidation half-cycle. The SR half-cycle exposed the various OC’s to CH4 and H 2 O under N 2 carrier gas, chemically reducing the OC’s as well as performing steam reforming. The carrier gas was used to enable elemental balances and to reach the minimum flow rates required by the analyser for accurate measurements. Oxidation was induced in the half-cycle of air-exposure.Appropriate conditions for chemical looping experiments were chosen by calculating chemical equilibrium compositions at assigned temperatures and pressures using the computer program CEA (Chemical Equilibrium with Application) developed by NASA Lewis Research Centre (McBride and Gordon, 1994). Conditions of 700 °C and a molar steam to carbon ratio (S:C) of 3 were found to maximize the equilibrium methane conversion and provide high hydrogen yield and purity while minimizing solid carbon deposition and steam flows.The reactor set-up used in the chemical looping experiments is shown in Fig. 1. CH4, H 2 and N 2 (BOC, purity 99.995%) and on-site compressed air were used as reactant gases. Mass flow controllers (Bronkhorst EL-FLOW, range 0.1–180 sccm) set the flow rate of CH4 and H 2 whereas N 2 and air were controlled by electric rotameters (Bronkhorst MASS-VIEW, range 0–2000 sccm). A programmable syringe pump was used to introduce distilled water into the heated zone of the reactor to satisfy the S:C required for each experiment. The 316 stainless steel (SS) reactor had an inside diameter of 13.2 mm and a length of 750 mm of which the bottom 500 mm was heated using a vertical tube furnace (Elite Thermal systems TSVH12/30/450) controlled by a temperature controller (Eurotherm 3216). Additionally a second K-type thermocouple was placed adjacent to the bottom basket and was used to monitor temperature in the catalyst bed to ensure the desired temperature was reached in each experiment. In each experiment the reactor was loaded with 2 g of OC. The bed volume was 12.5 cm3 and consisted of 10 stainless steel mesh baskets in which the OC’s were held. These baskets were placed in the reactor by resting on a small steel bar welded across the tube. Given the difference in bulk density between the two catalysts (conventional 18 wt% NiO was ∼ 1 g cm−3, the Saffil OC’s were ∼ 0.2 g cm−3) the conventional SR catalyst used for comparison purposes was diluted with silica sand to increase the bed volume to that of the Saffil catalyst.As the analysers used in the apparatus were highly sensitive to water, a condenser (jacketed heat exchanger using a cooling fluid consisting of 1:1 mix of water and ethylene glycol chilled to 2 °C) and a moisture trap (silica gel) were used to remove water prior to gas analysis. The outlet composition of the remaining gases was recorded in vol% every 5 s by an ABB Advanced Optima analyser using three modules; Uras 14 (CO, CO2 and CH4 measured by infrared adsorption), Caldos 15 ( H 2 via thermal conductivity), and Magnos 106 ( O 2 via paramagnetic analysis).All experiments were conducted at 700 °C with S:C = 3, a N 2 flow rate of 1000 sccm, CH4 flow rate of 111 sccm and liquid H 2 O flow rate of 0.25 sccm. The experiments were conducted via the following procedure. (1) The reactor was heated at 10 °C/min to 700 °C under a flow of 1000 sccm N 2 . (2) A mixture of 5 vol% H2 in N2 flowed through the catalyst bed at 1000 sccm to reduce the catalyst: reduction was inferred from a rise in the outlet concentration of H2 to 5 vol%). This ensured that the catalyst was in the reactive metal form rather than the non-catalytically active oxide phase. (3) The reactor was purged with N 2 until H 2 was no longer present in the outlet. (4) The initial SR cycle was then performed by switching on the water feed to pump water into the reactor. When water contacts the reduced catalyst as steam, H 2 is formed via oxidation of the catalyst (water splitting); when this was detected by the analysers, a flow of 111 sccm CH4 was added to the N 2 flow resulting in a total flow of 1111 sccm of reactant gas at 10 vol% CH4. This SR regime was continued for 20 min. (5) The CH4 and water flows were switched off and the reactor purged with N 2 for 5 min. (6) The N2 flow was switched off and the Air flow set to 1000 sccm for 5 min; in order to oxidize the catalyst. (7) The reactor was then purged for 5 min under 1000 sccm N 2. (8) The water and CH4 flows were switched on again and reduction of the catalyst occurred: the reduced catalyst then promoted SR and steps 5–8 were repeated until the catalyst had been reduced and oxidized 7 times.The outputs of the chemical looping experiments were defined by the following equations: (7) H 2 Y i e l d = n ̇ H 2 , o u t ∗ W ¯ H 2 n ̇ CH 4 , i n ∗ W ¯ CH 4 × 100 % (8) H 2 P u r i t y = x H 2 , o u t x C O , o u t + x CO 2 , o u t + x CH 4 , o u t + x H 2 , o u t × 100 % (9) CH 4 C o n v e r s i o n , X CH 4 = n ̇ CH 4 , i n − n ̇ CH 4 , o u t n ̇ CH 4 , o u t × 100 % (10) H 2 O C o n v e r s i o n , X H 2 O = n ̇ H 2 O , i n − n ̇ H 2 O , o u t n ̇ H 2 O , o u t × 100 % (11) r r e d , O C = n ̇ d r y , o u t x C O , o u t + x CO 2 , o u t + x O 2 , o u t − n ̇ H 2 O , i n X H 2 O Under both half cycles, molar flow rate of gas component i ( n ̇ i , o u t ) was defined as the dry gas mol fraction in the gas products x i measured online, multiplied by n ̇ d r y , o u t , the molar flow rate of total dry outlet gas from the nitrogen balance. ( n ̇ d r y , o u t ) was estimated using a nitrogen balance. During the methane/steam feed half cycle, methane conversion and the carbon balance error (where 100% results in a perfect balance) were estimated using a carbon balance. Water conversion was then estimated using a hydrogen balance. Finally, the rate of OC reduction ( r r e d , O C ) was estimated with the use of an additional oxygen balance. Total moles of metal oxide reduced over a given time were obtained from a time integration of the rate formula over the duration of the methane/steam feed. In-depth discussion of the elemental balances can be found elsewhere for a generic C n H m O k fuel (Pimenidou et al., 2010). For the air feed half cycle, a carbon balance would have determined if any carbon was present on the OC, as it would have oxidized to CO or CO2, but it will be seen that neither gases were detected during the air feeds. An oxygen balance then determined the rate of Ni oxidation, and from its integration over time, the moles of Ni oxidized during air feed. The nominal and chemically analysed metal loadings of the OC’s were generally in good agreement, Table 1, confirming the validity of the synthesis methods employed. X-ray powder diffraction showed NiO peaks and broad peaks of γ -Al2O3 (ICDD 00-010-0425) and SiO 2 (ICDD 01-080-6157) from the Saffil support, Fig. 2 (Peng et al., 2001). No CoO was detected in the mixed catalyst samples, as its concentration fell below the XRD detection limit. Important in the context of OC performance, any NiAl2O4 was also below XRD detection limits (< 5%).Scanning electron microscopy of uncoated Saffil indicated fibre diameters in the range 2– 5 μ m , Fig. 3a. Examination of the coated fibres showed the WI method was the least effective deposition route, giving erratic distributions of the metal oxide phase, Fig. 4a–b; some fibres were completely covered in a layer of densely packed metal oxide particles and there were examples of micron sized agglomerates on top of some regions of the coating. In terms of crystallite morphology, the WI route led to coatings composed of round, nodular crystallites with a wide size distribution, including some up to ∼ 100 nm in size, Fig. 4c. This variability is typical of a wet impregnation method as during drying a range of precipitation and growth conditions exist leading to non-uniformity in coating integrity and particle sizes (Neimark et al., 1981).The DP and HT synthesis routes each gave more uniform distributions of the deposited material with little or no evidence of uncoated areas. Fig. 4d and e show SEM images for DP. Micrographs of HT samples appear in Fig. 4g–i. Both coatings appeared as a series of folded layers, possibly representing a series of interconnected platelet crystallites, with edges normal to the surface of the fibres. The network of folded ‘ridges’ with solid walls up to ∼ 100 nm in width framed void spaces with lateral dimensions of up to ∼ 400 nm for DP and to 500 nm for the less densely folded HT coatings.The addition of Co seemed to have little effect on the morphology of the catalysts produced by any of the synthesis routes. Fig. 5a and b show typical SEM cross sections of a DP sample revealing a metal oxide layer thickness of ∼ 500 nm. The ridge structure extends from the outer surface to the fibre-interface indicating a progressive growth of the metal oxide layer perpendicular to the fibre surface.The BET specific surface areas of the uncoated Saffil fibre and the synthesized OC’s are shown in Table 1. The as-received Saffil fibres had a SSA of 106 m2 g−1. The deposition of NiO by the WI method reduced the SSA compared to the Saffil substrate. By contrast, the HT and DP OC’s generally exhibited a higher SSA, with one exception, the 18Ni DP sample (102 m2 g−1). The OC’s via the HT route produced the most consistent SSA results, with all HT OC’s showing an SSA in excess of 115 m2 g−1 and varying in SSA by only ∼ 5 m2 g−1. There appeared to be no effect on SSA from cobalt doping. The increase in measured SSA from DP and HT synthesis routes, relative to WI, is consistent with the porous coatings disclosed by SEM (Fig. 3).Transmission electron microscopy of uncoated Saffil fibres had indicated a crystallite size of approximately 4–6 nm, Fig. 3b. TEM of the 18Ni WI sample, Fig. 6a and b, revealed rounded particle profiles consistent with the nodular structures observed by SEM (Fig. 4b and c). The estimated average size of these crystals was 73 nm, with a range from 20 nm to 150 nm (based on 50 crystals measured). TEM images of the coatings in DP and HT samples revealed that the folds (wall thickness ∼ 100 nm) were composed of a sub structure of sub 10 nm crystallites, Fig. 6c–e. Some detached fibril-like nanoparticles were also evident. Corresponding HAADF STEM images and EDX maps are shown in Fig. 7. The location of the interface between fibre-and coating was located from EDX analysis (white and black arrows inset); from this the depth of Ni coating could be evaluated. The 18Ni WI sample indicated a deposited layer of < 200 nm whilst the coatings in the DP and HT samples were thicker at ∼ 300 nm and ∼ 400 nm respectively, consistent with the thickness estimated from SEM section (Fig. 5a). Analysis of electron diffraction patterns yielded d-spacings in agreement with the XRD analysis conducted earlier that was consistent with a NiO coating (refer to Figure S1 and Table S1).Aluminium was detected from TEM-EDX co-existing with Ni (and Co) even at the outermost surface of the metal oxide coating (Area 1, Fig. 7, Table 2 and Figure S2). Table 2, shows the semi quantitative analysis of Areas 1 and 2 in Fig. 7a–i. The quantity of Al at the surface of the deposited layer (Area 1) is least in the WI sample (15 at. %) and highest in the HT sample (35 at. %). Corresponding EDX spectra are shown in Figures S3–S5. The presence of Al at the surface of the deposited layer may arise from partial dissolution of Saffil during the chemical treatments involved in synthesizing the OC’s. It has been reported that addition of nickel and cobalt nitrate to water can cause some dissolution of pure powdered γ -Al2O3 supports due to the change in pH (Espinose et al., 1995). In the present work, the Al detected by TEM-EDX in the catalyst phase suggests some co-precipitation of Ni 2 + , Co 2 + and Al3＋ species ions occurred. This sequence may account for the very unusual ‘honeycomb’ structure to the DP and HT coatings on Saffil, as identified by SEM (Fig. 4). The Saffil fibres were only in contact with the precursor solution for a limited amount of time in the WI method (i.e. until the precursor solution had evaporated), thereby severely limiting any Saffil dissolution. The chemical conditions in DP, and more so HT synthesis methods are expected to promote increased Saffil dissolution.We tentatively ascribe the morphology of the coatings from DP and HT methods to a formation mechanism involving layered intermediate phases such as double layer hydroxides, DLHs. The soluble Al species and the presence of carbonate ions from urea decomposition, along with Ni/Co ions would provide a solution environment from which DLHs, [M 1 − x 2 + M x 3 + (OH)2][CO 3 x ∕2 ] ⋅ mH2O could form (Li et al., 2012; Feng et al., 2009; Xu et al., 2015; Christensen et al., 2006). The reactions may be represented as follows in Eqs. (12)–(15). (12) Al 2 O 3 + 3H 2 O + 2OH − → 2 [ Al(OH) 4 ] − (13) Ni 2 + + 2 O H − → Ni(OH) 2 (14) CO 2 + H 2 O → CO 3 2 − + 2 H + (15) ( 1 − x ) Ni(OH) 2 + xAl(OH) 4 − + x ∕ 2 CO 3 2 − → [ Ni 1−x Al 3 + x (OH) 2 ] x+ ( CO 3 − 2 ) x∕2 + 2xOH − Double layer hydroxides, formed in this case from Ni and Al ions (the latter from dissolution of Saffil fibres) as represented by Eqs. (12)–(15), typically crystallize with a characteristic platelet morphology due to their hexagonal crystal structure and low surface energy of ab crystal planes. For example Li et al. report platelet DLH crystals from urea derived synthesis of Ni/Mg catalysts on alumina particles that have broad similarities to the DP and HT product morphologies (Li et al., 2012). A high lateral growth rate of anisotropic LDH crystals, perpendicular to the Saffil support may in part explain the ridged, folded structure observed in DP and HT samples. However a future dedicated crystal growth study would be required to fully understand the growth mechanism.The outlet composition for the Saffil OC’s during the SR and the oxidation half cycle is shown in Fig. 8 for a DP sample, and is representative of all the OC’s tested. The inset graph in Fig. 8 shows the CH4 and H 2 O conversions and the rate of NiO reduction ( r red ,OC) derived from Eqs. (9)–(11) during the first 100 s of the SR half-cycle.The onset of OC reduction is clearly shown in the inset figure: r red ,OC increased after 10 s and peaked at 25 s coinciding with a rapidly decreasing negative H 2 O conversion, and an increasingly positive CH4 conversion. Negative H 2 O conversion signifies a production of H 2 O, this coupled with an increase in CH4 conversion implies that the global reduction of the OC through complete combustion (Eq. (16)) was favoured over other likely reduction reactions (Eqs. (17), (18) or (19)). After 25 s, an abundance of the reduced OC caused the initiation of the SR and WGS reactions (Eqs. (1) and (2)) thereby consuming H 2 O and thus increasing the H 2 O conversion. As the reduction neared completion, r red ,OC decreased, while H 2 O conversion increased and reached steady state as the SR and WGS reactions dominated. (16) CH 4(g) + 4NiO (s) ⇆ 4Ni (s) + CO 2(g) + 2 H 2 O (g) Δ H 298K = 135  kJ mol − 1 (17) CH 4(g) + NiO (s) ⇆ Ni (s) + CO (g) + 2 H 2(g) Δ H 298K = 213  kJ mol − 1 (18) CO (g) + NiO (s) ⇆ Ni (s) + CO 2(g) Δ H 298K = − 48  kJ mol − 1 (19) H 2(g) + NiO (s) ⇆ Ni (s) + H 2 O (g) Δ H 298K = − 15  kJ mol − 1 Fig. 8 shows no CO2 or CO peaks during the oxidation half cycle, a result shown across all the OC’s’ tested. This suggests the absence of complete or partial oxidation of solid carbon(Eqs. (20) and (21)), and implies a lack of carbon deposition during reduction and steam reforming, thus oxidation of the OC was favoured (Eq. (22)). (20) C (s) + O 2(g) ⇆ CO 2(g) Δ H 298K = − 393  kJ mol − 1 (21) C (s) + 0 . 5 O 2(g) ⇆ CO (g) Δ H 298K = − 110  kJ mol − 1 (22) 2Ni (s) + O 2(g) ⇆ 2 NiO (s) Δ H 298K = − 468  kJ mol − 1 Table 3 shows the maximum rate of OC reduction (maximum r red ,OC). The synthesis method did not have a pronounced effect on the maximum rate of reduction (Table 3), however the maximum r red ,OC increased slightly with decreasing Ni:Co ratio. All of the equations used to calculate these factors are available elsewhere (Pimenidou et al., 2010).The average H 2 purity and yield and CH 4 conversion over 7 CLSR cycles for all 9 OC’s is compared to the standard 18 wt% NiO catalyst and equilibrium values shows an error < 5% in the carbon balance between the molar flows of carbon in and out of the reactor across all experiments, indicating accuracy in the measurements and elemental balance analysis.All of the Saffil OC’s presented an improvement in terms of average CH4 conversion and H 2 yield over the conventional 18 wt% NiO SR catalyst, Fig. 9, Table 4. The 18Ni HT OC was the most effective, returning a 9.9% improvement in CH4 conversion and a 4.6% improvement in H 2 yield. The other Saffil OC’s improved CH4 conversion by between 3.4% and 8.5% while H 2 yield was improved by 3.1% to 4.2%. H 2 purity was largely unchanged. The CH4 conversion per cycle over the 7 CLSR test cycles is shown in Fig. 10 and indicates that the performance of the Saffil OC’s was stable and consistently superior to the conventional 18 wt% NiO catalyst (in granulated form), with HT having optimum performance, Table 4. The H 2 yield and purity performance over 7 cycles are shown in Figure S6 and show similar enhancements in the Saffil OC’s.There was no measurable difference in performance of OCs modified by cobalt. Additions of Co to Ni/Al2O3 have been reported by others to improve steam reforming performance by suppressing NiAl2O4 spinel formation (Hossain and de Lasa, 2007). A NiAl2O4 phase is normally formed during high temperature heat-treatments, thereby reducing the amount of available Ni catalyst. The absence of any effect of Co on steam reforming performance in the present work may relate to the milder heat-treatment schedules for OC synthesis and chemical looping which minimizes spinel formation—indeed no spinel phase was identified by X-ray or electron diffraction in any OC.In summary, the synthesis of Saffil supported OC’s via decomposition of urea (DP and HT) rather by simple liquid evaporation (WI) gives superior CLSR performance. The differences in the H 2 yield and CH4 conversion can be explained by the lower crystallite size, improved coating uniformity and more open texture. These factors are known to be important in conventional non-fibrous Ni/ γ -Al2O3 catalysts (Christensen et al., 2006; Ashok et al., 2008; Lucrédio and Assaf, 2006; Song et al., 2013). Reduced particle size results in a lower diffusional resistance to mass transfer between products and reactants in heterogeneous catalysis; the porous structure of the coating identified by SEM will aid gas diffusion, as will the relatively lose packing of the fibres. The role of any incorporated Al on the performance of the OC’s is uncertain at this stage.The low density Saffil OC’s developed in this project represent a different type of fibre-based catalyst to the densely packed fibre beds traditionally used for catalytic combustion of methane in space heating using platinum group metals (Radcliffe and Hickman, 1975; Trimm and Lam, 1980b,a) which are currently receiving widespread interest for used in microchannel reactors for the rapid production of H 2 for use in hydrogen fuel cells (Reichelt et al., 2014; Zhou et al., 2015).The performance comparisons highlight that for a given catalyst mass and volume, Ni/Co fibrous OC’s offer significant advantages over traditional catalysts in fixed bed CLSR. Furthermore the low density fibrous OC’s may provide other benefits for fixed bed processes. The low-density Saffil OC’s present a high surface area to volume ratio and high void fraction in the catalyst bed leading to an excellent compromise between mass transfer for the catalytic reactions and pressure drop through the bed. Additionally they are easily manipulated into various shapes and have a high thermal stability (Reichelt et al., 2014; Zhou et al., 2015; Reichelt and Jahn, 2017; Sadamori, 1999). These factors may prove beneficial to fixed bed CLSR processes, a research field in which fibre catalysts have not been applied. The lower thermal inertia offered by a less dense catalyst in conjunction with the reduction in diffusional resistance potentially allows fast and homogeneous heat provision during oxidation for the endothermic SR reactions allowing for efficient production of H 2 through the SR reactions. Further research is required to establish if these promising features can indeed facilitate a change in CLSR reactor geometry suited to small and medium scale hydrogen production.Ni/Co fibrous oxygen carriers (OC’s) have been fabricated utilizing low density mats of a polycrystalline alumina/silica fibre support (Saffil). The performance of oxygen carriers deposited on Saffil by urea homogeneous precipitation in chemical looping steam reforming was equal or better than the commercial 18 wt% NiO catalyst (in granulated form). The Saffil based OC’s synthesized in this work showed improved average values of methane conversion and hydrogen yield over the tested seven redox cycles. All of the OC’s were reduced by the fuel steam mixture and produced no solid carbon during reforming. Moreover the process advantages of fibre catalysts including high malleability, thermal stability, high surface to volume ratio and void fraction indicate that fibrous catalysts are a promising alternative to conventional catalysts in fixed bed chemical looping steam reforming. Future work will explore the long-term stability and durability of these fibrous OC and explore the kinetics and mass transfer properties associated with reduction and oxidation to further examine their suitability for CLSR in comparison to commercial pelletized catalysts. Nomenclature OC Oxygen carrier SR Steam reforming CLSR Chemical looping steam reforming WI Wet Impregnation DP Urea decomposition-precipitation HT Hydrothermal DP LDH Layered double hydroxide SEM Scanning electron microscopy TEM Transmission electron microscopy EDX Energy dispersive X-ray analysis BET Brunauer–Emmett–Teller SSA Specific surface area XRD X-ray diffraction HAADF High angle annular dark field AAS Atomic adsorption Spectrophotometry S:C Molar steam to carbon ratio SS Stainless steel n ̇ i , o u t dry moles in the reactor outlet of species i n ̇ i , i n dry moles in the reactor inlet of species i x i , o u t vol % in the reactor outlet of species i x i , i n vol % in the reactor inlet of species i W ¯ i Molar mass of species i X i Conversion of species i r r e d , OC Rate of OC reduction (mol s−1) Thanks to Jenny Forrester for her assistance with XRD analysis and to Stuart Micklethwaite for his assistance during SEM imaging. This work was supported by the EPSRC via the Low Carbon Technologies Doctoral Training Centre (EP/G036608/1) and the UKCCSRC via the Call 2 Capture Projects (UKCCSRC-C2-181). Further thanks must be extended to Jonathan Cross and Unifrax Ltd for supplying the CG Saffil material and for their input into this project.None.Supplementary material related to this article can be found online at https://doi.org/10.1016/j.egyr.2018.10.008.The following is the Supplementary material related to this article. MMC S1

Motivated by possible future applications in low pressure drop reactors for hydrogen production by fixed bed chemical looping steam reforming (CLSR), novel high porosity fibrous mats of aluminosilicate fibres have been investigated as a substrate for Ni/Co oxygen carriers (OC’s). When compared to granules of a conventional 18 wt% NiO steam reforming catalyst tested over seven redox cycles of CLSR, the fibrous OCs produced by homogeneous chemical precipitation routes increased the average methane conversion by up to 10% and hydrogen yield by up to 5%. All of the OC’s could be reduced by a CH 4 ∕ H 2 O mixture and produced no solid carbon during reforming.

Anion exchange membrane fuel cells (AEMFCs) have been investigated as a low-cost fuel cell alternative to proton exchange membrane fuel cells (PEMFCs) due to the potential use of non-platinum group metal (PGM) catalysts and the enhanced oxygen reduction kinetics on non-PGM catalysts under alkaline conditions [1–9]. The technical challenge for AEMFCs is that there is still much room for improvement in both performance and durability compared to PEMFCs. To obtain high performance and durability for AEMFCs, the chemical/mechanical stability and anion conductivity of the anion exchange membrane (AEMs) continue to be improved [10–18]. In recent years, various strategies have been devised to overcome these problems, such as microphase separation, cross-linking, and organic-inorganic composites [3,18,19]. This means that AEMs can be manufactured with higher anion conductivity while maintaining the same ion exchange capacity (IEC) [13,20]. In addition, Mandal et al. reported a high performance AEMFC with an anionic conductivity of 212 mS cm−1, cell performance of 3.5 W cm−2, and cell durability of more than 545 h at 80 °C due to the introduction of cross-linking and long alkyl spacers [21].Other important factors are the development of effective non-PGM catalysts. For anodes, mainly Ni-based non-PGM catalysts have been reported [4,6,22]. For cathodes, mainly non-PGM catalysts based on Fe, Co and other transition metals have been reported [7,23–29]. Among these, Hossen et al. reported the remarkable result that an Fe–N–C catalyst had the same performance as that of platinum-supported carbon (Pt/C), by combining the N–C materials used in the synthesis of catalysts and optimizing the ionomer content of the cathode catalyst layers (CLs) [26].In addition to the above components, water management at both the anode and cathode is also an important factor in AEMFCs. This is because water is produced at the anode by the hydrogen oxidation reaction (HOR) and consumed at the cathode by the oxygen reduction reaction (ORR) of AEMFCs. In other words, the AEMFC must provide sufficient water to maintain hydration of the AEMs and electrodes, without flooding the anode or drying the cathode. Also, water moves from the cathode to the anode, due to the electroosmotic drag associated with the movement of OH−, and moves from the anode to the cathode due to back-diffusing water [30–43]. Continuing from our previous paper [44], the present paper focuses on effective water management in the AEMFC. In the previous research, we discovered the current density-voltage (I–V) hysteresis phenomenon that accompanies an increase or decrease in the current density (CD) of a cell that uses a Fe–N–C catalyst as a cathode catalyst. This suggested that the hysteresis phenomenon resulted from the difference in the absorption capacity of liquid water in the cathode CL and affected the water supply at the reaction sites of the cathode. By more detailed Tafel slope component analysis, this I–V behavior can be characterized as a direct transition from kinetic control to combined gas-ion-water transport control combination, with a unique 8 × slope behavior, i.e., the intrinsic kinetic slope is multiplied by 8. These results also supported the importance of back-diffusing water.In the case of Fe–N–C being used in the cathode catalyst layer, one of the methods for improving water management performance, i.e., improving the flux of back-diffusing water utilized from the anode, is the use of a thin electrolyte membrane. It has been reported that thinning the membrane shortens the distance of anion conduction and increases the flux of back-diffusing water, improving AEMFC performance [2,3,5,9].In addition, Dekel et al. and Yassin et al. reported that by reducing the membrane thickness from 28 μm to 10 μm and increasing the water diffusion coefficient of the membrane, not only the power generation performance was improved, but also the cell life was extended, based on model calculations [45,46]. Also, Jiang et al. reported that the cell performance was improved by hydrophilizing the surface of the membrane via the creation of hydrophilic functional groups using a plasma [47].In addition, for using Fe–N–C catalysts, the active reaction sites of the cathode must have a balanced hydrophilicity/hydrophobicity in order to be accessed by both oxygen gas and water for the ORR. To achieve high ORR performance for Fe–N–C catalysts, it is important to have a hierarchical pore structure with macro/mesopores for reactant access to the Fe–N–C active sites present in the micropores [48,49]. We reported that Fe–N–C catalyst particles synthesized using carbon with a high specific surface area carbon, specifically, refluxing Black Pearls (BP-2000, Asian-Pacific Specialty Chemicals Kuala Lumpur) with nitric acid increased the number of mesopores, and show ORR activity equivalent to that of a Pt/C [28]. In addition to the catalyst layer structure, the interfacial interaction between the ionomer and the active sites of the catalyst is also an important factor. Using nanoparticle silver catalysts and quaternary ammonium functionalized triblock copolymer ionomers, Buggy et al. reported that ionomer-catalyst interactions may also have a significant effect on ionomer water uptake [50]. Santori et al. also reported, by means of Fe–N–C catalysts and RDE evaluations, that low ionomer gas permeability has a negative impact on catalyst bed performance because of the increased reactant consumption (O2) per active site for catalysts with low site densities [51].In this paper, we report two approaches for the improvement of the water management ability in order to suppress the I–V hysteresis phenomenon [44]. To fully understand the effect of AEMs on suppressing this AEMFC performance hysteresis phenomenon, it is necessary to test AEMs with various structures such as hydrocarbon-based membranes, cross-linking, and microphase separation. In this study, an electrolyte membrane (quaternized poly(arylene perfluoroalkylene), QPAF-4), which was developed by the University of Yamanashi and Takahata Precision Co., Ltd. [13], was used as the electrolyte membrane and binder. The QPAF-4 AEM with the molecular structure shown in Fig. S1 is suitable for this study because of its excellent microphase molecular structure, high gas permeability, high alkaline stability and high membrane mechanical strength. QPAF-4 is also soluble in methanol, which is highly volatile and minimizes the effect of the catalyst layer preparation. The QPAF-4 membrane was also compared with a cell using an A201 membrane (Tokuyama Corp.), which is widely used in AEMFCs. First, we sought to increase the flux of back-diffusion water from the anode by both thinning and hydrophilizing a QPAF-4 AEM, and subjecting this membrane to hydrophilic treatment. This study led us to realize again the importance of water transport at the interface between the membrane and the catalyst layer. Second, to develop a high ORR activity catalyst layer with a macro/mesopores layer accessible to both oxygen gas and water, we sought to improve the water supply to the reaction active sites by suppressing the absorption of water at the cathode by using a specially developed Fe–N–C catalyst provided by the Changchun Institute of Applied Chemistry (CIAC), Chinese Academy of Sciences [28]. This Fe–N–C is a catalyst with a high specific surface area of 1200–1500 m2 g−1 in Brunauer-Emmett-Teller (BET), and 30 nm Fe nanoparticles are uniformly present, based on transmission electron microscope (TEM) images. In addition, the I–V results reported for high-power AEMFCs, as shown in many reports, have typically been obtained for high gas flow rates, e.g., 1 L min−1. This is quite low in terms of hydrogen and oxygen flow rate utilization, 3% for hydrogen (stoichiometric ratio = 33.3) and 1.5% (stoichiometric ratio = 66.6) for oxygen at 1 A cm−2 [21,37,38], which are difficult to be implemented in practical fuel cell systems. Our cells have been operated at the low flow rate of 0.1 L min−1 (30% hydrogen utilization and 15% oxygen utilization), which is close to the actual operating conditions of an AEMFC. Chen et al. have proposed a new normalized efficiency metric of W cm−2 divided by the flow rate, W s cm−2 L−1 [52]. The cell using QPAF-4 in our previous paper achieved values of 120–198 W s cm−2 L−1, which is competitive with the performance reported in many AEMFC papers [21,37,38]. On the other hand, the importance of water balance is not new, as it has been mentioned in many previous reports [30–43]. However, we demonstrate and propose that the water absorption of the catalyst and the impediments to water transport at the membrane catalyst layer interface make this important water management challenge more pronounced, and thus a more serious problem, at the lower flow rates and lower pressure drops that are easier to achieve in practical systems.QAPF-4, which was used as an electrolyte membrane and binder, was synthesized based on the synthesis procedure of Ono et al. [13]. The catalyst inks for the anodes were prepared with Pt catalyst supported on carbon black (Pt/CB: TEC10E50E, Tanaka Kikinzoku Kogyo, K. K.), methanol and pure water by stirring for 30 min and use of a planetary ball mill containing 20 zirconia beads with a diameter of 5 mm. Subsequently, 5 wt% QPAF-4-MeOH (IEC = 2.0 meq g−1) binder solution was added to the slurry, and the mixture was further stirred with a planetary ball mill for 30 min. The weight ratio of QPAF-4 binder to support carbon was 0.8. In the same way, the catalyst inks for the cathodes were prepared with the Fe–N-Cc catalyst (synthesized and supplied by the CIAC from Black Pearls (BP-2000, Asian-Pacific Specialty Chemicals Kuala Lumpur)) and Fe–N-Cp catalyst (PMF-011904, supplied by Pajarito Powder), 5 wt% QPAF-4-MeOH binder solution (IEC = 2.0 meq g−1), methanol and pure water by use of a planetary ball mill. The weight ratio of QPAF-4 binder to support catalyst was set to 0.43. These catalyst inks were directly sprayed onto the microporous layers (MPLs) of the gas diffusion layers (GDLs) as the anode (W1S1010, Cetech Co., Ltd.) and cathode (29BC, SGL Carbon Group Co., Ltd.) by the pulse-swirl-spray (PSS, Nordson Co., Ltd.) technique to prepare the gas diffusion electrodes (GDEs). The electrode areas were 4.41 cm2, the Pt loading of the CL was 0.20 ± 0.02 mgPt cm−2, and these Fe–N–C loading of CLs were 0.50 ± 0.05 mgcat. cm−2. The prepared GDEs were immersed in 1 M KOH 80 °C for 2 days before measurement to ion-exchange to the OH⁻ form. Similarly, the QPAF-4 electrolyte membranes (IEC = 2.0 meq g−1, ca. 10 and ca. 30 μm) were also immersed in 1 M KOH aqueous solution at 80 °C for 2 days before measurement. To remove excess KOH aqueous solution, the GDEs and the electrolyte membranes were sandwiched between Kim Towels (Nippon Paper Cresia Co., Ltd.). Next, the GDEs and the electrolyte membranes were immersed in ultrapure water for approximately on hour, being constrained so as not to float, and then sandwiched between Kim Towels again to remove the ultrapure water. After the KOH was thoroughly removed, each set of GDEs and QPAF-4 membrane was pressed together in-cell to form the membrane electrode assembly (MEA) without hot pressing. The MEAs were sandwiched between two single serpentine flow graphite plates and 200 μm silicone/poly(ethyl benzene-1, 4-dicarboxylat/silicone gaskets (SB50A1P, Maxell Kureha Co., Ltd.) and were fastened at 10 kgf cm−2 with four springs. For the reversible hydrogen electrode (RHE), a 5 mm diameter disk was cut from the Pt/CB 29BC GDE prepared above and applied to the membrane on the cathode side. The hydrogen source for the RHE was the anode outlet, supplied through a heated (90 °C) gas line. The CL surface of the RHE was in contact with the electrolyte membrane in the cell, and the GDL surface was in contact with gold wire, which was connected to the anode and cathode by terminals through a multi-input data logger (NR-500, KEYENCE Corp.) and a high voltage measurement unit (NR-HV04, KEYENCE Corp.), respectively, and the polarizations of the anode and cathode were measured. Fig. S2 shows an overview of the cell with reference electrode.The cell voltages (V) as a function of current density (I) were measured with hydrogen and oxygen at 60 °C at various pressures. Hydrogen and oxygen gases were supplied to the anode and the cathode at a flow rate of 100 mL min−1. The flow rates of all gases were controlled by mass flow controllers. These gases were humidified at 100% relative humidity (RH) by bubbling through a hot water reservoir. The I–V curves were galvanostatically measured under steady-state operation by use of an electronic load (PLZ664WA and KFM2150, Kikusui Electronics Corp.) controlled by a measurement system (fuel cell characteristic evaluation device, Netsuden Kogyo Corp.). The measurement times in the direction of increasing current were 1 min up to 0.02 A cm−2, 3 min up to 0.1 A cm−2, 5 min up to 0.2 A cm−2, 7 min up to 0.3 A cm−2, and 10 min up to 1.0 A cm−2. The measurement times in the direction of decreasing current were just half those used for increasing current. Also, since resistances are difficult to measure with alternating current (AC) impedance at current densities below 0.1 A cm−2 (KFM2150, Kikusui Electronics Corp.), they were measured with a 1 kHz external resistance meter (MODEL 3566, Tsuruga Electric Corp.) For current densities of 0.1 A cm−2 or more, the membrane resistance was measured by AC impedance.An ozone/UV surface treatment device (EKBIO-1100, EBARA JITSUGYO Co., Ltd.) was used to hydrophilize the QPAF-4 electrolyte membrane. Ozone was generated by UV lamps (wavelengths of 245 nm and 185 nm) installed at the top of the chamber. The chamber temperature and water temperature were set at 40 °C, and the QPAF-4 membrane was placed 90 mm below the chamber ceiling. In an air atmosphere for 10 min, the back and front sides were turned over and hydrophilized twice in total. The ozone concentration in the chamber became steady after 2 min from the starting hydrophilization and showed a level of about 50 ppm.For the surface-conduction analysis by current-sensing atomic force microscopy (CS-AFM) on the QPAF-4 membranes of ca. 10 μm thickness, the GDE was first prepared by spraying a catalyst ink containing the Pt/CB and QPAF-4 binders (IEC = 2.0 meq g−1) as a binder on the 29BC GDL using the PSS technique in the same manner as described above. The weight ratio of QPAF-4 binder to support catalyst was adjusted to 0.8. The Pt loading of the electrodes was 0.2 ± 0.02 mgPt cm−2. The GDE was subsequently immersed in 1 M KOH aqueous solution at 80 °C for 2 days and then immersed in a saturated aqueous solution of NaHCO3 aqueous solution at 40 °C for 2 days, and then dried to ion-exchange it to the HCO3 − ion form. The membrane and the GDE in the HCO3 − ion form were pressed at 10 kgf cm−2 and 140 °C for 3 min. The ozone/UV surface treatment was carried out on the QPAF-4 membrane surface attached to the GDE.The CS-AFM setup was prepared according to prior literature [53–59] making use of a commercial AFM system (SPM-5500, Agilent) equipped with a home-made environmental chamber under a purified (CO2-free) air atmosphere at 40 °C and 70% RH. A Pt/Ir-coated silicon tip (NanoWorld) was used for the CS-AFM measurements. The morphological and current images were obtained in the contact mode, with a contact force of 5 nN on the membrane surfaces and a tip bias voltage of −2.0 V [59]. Before measurements, humidified air was supplied to the environmental chamber (dead volume = 500 mL) at 100 mL min−1 for 2 h. During the AFM measurements, the flow rate was reduced to 10 mL min−1.To ensure that there was no tip damage of the surface and that it was free of airborne impurities such as dust, we obtained the first and second scanned images at the same position for each CS-AFM measurement. The tip bias voltage was kept at −2.0 V during the image acquisition.The electrochemical measurements were carried out with an Automatic Polarization System (HZ-5000, Hokuto Denko Co.) using the rotating disk electrode (RDE) method. A Pt mesh was used as the counter electrode and the RHE was used as the reference electrode. The working electrode was prepared as follows: the catalyst ink was prepared by mixing 1 mg of each catalyst with 0.025 mL of ultrapure water and 4.975 mL of ethanol via ultrasonication. A 1 μL droplet of this ink was deposited onto a glassy carbon electrode (GC area = 0.283 and 0.196 cm2, Naito Rika Co. Ltd.) using a 1-μL syringe. The amounts of the Pt/CB ((4 μgPt cm−2, TEC10E50E), Fe–N-Cp (11 μg cm−2) and Fe–N-Cc (11 μg cm−2) catalysts was controlled by the number of drops. Subsequently, Nafion diluent obtained by diluting a 5 wt% Nafion solution with 75 vol% of ethanol was pipetted on the electrode surface, yielding an average thickness of 0.03 μm. The electrolyte solution, 0.1 M KOH, was prepared from reagent grade chemicals and ultrapure water. Polarization curves for the ORR were recorded in O2-saturated electrolyte at 5 mV s−1 (positive potential scan) and several rotation rates from 1000 to 2500 rpm.The wettability of QPAF-4 membranes with/without hydrophilization and CL surfaces were investigated by contact angle measurement (DM-501, Kyowa Interface Science Co., Ltd.). Reagents (wetting tension test mixture, Kanto Chemical Co., Inc.) having different surface tensions of 30, 40, 50, and 73 mN m−1 were pipetted on membranes and CL surfaces, and the contact angles were measured. The contact angle was evaluated by analysis software (FAMAS, Kyowa Interface Science Co., Ltd.) that can be used for the sessile drop method and a half-angle method. To reduce the measurement error, the measured value was calculated by averaging the values of 6 times measured the front and the back sides.We applied N2 adsorption to investigate the pore structures of the CLs. The N2 physisorption experiments were measured at 77 K by use of an automated gas sorption analyzer (Autosorb iQ, Anton-Paar GmbH). All the samples (0.1 g or more) were degassed at 60 °C for 24 h in an onboard degassing port, prior to the adsorption experiments. The N2 adsorption measurements were conducted in the P/P0 range 0.025–0.997, where P represents the gas pressure and P0 the saturation pressure. The specific surface areas and pore volume distributions were calculated by the Brunauer-Emmett-Teller (BET) and Barrett-Joyner-Halenda (BJH) methods, respectively. To obtain precise measurements of the values of the CLs and avoid the influence of the values of GDLs, catalyst-coated membranes (CCMs) were prepared by coating the catalysts on the QPAF-4 electrolyte membrane by the PSS method. The CCMs (6 cm × 6 cm) were divided into three parts and placed in the measurement cell. The specific surface area and pore size distribution were calculated from the obtained adsorption isotherm curves.We also applied water vapor adsorption to investigate the pore structures of the CLs. The experiments of water vapor physisorption were measured at 60 °C with water vapor sorption analyzers (Vstar, Anton-Paar GmbH). All the samples (0.1 g or more) were degassed at 60 °C for 24 h in an onboard degassing port prior to the adsorption experiments. The values of water vapor adsorption were measured in the P/P0 range 0.05–0.95. In the case of the CLs, these were formed on a PP film by the PSS method and were removed and filled into the cell.In our previous studies [44], it has been clarified that the supply of water to the catalytic active sites has a great influence on the hysteresis of the power generation performance of the cathode in the AEMFC. Therefore, in order to increase the supply of generated water to the reaction active site of the cathode, we sought to reduce the membrane thickness and increase the flux of back-diffusing of water.In Fig. 1 (a–e), the changes in the cell polarization curve, ohmic resistance, and anode cathode polarization curve using Pt/CB CL as the cathode CL are shown. As the thickness of the electrolyte membrane decreased from 33 μm to 11 μm by comparisons between circles and triangles in Fig. 1(a) and (b), contrary to our expectations, there was a large difference in the potential between increasing and decreasing current, which was the phenomenon we have termed I–V hysteresis. These I–V hysteresis tended to be larger for thinner membranes, as shown in Fig. S3. In addition, this hysteresis was not a tendency peculiar to the QPAF-4 membrane and was also observed in cells using the A201 membrane (Tokuyama Corporation) (Fig. S4). Based on the comparison in Fig. 1(c), the ohmic resistance decreased from 0.088 Ω cm2 to 0.066 Ω cm2 at 1.0 A cm−2 as the membrane became thinner.We hypothesized that one of the causes of I–V hysteresis was water transport at the interface between the membrane and the cathode, and therefore, to improve the transport, we hydrophilized the surface of the 11 μm membrane. In Fig. 1(a) and (b), the hysteresis observed for the hydrophilized 11 μm membrane decreased, and the ohmic resistance of the hydrophilized membrane in Fig. 1(c) was 0.080 Ω cm2 at 1.0 A cm−2, which was higher than that for the unmodified 11 μm membrane and less than that for the 33 μm membrane.In order to investigate the cause of the I–V hysteresis in detail, the polarization contributions of the anode and cathode were measured independently. As shown in Fig. 1(d and e), this I–V hysteresis occurred predominantly on the anode side and disappeared after hydrophilizing the membrane. The increase of I–V hysteresis for the 11 μm membrane with increasing CD is considered to be due to flooding by generated water in the anode. Fig. 1(f–l) show the results of investigating whether the effect of the hydrophilization is effective even in the case of cells using the Fe–N-Cp cathode catalyst. The cell using the Fe–N-Cp catalyst and the 11 μm QPAF-4 membrane showed hysteresis at both anode and cathode. However, as seen in Fig. 1(f) and (g), the I–V hysteresis decreased even in the cell using the Fe–N–C CL when the hydrophilized 11 μm QAPF-4 membrane was used. Regarding the ohmic resistance in Fig. 1(h), the resistance of the hydrophilized 11 μm membrane was 10% (increasing CD) and 20% (decreasing CD) smaller than that of the 33 μm membrane at 0.5 A cm−2 and increased significantly with increasing CD, in contrast with the case of the Pt/CB cathode in Fig. 1(c). These phenomena can be explained by the fact that, in the case of the Fe–N-Cp cathode, water absorption was larger than that for Pt/CB and this led to the increasing ohmic resistance. In Fig. 1(i), I-V hysteresis did not occur in the anodic polarization, and, in Fig. 1(j), another instance of I–V hysteresis, in this case, for the Fe–N-Cp cathode polarization with the 33 μm membrane, was suppressed by use of the hydrophilized 11 μm membrane. Therefore, it was found that the improvement of the water transport at the interface between the membrane and the cathode by use of the hydrophilization treatment on a thin electrolyte membrane is effective against the hysteresis phenomenon of both the anode and the cathode.In Fig. 2 and Fig. S5, the change in contact angle for various reagents with different surface tensions used for each MeOSO3 - form membrane showed that the hydrophilicity of the membrane surface depended on the thickness, and those for the hydrophilized 11 μm QPAF-4 membrane were an average of 21° lower than those for the non-hydrophilized QPAF4 membrane. As for the reason for the change in contact angle depending on the thickness of the AEM, when the membranes were dried in the same casting method on a substrate, the hydrophobic substructure would be more stable and thus more prevalent at the membrane/vapor interface, and the top surface were more hydrophobic. In a review of proton-based PFSA membranes, Kusoglu and Weber reported that at low hydration levels, the membrane/vapor interface layer was highly hydrophobic and resistive, and the fluorocarbon chains were oriented parallel to the membrane/vapor interface, which in some cases limits the mass transport of water and ions [63]. When imagined with a minimal molecular structure, i.e., two molecular chains, the ionic groups of the QPAF-4 AEM, which are hydrophilic, face inward, and the main chain structure, which is hydrophobic, faces outward. We estimate that the smaller the thickness of the bulk membrane, the more the hydrophobization would be enhanced. This result is also in agreement with the results of the OH⁻-form membrane, which is the same counter ion as that for the fuel cell evaluation. As shown in Fig. S5, the contact angle also increased with decreasing membrane thickness, but the contact angle decreased for all membrane thicknesses after the membranes were subjected to hydrophilization. Hydrophilization of the membrane surface reduces impediments to water movement at the membrane-cathode interface. These results support the result of promoting the movement of water by the hydrophilization of the membrane surface and suppressing the hysteresis of the I–V performance. Fig. 3 shows the results of CS-AFM analysis to investigate the effect of the hydrophilization on the 10 μm OPAF-4 membrane on the current distribution at the interface between the membrane and the cathode. There were no clear differences of more or less than 10 nm in the topography shown at the upper part of Fig. 3(a), but in the current image in the lower part, it was found that the hydrophilization treatment clearly improved the uniformity of the current distribution. As shown in Fig. S6, similar microscopic results were confirmed in other parts of the membrane surface. These results can also be clearly confirmed in the difference in the current profile on the lines shown in Fig. 3(b). The results of the effect of the hydrophilization of the membrane surface on the osmotic pressure in liquid water showed similar changes in Fig. S7. Therefore, from these analysis results, we determined that hydrophilization does not improve the diffusivity of water inside the membrane but contributes to the improvement of the surface anion conduction and water pathways. The results of the membrane improvement approach show that hydrophilization of the QPAF-4 membrane surface was more effective than thinning in increasing the amount of back-diffusing water produced at the anode. We can conclude that a membrane with a hydrophilic surface can also function with the effect of reducing the membrane thickness. This effect was confirmed in the power generation performance results shown in Fig. 1, even for the Fe–N-Cp catalyst.As a second approach to suppress the I–V hysteresis, we investigated the use of a new non-PGM catalyst that would have improved performance in comparison with that of the Fe–N-Cp material, used in the previous research [44], by using the Fe–N-Cc catalyst developed at the CIAC Laboratory. Fig. 4 shows the pore size distributions calculated by the BJH method and N2 adsorption isotherms of these Fe–N–C catalysts and Pt/CB catalysts, respectively. In Fig. 4, in the Fe–N-Cc CL, the hysteresis of the nitrogen adsorption/desorption isotherm is smaller than that for the Fe–N-Cp CL, suggesting that the pore volumes of the ink bottle structures [60,61] were also lower. In addition, the isotherm on the low-pressure side near P = 0 had a larger slope than those for the other CLs, and the cumulative pore distribution and the log differential pore volume distributions calculated from the adsorption/desorption isotherms indicated many pores of ∼5 and 40 nm. The Fe–N-Cc is synthesized from the high specific surface area carbon Black Pearls [28], and it is clear that the characteristics of the carbon material also affect the gas adsorption of the catalyst layer. Fig. 5 shows water vapor adsorption isotherms and contact angle changes of these Fe–N–C catalysts and Pt/CB catalysts, respectively. Comparing the water vapor adsorption isotherms of all of the CLs in Fig. 5(a), on the low-pressure side near P = 0, where the first layer of water molecules was adsorbed, the amount of water vapor adsorbed by the Fe–N-Cc CL was as low as Pt/CB CL. On the high-pressure side near P = 1, the Fe–N-Cc CL had a similarly large adsorption amount as the Fe–N-Cp CL, and the water adsorbed value increased with increasing pressure applied. These results indicate that water was not able to enter into the pores of the Fe–N-Cc CL despite the larger pore volume than that of the Pajarito CL at low pressure, close to a practical fuel cell condition. Fig. 5(b) shows that the Fe–N-Cc CL was more hydrophobic than the Fe–N-Cp CL from the contact angle measurements of the CLs and the hydrophobicity was comparable to that of the Pt/CB CL. The high hydrophobicity of the Fe–N-Cc material is thought to be attributed to that of Black Pearls, a furnace black carbon similar to Kejten Black, the carbon support material for Pt/CB.Polarization performances for the Fe–N-Cc (11 μg cm−2), Fe–N-Cp (11 μg cm−2) and Pt/CB ((4 μgPt cm−2, TEC10E50E) catalysts using the RDE technique are shown in Fig. 6 and S8. The amounts of catalysts for the RDE were prepared at the same ratio used for the cathodes of the MEAs. The plateau currents for the Fe–N–C catalysts did not reach the limiting current density for the 4-electron ORR, because the amounts of the catalysts were very small, and thus there was a limitation due to mass transport between catalyst agglomerates on the electrode surface [62]. Nevertheless, the ORR activity can be compared in terms of the half-wave potentials: the Fe–N-Cc catalyst exhibited 0.81 V at 0.05 mA cm−2, which was 0.18 V higher than that of the Pajarito catalyst but was 0.084 V lower than that of Pt/CB with the same catalyst ratio. Fig. 7 shows the cell performance using each cathode CL with a 30 μm QPAF-4 membrane. Table 1 shows the voltages at three different current densities, 0.2, 0.5 and 0.8 A cm−2. The voltage differences of the I–V hysteresis at 0.2 A cm−2 and 0.5 A cm−2 for the cell using the Fe–N-Cc CL were 0.05 V and 0.03 V, respectively, which were smaller than the corresponding values, 0.20 V and 0.15 V for the cell using the Fe–N-Cp CL, and the performance at current densities below 0.5 A cm−2 was comparable to that for the Pt/CB CL. In addition, the ohmic resistance of the Fe–N-Cc cell was 0.090 Ω cm2 at 0.5 A cm−2, which was similar to that of the Pt/CB CL.The reason why I–V hysteresis was suppressed in the Fe–N-Cc CL can be rationalized as being due to its high hydrophobicity (explained in Section 3.2.1), so that the volume of water absorbed by the catalyst is reduced, and the water required for the reaction is secured. In addition, the reason why the ohmic resistance of the Fe–N-Cc cell was low is considered to be that there was little water absorbed by the catalyst, and the water content of the membrane increased. However, in the higher current density region, for example, 0.8 A cm−2, the voltages of cell using the Fe–N-Cc CL decreased to 0.32 V compared with the case of 0.43 V using the Pt/CB CL and were nearly the same as those using the Pajarito CL (during decreasing current density). The reason for the performance decay at high current density is considered to be due to the fact that the catalyst layer was 3 times thicker than that of the Pt/CB, which was nearly the same as that of the Pajarito catalyst (Fig. S9), and thus the diffusion overvoltages for oxygen and water required for the cathode reaction increased. On the other hand, with a thinner CL, there are fewer Fe active sites, and more oxygen is needed to penetrate the catalyst layer through the ionomer [51]. This may affect the performance and the use of ionomers with high gas permeability should be considered.The insights gained from the two parts of the present study are summarized as schematic images of water management to suppress I–V hysteresis by use of the surface-hydrophilized electrolyte membrane in Fig. 8 (a) (Section 3.1) and by the use of the Fe–N–C catalyst with low water absorption in Fig. 8(b) (Section 3.2). Fig. 8(a) shows that we hydrophilized the surface of the QPAF-4 membrane, and this approach led to an improvement in performance due to the increased amount of water available for the ORR. By making the electrolyte membrane surface hydrophilic to improve the diffusivity of water at the surface rather than in the interior, it was possible to suppress the I–V hysteresis, even for the cell using the Fe–N-Cp catalyst, which has high water absorption [44]. Therefore, increasing the utilization of back-diffusing water from the anode is essential in the supply of water to reaction sites of the cathode [32,38]. These results indicate that the water diffusivity, not only in the interior of both the membrane and CL, but also at the interface between membrane and cathode, is important. Fig. 8(b) shows diagrammatically the effect of changing the cathode catalyst from the Fe–N-Cp to the CIAC-developed counterpart. The Fe–N-Cc material exhibited lower water absorbability than the Pajarito material. The use of this material led to decreased absorption of the back-diffusing water into the interior of the carbon in the CL and an increased volume of water supplied to reaction sites on the surface of the carbon, which are those with the shortest oxygen diffusion distance. These improvements demonstrate that water transport is the main limitation responsible for the previously reported I–V hysteresis [44] and provide strategies to achieve higher performance AEMFCs through proper water management and formation of water transport pathways.In previous studies, we clarified that the supply of water to the catalytic active sites was responsible for the severe voltage drop observed as the current density was increased and thus had a great influence on the hysteresis of the I–V performance of the cathode in AEMFCs. In the present work, we examined two approaches for the improvement of the water management ability, with the aim of suppressing the I–V hysteresis phenomenon.First, to increase the supply of generated water to the reaction active sites of the cathode, we decreased the membrane thickness and hydrophilized the membrane surface in order to increase the flux of back-diffusing water. These improvements of the water transport at the interface between the membrane and the cathode by use of the hydrophilization on a thin electrolyte membrane were effective in eliminating the I–V hysteresis phenomenon at both the anode and the cathode. Based on various types of membrane characterization, including CS-AFM, contact angle and osmotic pressure, we also clarified that hydrophilization does not improve the diffusivity of water in the interior but contributes to the improvement of surface anion conduction and water transport pathways. This effect was confirmed in the power generation performance measurements, even with the Fe–N–C catalyst with high water uptake that was associated with the initial observation of the I–V hysteresis.Second, from the viewpoint of the catalyst layer, to suppress the I–V hysteresis, we sought to improve the performance in comparison with that obtained for the Fe–N-Cp catalyst used in the previous research by using the recently developed Fe–N-Cc catalyst. The nitrogen adsorption results indicated that water was not able to enter into the pores of the Fe–N-Cc CL despite the larger pore volume than that of the Fe–N-Cp CL at low pressures, close to the practical fuel cell condition. From water vapor adsorption and contact angle evaluation, the high hydrophobicity of the Fe–N-Cc catalyst was assigned to the hydrophobicity of Black Pearls, the furnace black carbon, which is similar in its characteristics to Kejten Black, which is the carbon support material of Pt/CB. From results, the reason for the I–V hysteresis suppression by the Fe–N-Cc CL was able to be assigned to its high hydrophobicity, so that the volume of water absorbed by the catalyst was decreased, and the water required for the reaction was secured. In addition, the reason for the low ohmic resistance of the Fe–N-Cc-based cell was considered to be that the volume of water absorbed by the catalyst was suppressed, and thus the water content of the membrane increased.These improvements have demonstrated that water transport is the main limitation responsible for the voltage drop observed as the current density is increased, with the resulting I–V hysteresis, and thus have provided strategies for achieving higher performance AEMFCs through proper water management and formation of water transport pathways. Kanji Otsuji: Conceptualization, Formal analysis, Investigation, Writing – original draft. Yuto Shirase: Investigation. Takayuki Asakawa: Investigation. Naoki Yokota: Resources. Katsuya Nagase: Resources. Weilin Xu: Resources. Ping Song: Resources. Shuanjin Wang: Resources. Donald A. Tryk: Writing – review & editing. Katsuyoshi Kakinuma: Conceptualization, Validation. Junji Inukai: Conceptualization, Validation. Kenji Miyatake: Conceptualization, Validation, Resources, Funding acquisition. Makoto Uchida: Conceptualization, Methodology, Validation, Resources, Writing – review & editing, Supervision, Project administration, Funding acquisition.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This project was partly supported by the New Energy and Industrial Technology Development Organization (NEDO) Japan through funds for the “Advanced Research Program for Energy and Environmental Technologies,” by the Japan Society for the Promotion of Science (JSPS) and the Swiss National Science Foundation (SNSF) under the Joint Research Projects (JRPs) program, and by the Japan Science and Technology (JST) through Strategic International Collaborative Research Program (SICORP).We are deeply grateful to Pajarito Powder supplying the Fe–N-Cp catalyst.The following is the Supplementary data to this article: Multimedia component 1 Multimedia component 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.jpowsour.2022.230997.

Anion exchange membrane fuel cells (AEMFCs) are vulnerable to water management problems, since water is produced at the anode and consumed at the cathode. Previously we found severe voltage losses when increasing the current density in an AEMFC with a commercial Fe–N–C cathode catalyst. In the present work, we have clearly identified the problem as being related to water management and developed two approaches to alleviating the problem: by use of a thin hydrophilized membrane, the diffusivity of water at the surface was improved, and the severe I–V hysteresis was suppressed, despite the cell using an Fe–N–C cathode catalyst with a high water absorption rate. The voltage loss was also alleviated by the use of a recently developed Fe–N–C catalyst with higher hydrophobicity, which decreased the absorption of back-diffusing water into the catalyst layer and increased the amount of water supplied to the reaction sites These improvements have demonstrated that water transport is the main limitation for the previously reported hysteresis and provide strategies to achieve higher performance AEMFCs through proper water management and formation of water transport pathways.

Volatile organic compounds (VOCs) are defined by the WHO as a group of organic compounds whose melting point is lower than room temperature and boiling point is between 50 and 260 °C, including alkanes, olefins, aromatics, and so on. It has been proved that VOCs are primary pollutants of PM2.5 and photochemical smog [1], and most of VOCs are harmful to human health, such as benzene, toluene, so more and more research on VOCs removal are been carried out. At present, the main methods of VOCs removal include adsorption, catalytic combustion, thermal combustion, biodegradation, and so on [2]. The catalytic combustion is proved the most efficient technology for the removal of VOCs because of their lower-temperature operation, high degradation activity and less secondary pollution [3], and the core of catalytic combustion is to design and prepare stable and high efficient catalysts [4]. Many scholars pay attention on research of perovskite with ABO3 when the B site was occupied by Mn element because it could apply abundant lattice defects and higher oxygen mobility [5–7] compared with other transition metal elements [8,9]. When the B element is fixed, changing the A site element of the perovskite can increase the surface oxygen species concentration, specific surface area and low temperature reduction performance of the catalyst, thereby changing the catalytic activity of the catalyst [10–12]. Using of three-dimensionally ordered macroporous (3DOM) catalyst in the catalytic combustion of VOCs not only helps to increase the effective active specific surface area of the catalyst, but also facilitates the diffusion of reactants and products in the catalyst pores and enriches the macroporous structure [13,14]. Designing perovskite into 3DOM structure has very important basic research significance and practical significance for catalytic combustion of VOCs [15–18]. Three-dimensional (3D) ordered mesoporous manganese dioxide with high specific surface area and cubic symmetry prepared using the hard template method can completely oxidize formaldehyde at 130 °C [19]. Jiang et al. [20] supported Mn3O4-Au nanoparticles on 3DOM La0.6Sr0.4CoO3, which made the catalyst possess higher specific surface area, adsorbed oxygen concentration and good low temperature reducibility.The work was focused on discussing the effect of A-site element on catalytic performance of manganese-based perovskite catalyst. All the prepared samples were characterized by the Brunauer–Emmett–Teller method (BET), scanning electron microscopy (SEM), transmission electron microscopy (TEM), selected area electron diffraction (SAED), X-ray diffraction (XRD), H2-temperature programmed reduction (H2-TPR), X-ray photoelectron spectroscopy (XPS) and the activity of the materials for eliminating toluene was investigated by catalytic combustion technology to obtain (3DOM AMnO3 catalysts with the best performance.All chemicals used were A.R. grade without further purification. Methyl methacrylate (MMA) was obtained from Macklin. Cerium (III) nitrate hexahydrate, Nickel Nitrate, Lanthanum nitrate, Manganous nitrate, Polyethylene glycol-400, methanol, Potassium persulfate, P-Hydroxybenzoic acid, were purchased from Sinopharm Chemical Reagent Co.,Ltd.The polymethyl methacrylate (PMMA) colloidal crystal microspheres were synthesized using emulsifier-free emulsion polymerization approach [17,21,22]. A three-necked round-bottomed alaskite reactor (1000 mL) filled with 650 mL of deionized water equipped with a magnetic stirrer was heated by a hot water bath. In order to remove the air in the reactor, a pipet for pure N2 introduction was also connected to the vessel. Under constant stirring (350 rpm) and with N2 bubbling, the water was kept at 70 °C for 30 min and then 57 mL of methyl methacrylate monomer inhibited with ca. 0.03% p-hydroxyl benzoic acid was poured into the reactor through the third opening which was otherwise closed with a stopper. After further stirring and N2 bubbling at 70 °C for 15 min, a solution of potassium persulfate initiator (0.20 g dissolved in 20 mL of deionized water) preheated to 70 °C was added. With N2 bubbling and stirring, the reaction was allowed to run at 70 °C for 45 min, and then the emulsion was cooled to room temperature and mixed with 1300 mL deionized water. The PMMA colloidal crystal microspheres were left suspended in the liquid medium.PMMA- template was prepared by constant temperature suspension film forming method [23]. The emulsion was centrifuged at 7000 rpm for 40 min. After mixing the solid layer and deionized water into a homogeneous emulsion, it was dried at 80 °C by a hot water bath and the PMMA hard-templating with surface gloss and orderly accumulation was obtained.The 3DOM AMnO3 support was prepared using the PMMA-templating strategy [24–26]. In a typical method. Nitrate salts Ce(NO3)3 6H2O, La(NO3)3 6H2O, Ni(NO3)2: Mn(NO3)2 according to the molar ratio of 1:1, were dissolved in 20.9 mL of methanol (MeOH) and 3.0 mL of polyethylene glycol-400 adjust the total metal concentration of the precursor solution to 2 mol/L. At room temperature (RT) under stirring for 2 h to obtain a transparent solution. 2.0 g of the PMMA template was soaked in the above pre-cursor solution for 4 h. After the mixture was filtered, the obtained wet PMMA template was dried in air at RT for 48 h, The thermal treatment process was divided into two steps: (i) the dried PMMA was first claimed in a RT flow at a ramp of 1 °C/min from RT to 300 °C and kept at this temperature for 4 h, and (ii) the sample obtained after step (i) was claimed in an air flow at a ramp of 1 °C/min from RT to 600 °C and maintained this temperature for 5 h, thus obtaining the 3DOM AMnO3 support.The morphology and size of samples were observed using S-4800 field emission scanning electron microscope (SEM, Hitachi Co; Japan) and JEM-2100UHR transmission electron microscope (TEM, JEOL; Japan). The Brunauer-Emmett-Teller (BET) surface area, pore volumes and pore diameters were performed with N2 adsorption/desorption isotherms on Micromeritics ASAP2020M analyzer. The crystal structure of samples was determined using X’Pert PRO MPD using Cu Kα radiation (λ = 0.1541 nm) from 10° to 80°. X-ray photoelectron spectroscopy (XPS) was measured on a Thermo ESCALAB 250Xi XPS system from a monochromatic aluminum anode X-ray source with Kα radiation (1486.6 eV), and the spectra were calibrated with the C1s peak at 284.6 eV as an internal standard. Hydrogen temperature-programmed reduction (H2-TPR) experiments were carried out on a chemical adsorption analyzer (PAC-1200).The catalytic activities of the samples were evaluated in a continuous fixed bed reactor with a diameter of 28 mm and a length of 500 mm. To minimize the effect of hot spots, 5 mL of quartz sands (40–60 mesh) was used to dilute 1 mL of the sample (40–60 mesh). The reactant was com-posed of 300 ± 50 mg/m3 toluene. Reactants and products were analyzed online by a gas chromatograph (Agilent 7890B) equipped with a flameionization detector (FID) and a TCD. When the reaction temperature was below 350 °C, toluene showed less signs of conversion, indicating that the self-decomposition of toluene at high temperature could be excluded in this experiment.The XRD results of the catalysts were shown in Fig. 1 , It was found that all three catalysts (3DOM LaMnO3, 3DOM CeMnO3, 3DOM NiMnO3) reveal the peak value at 24.8°, 33.1°, 36.7°, 47.6°, 50.8°, 55.2°, 59.2° and 69.5°. Previous studies have shown that the peaks correspond to the characteristic peaks of the perovskite crystal phase, which indicates that the study has successfully synthesized a catalyst with a perovskite crystal phase. Small differences exist in these diffraction patterns, especially for the sample of 3DOM LaMnO3 with unclear peak of perovskite, meanwhile, the diffraction peaks of the 3DOM CeMnO3 are sharp and symmetrical with increasing half width, and moving to the higher angle. In addition to the characteristic diffraction peaks of perovskite. The sample of 3DOM CeMnO3 with additional peaks of the CeO2 (2θ = 28.6°, 56.2°) and MnO2 (2θ = 65.8°). Phases indicates that formation of CeMnO3 may lead to small amount of residual CeO2 and MnO2. Because Ce4+ is the constant valence state of Ce and higher than the conventional perovskite A-site element [27], the perovskite structure of the CeMnO3 catalyst is unstable. Some Ce and Mn elements are retained, resulting in the formation of CeO2 and MnOx. The presence of oxides makes the catalyst have more lattice oxygen and surface oxygen vacancies [28], which is conducive to the catalytic combustion of toluene.The presence of macropores in 3DOM AMnO3 catalysts is further assessed by SEM in Fig. 2 . 3DOM CeMnO3 shows high quality three-dimensional ordered macropore (Fig. 2a, b). The samples exhibit the macroporous materials contain a skeleton surrounding uniform periodic arrange voidswith an average diameter of 200 ± 50 nm and a wall thickness of 16–18 nm [17]. However, the pore structure of 3DOM LaMnO3 (Fig. 2c, d) has low porosity, poor permeability and irregular macroporous structure. 3DOM architecture agrees well with TEM analysis (Fig. 3 ). It can be seen that the macroporous structure formed by 3DOM CeMnO3 was the most complete and the SEM of NiMnO3 showed that the catalyst didn’t form macroporous structure. The result also agree well with BET. In Fig. 3, (a) and (c) represent 3DOM CeMnO3, and (b) and (d) represent 3DOM LaMnO3. From the HRTEM characterization diagram of 3DOM CeMnO3, the (2 0 0) and (1 1 2) crystal planes of orthorhombic perovskite and (1 1 1), (2 0 0) crystal plane and (2 1 1) crystal plane of Mn3O4. Moreover, the arrangement of each crystal plane is intricate and shows a small amount of lattice distortion. It can be inferred that the 3DOM CeMnO3 catalyst has more surface oxygen defects, which is consistent with the results of XRD and XPS characterization. Figs. 4 and 5 shows the N2 adsorption-desorption isotherms and pore size distribution of catalysts prepared. Except 3DOM NiMnO3 catalyst, all adsorption-desorption isotherms display a type II isotherm with type H3 hysteresis loop (P/P0 = 0.7–1.00), indicating the existences of mesopores or macropores. In the catalysts [29], which is consistent with the SEM images in Fig. 2. At the same time, it can be seen that the adsorption amount of the 3DOM CeMnO3 catalyst was faster than the other sample (Fig. 1A), which corresponds to the specific surface area shown in Table 1 (3DOM CeMnO3 is 48.8 m2/g, 3DOM LaMnO3 is 35.6 m2/g and 3DOM NiMnO3 is 11.05 m2/g.).When the specific surface area is much larger, the adsorption and transfer rate of toluene on the surface of the 3DOM catalyst is greatly increased and more active sites are exposed to increase the catalytic reaction rate. Fig. 5 shows the pore size distribution for all catalysts where clearly defined peaks around 0–5 nm are observed and confirms the existence of mesoporous as is mentioned above. It is worthy to mention [30] the size of the specific surface area is related to the integrity of the pore structure. The results show that the A-site element of perovskite has a great influence on the apparent structure.The oxidation state of manganese and the properties of oxygen species in catalysts can be analyzed by XPS. The composition of surface elements of catalysts can be discussed. The corresponding results were shown in Fig. 6 and Table 2 . Among all the catalysts investigated, the Mn4+/Mn3+, and Oads/Olatt ratios over 3DOM CeMnO3 were the highest showing a strong mutual effect among manganese, oxygen, and Cerium. As shown in Fig. 6 (a), the Mn 2p3/2 signal peaks of the catalysts can be decomposed into two components: binding energies at 641 eV and 642.7 eV can be respectively classified as surface Mn3+ species and satellite peaks of surface Mn4+ species, indicating that Mn3+ and Mn4+ coexist in all samples. The O1s signal peaks of the catalysts can be decomposed into two components. The binding energies of the peaks are 529.2 eV and 531.4 eV, respectively. They belong to the species of surface lattice oxygen (Olatt) and the species of surface adsorbed oxygen Oads (O, O2– or O2 2–) [31]. NiMnO3 catalyst has a peak of binding energy at 532.8 ev, which belongs to the carbonate species [32–34]. This peak indicates that PMMA template is prone to carbonization in the process of calcination of the catalyst, so it is not conducive to the formation of macroporous structure, which is also consistent with the porous structure shown in the SEM diagram of 3DOM NiMnO3 catalyst (Table 3 ).The oxygen species can be strongly adsorbed in oxygen defects on surfaces of perovskite catalysts [34] and high concentrations of adsorbed oxygen might promoted the catalytic activity [21]. Because of the stronger Ce–O and Mn–O interactions, the 3DOM CeMnO3 catalyst exhibited the highest Oads/Olatt ratio. The large amount of active absorbed oxygen improved the catalytic performance for the oxidation reactions. Interestingly, the peak of lattice oxygen in the 3DOM CeMnO3 catalyst shifted to a higher BE by 0.5 eV. This denoted the enhancement in the O2– mobility of the catalyst, which facilitated the combustion of toluene.During catalytic combustion reaction, the oxygen adsorbed on the catalysts will be converted to lattice oxygen, while the surface Mn4+ will be reduced to Mn3+. At the same time, some lattice oxygen on the surface will be consumed. Therefore, the catalyst has high concentration of Mn4+ and high surface adsorbed oxygen can improve the performance of catalytic oxidation of VOCs. The formation of perovskite by cerium-manganese catalyst is unstable, which results in more non-stoichiometric perovskite structure in the catalyst. This will make perovskite crystal have more oxygen defects [35], which is more conducive to the adsorption of oxygen species on the surface, and can greatly promote the catalytic combustion of toluene. As shown in Table 2, the adsorption oxygen and Mn4+ concentration of 3DOM CeMnO3 catalyst are the highest and the catalytic performance of this catalyst is the best, which is consistent with the results of H2-TPR.H2-TPR was used to characterize the redox ability of the catalyst, the characterization results were shown in Fig. 7 . The peaks can be attributed to the reduction of Mn species in B-site of perovskite-type catalysts [25]. The low-temperature reduction peaks of 3DOM CeMnO3, 3DOM 3DOM NiMnO3 and 3DOM LaMnO3 catalysts occur at 425 °C, 440 °C and 475 °C respectively, which are mainly attributed to the reduction of Mn4+ species (Mn4+ → Mn3+) transformation in B-site of perovskite-type catalysts. The high-temperature reduction peaks appeared at 500 °C, 530 °C and 570 °C respectively, which can be attributed to the single electron reduction of Mn3+ which is in coordination unsaturation Translate into Mn2+. For manganese-based catalysts, Mn4+ is the main substance to promote the catalytic reaction [36,37]. After changing the A-site element to Cerium, the reduction peak becomes wider and the reduction temperature moves to a lower temperature. Obviously, the intensity of reduction peak of Mn3+ in 3DOM CeMnO3 catalyst is higher than that of the other two catalysts, and there is almost no peak [27] in 3DOM NiMnO3. This means that 3DOM CeMnO3 has better redox performance at low temperature and is more conducive to the removal of toluene [38]. Fig. 8 showed the effect of the A site element on the catalytic combustion of toluene. Although the removal rate of toluene hardly changed at lower temperatures, as the reaction temperature increased, obvious differences gradually appeared. When the conversion of toluene is 50%, the order of removal efficiency of toluene is 3DOM CeMnO3 > 3DOM LaMnO3 > 3DOM NiMnO3. When the conversion of toluene is 90%, the activity of catalyst is 3DOM CeMnO3 > 3DOM NiMnO3 > 3DOM LaMnO3. The results show that although all the samples have highly catalytic activity, the catalytic activity of different elements still varies greatly. According to Table 1, Positive correlation between catalytic activity and specific surface area of the sample. The 3DOM CeMnO3 exhibited the highest catalytic activity, the ignition temperature (T50%) and complete conversion temperature (T90%) were 100 °C and 172 °C, respectively. For the 3DOM NiMnO3 catalyst, T50% is lower than the 3DOM LaMnO3 catalyst. In contrast, T90% is higher than the 3DOM LaMnO3 catalyst. This indicates that 3DOM NiMnO3 has better catalytic activity at high temperature, which can be consistent with the results of H2-TPR and XPS.Studies have shown that the catalytic combustion process of toluene belongs to the first order kinetics. In the presence of excess oxygen, there is a relationship between the concentration of VOCs (c) and other parameters, as shown in formula (1-1), where the parameters r, k, A, and Ea represent the reaction rate (mol·s−1), Reaction rate constant (s−1), antecedent factor (s−1) and apparent activation energy (kJ·mol−1). (1-1) r = - k c = - A e x p - E a / R T c Fig. 9 show Arrhenius curves of samples at different firing temperatures (obtained before the toluene conversion is below about 20%). As can be seen from the figure, there is a good linear relationship between Lnk and 1 / T. R2 is 0.9693 (3DOM CeMnO3), 0.936 (3DOM LaMnO3), and 0.891 (3DOM NiMnO3). The 3DOM CeMnO3 catalyst with the best pore structure and the largest specific surface area (38.8 m2·g−1) also showed the lowest apparent activation energy (34.51 kJ·mol−1, SV = 15,000 h−1).Three kinds of Mn-based perovskite structure catalysts, 3DOM CeMnO3, 3DOM LaMnO3 and 3DOM NiMnO3, prepared by the PMMA Hard-Templating- Excessive impregnation method were employed as the catalyst for simultaneous toluene removal from air. The research results demonstrated that 3DOM CeMnO3 has the most complete macroporous structure, regular channels, strong permeability and high porosity. Its specific surface area reaches 48.8 m2/g−1, which is much larger than that of 3DOM NiMnO3 (11.1 m2/g−1), thus providing more adsorption sites for toluene molecule. 3DOM CeMnO3 had the best toluene removal performance among the catalysts. The superior catalytic performance of 3DOM CeMnO3 resulted mainly from the abundance of oxygen vacancies and the strong interaction between CeO2 and MnOx formed during calcination. The 3DOM CeMnO3 sample showed lower apparent activation energy (34.51 kJ·mol−1, SV = 15,000 h−1) and the best catalytic activity for toluene combustion, with the reaction temperatures (T50%, and T90%) required for achieving toluene conversions of 50%, and 90% being 100 °C, 172 °C. Toluene conversion of 3DOM CeMnO3 could be at least 90% in the air atmosphere at 160-180 °C, which was the optimal temperature range for the toluene removal, exhibiting prominent low-temperature activity. Therefore, 3DOM CeMnO3 perovskite oxide has broad prospects and can be used as an active catalyst in a natural carrier such as cordierite to form a monolith catalyst for industrial production.The author is grateful for the financial support of the Natural Science Foundation of Shandong Province [ZR2019MEE112].

Three-dimensionally ordered macroporous manganese-based perovskite catalyst (3DOM AMnO3, A = Ce, La, Ni) were synthesized by PMMA hard-templating and impregnation method. Physicochemical properties of the samples were characterized by means of various techniques including XRD, BET, SEM, TEM, XPS and H2-TPR, and their catalytic activities were evaluated by toluene combustion. It was found that the 3DOM AMnO3 in each of the samples was perovskite in crystal structure, and only the samples possessed a good quality 3DOM architecture with a surface area of 48.8 m2/g. Due to the highest adsorbed oxygen species concentration (Oads/Olatt = 2.330), the best low-temperature reducibility (The low-temperature reduction peaks of 3DOM CeMnO3 catalysts occur at 425 °C) and the strong interaction between CeO2 and MnOx formed during calcination. The 3DOM CeMnO3 sample showed lower apparent activation energy (34.51 kJ·mol−1, SV = 15,000 h−1) and the best catalytic activity for toluene combustion, with the reaction temperatures (T50%, and T90%) required for achieving toluene conversions of 50%, and 90% being 100 °C, 172 °C at SV = 15,000 h−1, respectively.

In catalysis science field, transition-metal heterogeneous catalysts can be considered one of the most important and far-reaching scientific developments up to now, since they promote the development of energy storage & conversion, chemicals manufacture, as well as prevention & treatment of environment pollution. The development of chemical industry field requires identification of improved transition metal catalysts with improved efficiency, selectivity and durability for each reaction processes, and made from earth-abundant elements. However, the trial-and-error approaches are still a common practice to search for new catalysts, largely due to the lack of deep insight into the fundamental relationship between structure feature of reactions site and its catalytic performance [1,2]. In heterogeneous catalysis, the atomic arrangement around the active sites is of significant influence on the processes of reaction cycle. Therefore, an important challenge in achieving a rational design of optimal transition metal heterogeneous catalysts, is to unveil the structural stability of the active centers and to explore how their configuration transform upon exposure to the realistic environments, attracting increasing research effort [3].Since two decades ago, first-principles-based or density functional theory (DFT) calculation tools have been used to tackle this challenge and take advantage of a predictive and analytical ability without the help of experiment [4]. They base on quantum chemical theory to reach a quantitative description of the catalytic process in present of ultra-high vacuum and 0 K. However, aiming to address issues of catalysis, studies with consideration of the impact of reactant partial pressures and temperature, cannot be solely finished by first-principles calculations. In order to involve the impact of realistic environments, DFT calculations are accompanied with concepts of chemical potential, which is an important concept in thermodynamics, since all of the thermodynamic properties of a material at a given temperature and pressure can be obtained from its chemical potential. Chemical potential-based model can construct a frame to assess intrinsic thermodynamic tendency and criteria for certain chemical process and map out the influence of external environment. In the meantime, the energetics for specific system or object in the thermodynamic model would be calculated precisely and conveniently by DFT calculations. Notably, first-principles calculation tools and thermodynamic theory have been coupling, and have been firmly adopted in the conceptual toolbox of theoretical surface catalysis [5]. The first-principles-aided thermodynamic models provide access to determine the stability of catalyst and the surface structure reconstruction under given thermodynamic condition, which is regarded as a prominent hallmark.Here, we briefly review recent theoretical works on transition metal heterogeneous catalysts by means of first-principles computation-aided thermodynamic models. According to the common concerns on transition metal heterogeneous catalysts, first, we will outline the whole thermodynamic framework to obtain the chemical potential of specific catalysts and analyze their thermodynamic stability. Then, we will demonstrate representative examples showing how to identify structure evolution of catalyst under operando reaction conditions.To be more feasible to interact with reactants, the size of transition metal catalysts usually disperse in the form of micrometer, nanometers or even atomic level scale. In the solid–gas thermocatalysis, the metal species tend to disintegration and/or sinter due to high temperature and adsorbed reactant, either through migration and agglomerate of metal atoms and nanoparticles, or through Ostwald ripening [6–8]. In the end, the lack of active surface area leads to the decrease and deactivation of overall activity of the metal species. As for solid–liquid electrocatalysis, especially in acid media, the catalytic activity fading of metal nanoparticles is caused by various processes, such as dissolution, agglomeration, and detachment [9,10]. These factors are tangled and often happen simultaneously. For example, the dissolved metal species from small particles may be redeposited on large nanoparticles. These mechanisms result in the loss of electrochemical surface and decayed performance. A thermodynamic understanding or evaluation of the structural destabilization tendency at atomic level would be helpful to propose an efficient method for suppressing deactivation processes and increasing the durability of transition metal catalysts. Extensive studies have employed first-principles computation-aided thermodynamic models, and provided a thermodynamically quantitative description of structural stability of transition metal with consideration of the effect of temperatures, pressures and the presence of reactants.An insight into Ostwald ripening and disintegration of metal nanoparticles under reaction conditions is regarded as the central issue on the thermal stability of metal nanoparticles heterogeneous catalysts for durable practical implementation. To reach a precise analysis on thermal stability of metal nanoparticles, Li et al. proposed a first-principles computation-aided thermodynamic model to quantitatively describe Ostwald ripening and disintegration processes with consideration of the reactant, particle size and morphology [11,12]. This theory model contains chemical potential of supported metal nanoparticles and formation energy of monomers on supports, as well as corresponding sintering thermodynamic tendency, in the presence of given temperature and reactant partial pressure. As display in Fig. 1 a, a supported nanoparticle is modeled by a spherical segment, the chemical potential of which as a function of curvature radius R is derived by Gibbs–Thomson (G–T) formula [13]. (1) μ N P = 2 Ω γ N P / R where Ω is the molar volume of bulk atom and γ NP is the weighted surface energy of the nanoparticles.Taking into account that nanoparticle expose various facets i with area ratio f i and corresponding surface energy γ i , the weighted surface energy of nanoparticles γ N P could be, (2) γ N P = ∑ i f i γ i Ostwald ripening of nanoparticles proceed through the decomposition of small particles to form monomers (mono-atom or multi-atom complexes), as well as the diffusion and the attachment to larger particles of monomers on the support. Therefore, as shown in Fig. 1b, in addition to determining the thermodynamic priority of detachment/attachment of the monomers from/toward the nanoparticles, the formation energy of the metal monomers (n atoms, ΔEf MO) has an impact on the concentration of monomers, which is expressed by the relative chemical potential of monomers to nanoparticle. (3) Δ E M O f ( R ) = μ M O − μ N P (4) μ M O = E M O @ s u p p o r t − E s u p p o r t − n E b u l k n where μ MO is the chemical potential of the metal monomers with relative to bulk, EMO@support is the total energy of supported monomers, Esupport is the energy of the substrate and Ebulk is the energy per bulk atom.The concentration of the metal monomers with respect to nanoparticles of curvature radius R is defined by, (5) c M O ( R ) = exp ( − Δ E M O f ( R ) k B T ) a 0 2 = c M O e q exp ( − μ N P k B T ) where a 0 is the lateral lattice constant of support, and c M O e q = exp ( − μ M O ( R ) k B T ) a 0 2 is the concentration of the monomers in equilibrium relative to the bulk. Hensen et al. carried out first-principles-aided thermodynamic models to study the dependence of ripening mechanism of CeO2(111)-supported Pd nanoparticles on size [16]. Particle coalescence is possible only for nanoparticles with less than 5 Pd atoms, while Ostwald ripening is the dominant sintering mechanism for larger cluster. Since metal-support interaction plays an important role in stabilizing catalysts, first-principles-aided thermodynamic modeling has been applied to reveal influence of facets and crystal phases of substrate on structural stability of supported metal nanoparticles. For example, Li et al. studied thermodynamic tendency of Pt nanoparticle ripening on various pristine TiO2 surfaces of both anatase and rutile phases [17].Under reaction conditions, reactants adsorbing on supported nanoparticles would decrease the surface tension and stabilize the nanoparticles, which affects subsequent Ostwald ripening and disintegration (Fig. 1b). The correct of surface tension on the facet i at given environment temperature and reactant partial pressure is expressed by, (6) Δ γ i ( T , P ) = θ i [ E r e a c t a n t a d ( θ i ) − μ r e a c t a n t ( T , P ) ] / A i where θ is the coverage of adsorbate, Ai is unit area of the surface i, and μ r e a c t a n t ( T , P ) is chemical potential of free reactant molecules. The coverage dependence E r e a c t a n t a d ( θ i ) (average adsorption energy of reactant) can be evaluated by first-principles theory calculation and fitted with polynomial functions. The coverage θi at specific T and P would be derived from the differential adsorption energy of reactants [15] (Fig. 1c–d). (7) E r e a c t a n t d i f ( θ i ) = d [ θ i × E r e a c t a n t a d ( θ i ) ] d θ i = μ C O ( T , P ) By substituting the revised surface energy of supported nanoparticles with adsorbates, chemical potential of nanoparticles turn to be a function of T and P, which has impacts on the Ostwald ripening and disintegration processes. (8) μ N P ( r e a c t a n t ) = 2 Ω γ N P ( r e a c t a n t ) R = 2 Ω ∑ i f i [ γ i + Δ γ i ( T , P ) ] R Fig. 1e–f show contour plot of μNP(CO) correlating with curvature radius, temperature, and pressure [14]. In general, the smaller radius, higher temperature, and lower pressure will bring about a higher μNP(CO), while the thermal stability of Au nanoparticles is only sensitive to particle size when it grows larger.Apart from reducing the surface energy and stabilize the nanoparticles, reactants would stabilize metal monomers detached from nanoparticles by forming metal-reactant complexes presented in Fig. 1b, which assist Ostwald ripening behavior. Likewise, first-principles-aided thermodynamic model can be applied to put insight into the effect of the adsorption of reactants on Ostwald ripening. The Gibbs free energy of adsorption Δ G r e a c t a n t would be corrected. (9) Δ G r e a c t a n t = E a d s ( r e a c t a n t ) − μ r e a c t a n t ( T , P ) To satisfy exothermic adsorption of reactants on metal monomers ( Δ G r e a c t a n t < 0), μ r e a c t a n t ( T , P ) is required to meet the following criteria. (10) μ r e a c t a n t ( T , P ) > E a d s ( r e a c t a n t ) The chemical potential of monomers with reactant adsorbed, μMO (reactant), is expressed by, (11) μ M O ( r e a c t a n t ) = μ M O + Δ G r e a c t a n t where Eads (reactant) is reactant adsorption energy on metal monomers, μ reactant (T, P) is the chemical potential of free reactant molecule as a function of pressure and temperature. The chemical potential of monomers upon reactant adsorption is lowered by Δ G r e a c t a n t . The concentration of monomers upon reactant adsorption is, (12) c M O ( r e a c t a n t ) ( R ) = c M O e q exp ( − μ N P ( r e a c t a n t ) + Δ G r e a c t a n t k B T ) a 0 2 Apart from accelerating Ostwald ripening, reactants could decompose supported nanoparticles into the metal-reactant complexes distributing on substrate. Reactant-assisted disintegration of metal particles is usually used to regenerate the sintered catalysts. Corresponding first principle-involved thermodynamic study has also been developed. In order to assess whether reactant-induced decomposition of nanoparticle into the metal-reactant complexes is a thermodynamically spontaneous process, the thermodynamic tendency is defined by, (13) Δ G N P d i s ( R , T , P ) = E c o m f − μ r e a c t a n t ( T , P ) − μ N P ( r e a c t a n t ) − T S where S is the configurational entropy of metal-reactant complexes, and E c o m f = μ M O + E a d s ( r e a c t a n t ) is the formation energy of the metal-reactant monomers on substrate relative to the bulk and reactants in gas phase. Since the value of Δ G N P d i s is controlled by the formation energy of the complexes, the chemical potential of nanoparticle and free reactant molecule, and the configuration entropy due to the decomposition, these factors have the following influence and implication. The composition of reactant gas should be varied according to diverse supported nanoparticle systems, which guarantee reactant can interacts strongly with metal monomers for desired disintegration, since E c o m f is dominated by the interaction among reactant, metal and support. As for specific catalysts and supports, the calcination in corresponding oxidizing conditions had been adopted to improve the dispersal of metal catalysts in experiment. Besides, Δ G N P d i s is determined by the reaction environment and the radius of nanoparticles. Therefore, chemical potential of reactants should be reached for thermodynamically spontaneous disintegration ( Δ G N P d i s ) is, (14) μ r e a c t a n t d i s ≥ E c o m f − μ N P ( r e a c t a n t ) − T S As for a specific T and P, the nanoparticles of the radius less than R(dis) will be disintegrated. (15) R ( d i s ) ≤ 2 Ω γ N P ( r e a c t a n t ) E c o m f − μ r e a c t a n t ( T , P ) − T S The DFT-based thermodynamic theory developed above was employed to study CO surrounded TiO2(110) supported Rh nanoparticles, of which sintering and disintegration behavior coincide with observation in experiments, and demonstrates how the metal-carbonyl monomers affect Ostwald ripening and disintegration of supported nanoparticles [11]. The dependence of Δ G N P d i s for both Rh(CO) and Rh(CO)2 as metal-reactant monomers on T at experimental P = 10−1 mbar are shown in Fig. 2 a. Rh(CO) and Rh(CO)2 will become the main complexes during 750–770 and 370–750 K, respectively. When T ≤ 370 K, the Δ G N P d i s become negative, Rh nanoparticles disintegrated to individual Rh(CO)2 become thermodynamically spontaneous process. The impact of pressure on agglomeration and disintegration at T = 300 K is plotted in Fig. 2b. Rh(CO) becomes the dominant monomers at most experimental conditions [10−25, 10−24] mbar, where Rh(CO)2 becomes thermodynamically favorable monomers in the range of [10−24, 10−4] mbar. When P > 10−4 mbar, Rh nanoparticles of d = 20 Å tend to decompose into Rh(CO)2 fragments in thermodynamics. At 300 K and 10−1 mbar, the corresponding Δ G N P d i s of nanoparticles into Rh(CO)2 monomers versus diameter is shown in Fig. 2c, which indicates that Rh nanoparticles decompose into Rh(CO)2 spontaneously when the diameter is smaller than 60 Å. Similar theoretical works investigated the detachment of Cu monomers from CeO2(111)-supported Cu nanoparticles onto the substrate with and without CO adsorption, which show that the adsorption of CO reduces detachment energy and promote the formation of metal monomers species on ceria [18]. First-principles-aided thermodynamic models have also been employed to understand the stability of TiO2-supported Rh, Pd, and Pt nanoparticles in NO or CO atmosphere [19], by studying the thermodynamic tendency for disintegration of nanoparticles into metal-reactant complexes. NO is found to be a more efficient reactant for nanoparticles disintegration and redispersion than CO. And Rh nanoparticles are found to be most sensitive to either NO- or CO-induced disintegration. The study of reactant-involved disintegration and redispersion process of FeO/Pt (111) supported Au particles in CO atmosphere is another example [20]. It is found that CO stabilizes the ripening Au monomer by forming Au carbonyls according to phase diagram at a wide range of temperatures and CO pressures. CO decrease the onset temperature of ripening by a few hundred kelvins.As a frontier in heterogeneous catalyst field, the single-atom catalysts (SACs) has brought about much interest, due to the efficient utilization of expensive metals, as well as excellent selectivity and high activity for specific reactions compared to supported nanoparticles [21]. The precondition of advantages of SACs is the thermal stability of isolated metal atom against aggregation under catalytic reaction conditions [22]. Therefore, the binding strength between isolated metal atom and substrate is significant for sinter-resistant SACs. In thermodynamic aspect, the support-induced lower chemical potential of single atoms compared to nanoparticles lead to the thermal stability of SAC against sintering, since single atoms can't disperse and form nanoparticles spontaneously. Similar to the thermodynamic framework of Ostwald ripening of nanoparticle, a quantitative description on thermal stability of SACs has been achieved by combining first-principle calculation and chemical potential-based thermodynamic model, considering the environment condition, substrate, and metal-reactant interaction [14]. According to first-principle-aided thermodynamic model, the chemical potential of supported isolated atom is approximately expressed by the difference between formation energy of supported single atoms and the bulk (μbulk = 0): (16) μSA  = (ESA@support  − Esupoort  − Efree atom ) – (Ebulk  − Efree atom ) = ESA@support  − Esupoort  − Ebulk where ESA@support is the total energy of supported single atoms, Esupport is the energy of the support, Efree atom is the energy of free-standing atom and Ebulk is the energy per bulk atom. The energy change for accreting a metal single atoms into metal nanoparticles is determined by the chemical potentials of nanoparticles relative to single atoms: (17) ΔEf SA  = μNP  − μSA Here, ΔEf SA (R) reflects the thermal stability of supported metal single atoms against aggregation. A larger value (especially positive value) generally implies that metal single atoms thermodynamically tend to resist aggregation. The concentration of metal single atoms with respect to nanoparticle of radius R is defined as, (18) c S A = exp ( − Δ E S A f k B T ) a 0 2 where a 0 is the lateral lattice constant of support. Based on the expressions above, the thermal stability of metal single atoms is associated with temperature, nanoparticle size, and embedding location of single atoms.Li et al. applied the first principle-aided thermodynamic model to design thermally stable Au SACs on series of oxide support under CO oxidation reaction [14]. It is shown in Fig. 3 a, that the thermal stability of Au SACs can be adjusted by diverse factors including temperature, pressure, nanoparticle size, and the reducibility of the substrate, according to the chemical potentials of Au single atoms with respect to Au nanoparticles. Gao et al. reported the effect of introducing nitrogen atoms into carbon-based support on the thermodynamic stabilities of single-atom iron catalysts by first principle calculations. The combination of chemical potential calculation and electronic structure analysis indicates that the doped N promotes charge transfer and, accordingly, improves thermodynamic stabilities [23]. Likewise, Li et al. found that a defect site of carbon support makes the chemical potential of Au single atoms much lowered, which unveils the great impact of defects on binding and stabilizing the Au single atoms [24]. Senftle and his group recently revealed that the degree of metal single atoms anchored to supports is governed by the chemical potential of binding process [25]. Metal atoms with strong exothermic adsorption on a support possess higher thermal stability and are hindered to diffuse or agglomerate. Based on first principle-aided thermodynamic model, Wei et al. proposed that the atomization and agglomeration of metal species depend on binding strength between metal-support (chemical potential of single atoms) and metal–metal (chemical potential of nanoparticles) [26]. So they explored the origin of thermally stable single atoms supported on N-doped carbon derived from noble metals nanoparticles.Similar to nanoparticles, the chemical potential of single atoms is influenced by reaction conditions. The chemical potential of single atoms with reactant adsorption, μSA (reactant), is expressed by, (19) μ S A ( r e a c t a n t ) = μ S A + E a d s ( r e a c t a n t ) − μ r e a c t a n t ( T , P ) where Eads (reactant) is reactant adsorption energy on single atoms; μreactant (T, P) is the chemical potential of free reactant molecule as a function of pressure and temperature. The reaction Gibbs free energy of metal single atoms aggregation upon reactant adsorption, ΔGagg , can be estimated by, (20) Δ G a g g = μ N P ( r e a c t a n t ) − μ S A ( r e a c t a n t ) ΔGagg is also an index that represents the thermal stability of supported metal single atoms against aggregation. A larger value (especially positive value) suggests that single atom is a more thermodynamic preferable form for metal species than nanoparticle upon reactant adsorption. Fig. 3b shows the tendency of μSA(CO) and μNP(CO) (R) (R = 20 Å) with varying temperature at given CO partial pressure, for Au single atoms supported on various oxide [14]. It is shown that μSA(CO) on MgO(100) is always 1 eV higher than μNP(CO) (R) at all studied temperature range, which suggests that the Au single atoms highly tend to cluster. The weak CO binding to Au atom at the ceria oxygen vacancy resulted in the similar between μSA(CO) and μSA. On the contrary, with the presence of CO adsorption, Au single atoms on CeO2(111) and ceria steps becomes more stable and resist sintering below 200–300 K, due to a positive ΔGagg. As shown in Fig. 4 , Hensen et al. applied ΔGagg to determine the threshold of CO partial pressure and temperature, where isolated Pt atoms are thermally stable against sintering on CeO2(111), and the threshold size of Pt nanoparticles with thermodynamic stability relative to isolated Pt single atoms [15].Comprehension of the electrochemical stability, especially dissolution of transition metal catalysts is crucial in solid–liquid heterogeneous electrocatalysis. Previous theoretical works mainly focus on transition metal catalysts used in acid oxygen reduction reactions (ORR) of proton exchange membrane fuel cells (PEMFCs). One of the most critical issues on electrochemical stability is the acid dissolution of metal species on fuel cells working potential. The first-principles-aided thermodynamic approach has been employed to understanding the electrochemical stability and predict the dissolution potential of transition metal electrocatalyst [27,28]. The formula of metal bulk dissolution is expressed by, M↔M 2+ + 2e − The dissolution process is investigated for an n-atom nanoparticle by applying the same method, M n ↔M n-1 + M 2+ + 2e − Thus, the reaction free energies at a given potential U is defined by (21) ΔG = G(M 2+,aq) − 2eU − G(M,s) (22) ΔG = G(M n-1) + G(M 2+,aq) − 2eU − G(Mn ) The chemical potential of cations G (M2+,aq) is derived from both experimental standard redox potentials U0 and calculated chemical potential of bulk metals G (M,s) (23) G(M 2+,aq) = G(M,s) + 2eU 0 Based on Nernst equation, U = U 0 – RT/zF × log ([M 2+]), reduction potentials U can be modified according to different concentration of M2+(aq) in the electrolyte. It requires ΔG < 0 for the dissolution proceeding spontaneously. Therefore, the equilibrium potential for the dissolution of catalyst, named by the dissolution potential, can be estimated in the following, (24) U bulk  = 1/2e [G(M 2+,aq) − G(M,s)] (25) U n  = 1/2e [G(M n-1) + G(M 2+,aq) − G(M n )] = U bulk  + 1/2e [G(M n-1) + G(M,s) − G(M n )] As seen in Fig. 5 a, Jinnouchi et al. reported the site-dependent dissolution potentials of the Pt particles by first-principles-aided thermodynamic approach, which found that Pt atoms at edges with higher d-band centers dissolve more facilely than those at flat surfaces [27]. Pt atoms attaching to the carbons have lower redox potential. Seo et al. demonstrated the obvious size-dependent dissolution process for Pt nanoparticles with diameters smaller than 3 nm (Fig. 5b) [29]. The dissolution process begins in edges and vertices and makes more (111) facets exposed. These results are in line with the consensus that the catalytic performance of Pt in nanoscale is restricted under the PEMFC environment due to their inferior electrochemical stability. In Fig. 5c, Ceder et al. explored the origin of lower dissolution potential of Pt nanoparticles relative to Pt bulk, since Pt nanoparticle dissolve through a different mechanism from that of bulk Pt, where dissolution of Pt nanoparticle takes place through electro-oxidation of Pt oxide to Pt2+ cations [28]. Li et al. applied first-principles-aided thermodynamic models to explore the factors influencing the dissolution of the Pt nanoparticles [30]. The oxygen chemisorption can hinder the dissolution by decreasing the surface energy and increasing the standard redox potential of dissolution/deposition. In addition, the dissolution is accelerated obviously at higher electrode potential, while it is impeded by large particle size.The dissolution of Pt-based alloy was likewise studied based on first-principles-aided thermodynamic approach. A DFT study on Pt–Co nanoparticle showed that the Pt shell-Co core structure possess a stronger anti-dissolution ability than pure Pt nanoparticles [31]. Another theoretical work on the Pt–Au nanoparticle concluded that the Pt dissolutions is hindered by substitution of edge Pt by Au atoms [32]. Balbuena et al. showed that the impact of Pt on the anti-dissolution ability of the second metal component in the alloy has the same order as that in pure metal, with PtIr and Ir possessing the highest dissolution potential [33]. Remarkably, based on a thermodynamic analysis on electrochemical stability by predicting the threshold potential of dissolution for a series of supported transition metal single atoms or clusters, Li et al. predicted a method for synthesizing high-purity and high-loading single atoms or clusters, shown in Fig. 6 [34]. They demonstrated the applicability of an electrochemical potential window, by which any metal species beyond electrochemical potential window are corroded away from substrate, while the high stability against dissolution of single atoms remains on the support.More and more significant findings emerging from experimental studies in situ and operando conditions indicate a dynamic nature of the catalyst surface under the catalytic reaction conditions. The catalyst structure may constantly vary before, during and after the catalytic process attributed to the interaction with reactant species. For example, ultra-thin oxide layers would form and cover the transition metal particles in an oxidizing environment [35]. The surface metal component of alloy would change upon exposure to reaction environments due to surface segregation, resulting in the enrichment of one metal on the shell and the other in the core [36]. The presence of reactants may adjust the ratio of exposed crystal facet so that the dynamic morphology reshaping of the metal particle could occur under reaction conditions [37]. All structural evolution of catalyst mentioned above would provide different active sites from those in absence of reaction condition, which could not be described precisely by 0 K/Ultra High Vacuum (UHV) surface science experiments or theoretical simulation. An insight into the relationship between configuration reconstruction of the transition metal catalyst and the reaction environment is crucial for a deep comprehension of catalysis and for a rational design of catalysts. Nowadays, theoretical methods are moving from 0 K/UHV models to operando investigations, including first-principles computation-aided thermodynamic framework, which precisely predicts the reconstructed configuration of transition metal catalysts with thermodynamic tendency upon the reaction environment.In a given reaction environment, the reactant gas may oxide or reduce the catalyst surface, and reactant molecules may combine with catalyst surface and lead to surface termination phase transition, such as thin oxide-like structures on exposure to high oxygen pressures [38]. The influence of the reactant partial pressures and temperatures on the reactant-involved surface phase transition have been taken into account by employing first-principles-aided thermodynamic model proposed by Reuter [5,39,40]. The first-principles-based thermodynamic framework is briefly demonstrated below. The Gibbs ensemble is suitable to identify the most stable system geometry and composition at given reaction condition by the minimum surface free energy of unit area A: (26) γ ( T , P ) = 1 A [ G ( T , P i , N i , N j ) − N i μ i ( T , P ) , N j μ j ( T , P ) ] where G ( T , P i , N i , N j ) is the Gibbs free energy of a specific surface models containing N i species i and N j species j, μ i and μ j are the chemical potentials of individual reservoirs of each components. It represents the energy needed to create certain surface termination for taking all atoms out of corresponding reservoirs. After calculating γ ( T , P ) of possible reactant-involved surface termination configuration, the one which possesses the lowest γ ( T , P ) is the thermodynamically most preferable configuration at given reaction conditions. Therefore, the excess energy relative to suitable reference is used to evaluate various surface configurations: (27) γ ( T , P ) − γ 0 ( T , P ) = 1 A [ G ( T , P i , N i , N j ) − G 0 ( T , P i , N i ' , N j ' ) − Δ N i μ i ( T , P ) − Δ N j μ j ( T , P ) ] The Gibbs free energies is approximated by total DFT energies accompanied with vibrational contribution, which makes it simplified to determine thermodynamically stable surface phase, by only evaluating DFT energy difference. (28) γ ( T , P ) − γ 0 ( T , P ) ≈ 1 A [ Δ E t o t − Δ N i Δ μ i ( T , P ) ] where E s e g − n − m the vibrational contribution is contained in the free energy part Δ N i Δ μ i ( T , P ) . (29) Δ E t o t = E t o t ( N i , N j ) − E 0 t o t ( N i ' , N j ' ) − Δ N i E t o t ( i ) − Δ N j E t o t ( j ) The first-principles-aided thermodynamic model has been applied to confirm the thermodynamically preferable atom arrangement of the surfaces termination under different reaction environments and it is routinely used to describe the surface phase transition of complex multicomponent systems. Since transition metals act as common catalysts for many thermal catalytic oxidation reaction, previous studies mainly focus on the surface phase transition of close-packed transition metal slabs under oxygen atmosphere. The growth of ultra-thin oxide covering transition metal, which bears a little resemblance to the bulk oxides, have been displayed, including Ag (111) [41,42], Pd (111) [43], Pd (100) [44], Rh (100) [45], Rh (110) [46], Pt (110) [47], Rh (111) [48], Ni(110) [49], Cu(111) [50], Cu(100) [51] and Au (111) [52]. These researches explored the surface atomic arrangement that tend to exist thermodynamically, and how they are determined by the reactant pressure and temperature. Li et al. employed DFT calculations on possible oxygen-induced reconstructed configurations at Ag (111) [53,54]. Through ab initio-based thermodynamic model, the surface free energies of Ag surface structures with various O coverage as a function of oxygen chemical potential were plotted (Fig. 7 ), by which some thermodynamic steady oxygen-involved reconstructed Ag surface, such as the p (4 × 4) phases, have been identified by theoretically and experimentally confirmed in scanning tunneling microscopy STM studies. Analogous theoretical models were applied on electrocatalytic oxygen reduction reaction of Pt nanoparticle, which indicated that hydroxide radical formation at (100)-edges at 0.5 V (RHE) would be substituted by atomic oxygen at 0.75 V, and the atomic oxygen aggregate to form PtO2 chains sinking into the subsurface with increasing electrode potential till 1.18 V [31].Apart from pure transition metal, first-principles-aided thermodynamic model has also achieved a deep understand on surface phase transition of bimetallic alloy under reaction condition [55]. Scheffler et al. plotted the surface phase diagram of the (111) crystal facets of Ag–Cu alloy as a function of oxygen chemical potential depending on temperature and pressure, as well Cu surface content (Fig. 8 a) [56,57]. According to the surface phase diagram, the most thermodynamic stable at given temperature and pressure as a function of surface element composition can be determined. For example, copper impurities in silver host prefer to stay in subsurface in the absence of oxygen, while in the oxygen atmosphere typically used in ethylene epoxidation, surface phase suggests that, with different the copper surface concentration, clean Ag (111), thin copper oxide layers, and thick oxide-like structures can coexist. In a multicomponent atmosphere, the presence of different reactant gas plays an individual role in surface termination reconstruction of catalyst, so that chemical potential of involved reactants need to be accounted for in order to assess the precise surface configurations. A typical case is a study by Scheffler et al. to identify preferable surface structures of Ag–Cu alloy determined by chemical potential of both ethylene and oxygen during ethylene epoxidation reaction, which indicates a dynamical coexistence of CuO layer and AgO–CuO shell (Fig. 8b) [58]. Another example is the surface phase diagram of the Pd (100) slab under a reactive condition of CO oxidation. The stable surface configurations in CO oxidation condition locates at the boundary between phases reactant binding to surface oxide and reactant attaching on the metallic slab (Fig. 8c) [59].In addition to oxidizing environment, first-principles-aided thermodynamic model have been successfully applied to describe surface phase transition under the other atmosphere. Scheffler et al. showed the surface phase diagram of rutile RuO2(110) under a humid environment containing oxygen and water vapor [60]. Catherine Stampfl et al. demonstrated the effect of N chemisorption on the reconstruction of Cu (111) (100) and (110) surfaces in N2 atmosphere [61,62]. A peculiar atom arrangement for 0.75 ML coverage of N atoms is predicted, namely a metastable ‘‘N-trimer cluster’’ on the metal slab. A surface nitride-like configuration is found to be energetically favored in surface phase diagram, and exists within a narrow scope of nitrogen chemical potential until the growth of bulk Cu3N. The atomic morphology change of NiCr alloy surface induced by fluorine chemisorption is investigated by first-principles-aided thermodynamic models to explore the early-stage corrosion processes of nickel-based alloys in strong oxidizing environment [63].A synergistic effect of transition metal alloys is strongly related with the surface composition, which may promote or hinder desirable and undesirable chemical reactions. The surface element content would change upon exposure to certain reaction environments, where one of the metal components enrich the surface region and the other stay in core, known as surface segregation [64]. Thermodynamically, surface segregation is derived from disparity of surface energy among diverse metals, so that the reaction environment would cause the transition of surface composition by tuning the surface energy. First-principles-aided thermodynamic model have been developed well to determine thermodynamic steady surface composition of alloy by defining segregation energy (Eseg), which is defined as the energy needed to overcome for moving solute metal from the interior to the surface layer of the host metal and calculated by the following equation: (30) E s e g − n − m = E n t h − E m t h (31) E s e g − n − b u l k = E p u r e b u l k + E n t h − E i m p u r i t y i n b u l k − E p u r e s u r f a c e where E s e g − n − m is the total energy difference of alloy surface model with solute metal atoms in the nth layer and mth layer. E s e g − n − b u l k , is the total energy difference between solute metal atoms in the nth layer and bulk. Negative Eseg indicates solute atoms in bulk prefer diffusing to the surface, while a positive Eseg implies solute atoms tend to stay in the bulk interior. Depending on the segregation energy, element-dependent segregation behavior of transition metal alloys for extended surface model and cluster model were transformed to a color-coded matrix [65,66].The thermodynamic steady surface composition would vary due to the environmental atmosphere. O2 and H2 tend to drive different metal elements to migrate to the surface when alloy catalysts are exposed to an O2 and H2 atmosphere, respectively, which is very meaningful for the applications of Pt-based alloy systems on PEMFC electrocatalysis. DFT calculation results have predicted a reversible surface segregation behavior of Pt–Ni nanoparticles upon alternating H2 and O2 environments [67–69]. Ni surface segregation is not obvious under a reducing atmosphere (such as H2) resulting from the weak bonding between H and Ni, while Ni prefer segregating onto the surface in the form of NiO under O2 atmosphere. Byungchan Han et al. found that a high surface coverage of oxygen accounts for the Co segregation of thermodynamically stable Co core-Pt shell nanoparticles under the working condition in fuel cells (Fig. 9 a) [32]. Balbuena and his co-worker investigated surface component transition of Pt-based alloy catalysts upon acidic oxygen reduction reaction conditions, and discovered surface segregation adjust metal dissolution through tuning oxidation state of the subsurface atoms [70]. They also explored surface segregation of Pt3M (M = Fe, Co, and Ni) alloys in oxidizing condition, which suggested that both the Pt-segregated and M-segregated slabs only become stable than the original one when O coverage reaches over 1/4 monolayer [71]. Scheffler et al. combined DFT calculation and thermodynamic method to describe Ag3Pd(111) surface under an oxygen atmosphere, analyzed profiles of segregation, adsorption, and surface free energies in increasingly oxygen-rich environments [72].Apart from O2 and H2 atmosphere, transition metal tends to have strong interaction with other reactant gas, such as CO, which causes surface segregation. For example, as for Cu–Pt bimetallic catalysts under CO atmosphere, the strong affinity with CO accounts for Pt pulled out to the top surface and formation of a Cu–Pt near-surface alloy thermodynamic steady state (Fig. 9b) [73]. Donna A. Chen et al. combined DFT calculations and thermodynamic model for Ni1Au121 clusters with the Ni atoms in the center and confirmed that the most stable structure under CO atmosphere belongs to CO binding to a surface segregated Ni atom. CO-adsorption-induced Pd surface segregation, dynamics of PdAu swapping, and Pd clustering in AuPd bimetallic surfaces are demonstrated by Graeme Henkelman's group [74].Recent in situ experimental characterizations have observed the reversible reshaping behavior of transition metal nanoparticles in the presence of reactant environment [75]. The shape of nanoparticles can affect the number of active sites and, further, the catalytic reactivity. The crystal facet stability will change under reactant atmosphere, where more open facets are exposed at the expense of close-packed surfaces, increasing the number of coordination-unsaturated sites. A proper theoretical modeling is an effective way to reveal the physical insight behind the geometry reconstruction and to help achieve shape control for higher catalysis performance. Recently, a first-principles-aided thermodynamic model was proposed by Gao and his group, containing the Wulff construction theory, adsorption isotherms theory, and DFT calculations [37]. This model achieved great success in quantitatively describing the precise equilibrium shape of nanoparticles at different temperatures and gas pressures, such as the reshaping of Cu nanoparticles in the water vapor condition (Fig. 10 a) [76] and the shape evolution of Ru, Pt, Pd, Cu and Au nanoparticles under CO and NO environment (Fig. 10b) [77,78].Based on Wulff construction, the optimal geometry of the crystal with the lowest total surface free energy is built according to the surface free energy of every crystal surface. Under a given reactant condition, the surface free energy of a clean surface γ h k l should be revised by considering the interface tension from reactant adsorption: (32) γ h k l i n t = γ h k l + θ ( T , P ) E a d s A a t where E ads and A at are average adsorption energy and surface area per atom, respectively. (T, P) is the reactant molecules coverage on surface determined by the temperature and reactant partial pressure, expressed by the Langmuir adsorption isotherms. (33) θ 1 − θ = P K where K is the adsorption equilibrium constant, expressed by: (34) K = exp ( − Δ G R T ) = exp ( − E a d s − T ( S a d s − S g a s ) R T ) where S gas is the entropy of the gas in the gas phase and S ads is the adsorption entropy. (35) γ h k l i n t = γ h k l + θ ( T , P ) ( E a d s − z w θ ) A a t (36) θ 1 − θ = P K e x p ( − z w R T θ ) where z is the number of first-neighboring molecules of adsorbates at 1 ML adsorption coverage, w is interaction energy between two nearest adsorbed molecules. After obtaining corrected surface tension of each crystal facet, the Wulff theorem is applied to construct the equilibrium geometry of nanoparticles at a given T and P.In addition to mono-component reactant gas, the thermodynamic model is feasible for the environments including multi-component atmosphere [79]. First-principles-aided thermodynamic model has made great success in reproducing in situ experimental characterization on nanoparticle exposed to H2 and O2 conditions (representative reductive and oxidative gases). The simulated shape evolution of Pd nanoparticles up to ambient conditions reached perfect agreement between the in situ environmental transmission electron Microscopy (ETEM) observation at the same T and P [80]. According to first-principles calculations, the surface tension of the Pd (110) and (100) surface is stabilized under O2 atmosphere, while the surface free energy of the (111) facet changes to the most unstable surface, which leads to surface area proportion transition of low-coordination sites up to 82%. Likewise, the prediction of thermodynamic model shows that the facet proportion of Pd (100), (110) and (111) at the equilibrium of state for Pd nanoparticles in H2 atmosphere, are 22%, 48% and 30%, respectively, which coincide with the experimental observation. As for the shape reconstruction of the Au nanoparticles caused by O2 atmosphere, first-principles-aided thermodynamic model described the equilibrium geometry evolution of Au nanoparticles upon different temperature [81]. The reshaping of Au nanoparticles with truncated octahedron began with the corner when decreasing the temperature, and eventually became a round shape, which is in line with the observation of in situ TEM. On the contrary, the morphology evolution of the Au nanoparticles in H2 atmosphere was not obvious. In addition to strong reductive and oxidative gases, N2, known as an inert gas, was also predicted to affect the equilibrium geometry of nanoparticles at ambient conditions, matching the experimental results. The coverage of N2 on the Pd (110) surface is much higher than that (111) and (100) surfaces, which makes the surface free energy of the (110) slab decreased and induces an increase of facet proportion [82]. First-principles-aided thermodynamic model were also reported to describe the reshaping of multi-metallic nanoparticles. The H2 environment-induced morphological evolution of PdCu alloy nanoparticles transforming from spheres to truncated cubes was visualized by theoretically calculating the surface free energy of PdCu alloy surface with or without hydrogen adsorption (Fig. 10c) [83]. In addition to transition metal, the equilibrium morphology of transition metal oxide under realistic reaction conditions has been successful predicted by first-principles-aided thermodynamic models [84].Heterogeneous catalysts are generally supported on a high-surface-area substrate. Supports can alter morphology of nanoparticle, and furthermore the perimeter of the interface between the nanoparticle and the substrate providing active centers. It is helpful to theoretically predict the reshaping of interface between metal particle and substrate under reaction conditions. The contact-surface tension between the nanoparticles and the substrate for supported nanoparticles system, can be evaluated by the Wulff-Kaischew theorem. (37) γ c − s E = γ A − E a d h − ( θ B E a d s B A a t B ) where γ A is the surface tension of the facet of nanoparticles adhering support, E adh is the binding energy between nanoparticles and support,θ B is the coverage of reactant on the substrate, E a d s B is the adsorption energy of the reactant on the substrate, and A a t B is the surface area of the substrate.Pt@SrTiO3, as a model system of supported nanoparticles, was studied by first-principles-aided thermodynamic model under a H2 environment [85]. Theoretical prediction show that Pt nanoparticle display a clear wetting on a SrTiO3(110) surface in absence of H2, consistent with TEM images. The H2 adsorption-involved structural transformation was modelled in Fig. 10d. Since H2 molecules compete with the Pt nanoparticle for attachment on the SrTiO3 substrate, which results in the dewetting of the supported Pt nanoparticle. As a result, the shape of the contact surface and the number of atoms at interface perimeter are both adjusted by the gas environment.First-principles-aided thermodynamic model has taken a dramatic development and established a computational strategy in evaluating structural stability and exploring structural reconstruction of transition metal catalysts under reaction condition, which was briefly summarized. First-principles computation-aided thermodynamic framework provides thermodynamic understanding or evaluation of the structural destabilization tendency, which is helpful to propose measures for suppressing deactivation processes and increasing the durability of transition metal catalysts. In addition, it belongs to computational operando investigations, precisely predicts the reconstructed configuration of transition metal catalysts with thermodynamic tendency in the reaction environment.A new computational approach, which can manage big database of accumulated by DFT calculations, is suggested to essentially develop in the future, by which energetics parameters in thermodynamic model can be correlated with readily available physical properties of the reactant, metal and the support, and identify key parameters known as descriptor. The descriptor influencing reactant-metal, metal–metal and metal–support interactions is promising to be used in conjunction with Machine learning to develop a predictive model for screening high-stability catalysts and understanding their structure evolution in future rational catalyst design.We declare that we have no financial and personal relationships with other people or organizations that can inappropriately influence our work, there is no professional or other personal interest of any nature or kind in any product, service and/or company that could be construed as influencing the position presented in, or the review of, the manuscript entitled “First-principles-aided thermodynamic modeling of transition-metal heterogeneous catalysts: A review”.This work is supported by the National Natural Science Foundation of China (Grant Nos. 21822801), China Postdoctoral Science Foundation (2019TQ0021) and the Fundamental Research Funds for the Central Universities (XK1802-1 and XK180301).

Over the past decade, the first-principles-aided thermodynamic models have become standard theoretical tools in research on structural stability and evolution of transition-metal heterogeneous catalysts under reaction environment. Advances in first-principles-aided thermodynamic models mean it is now possible to enable the operando computational modeling, which provides a deep insight into mechanism behind structural stability and evolution, and paves the way for high-through screening for promising transition-metal heterogeneous catalysts. Here, we briefly review the framework and foundation of first-principles-aided thermodynamic models and highlight its contribution to stability analysis on catalysts and identification of reaction-induced structural evolution of catalyst under reaction environment. The present review is helpful for understanding the ongoing developments of first-principles-aided thermodynamic models, which can be employed to screen high-stability catalysts and predict their structural reconstruction in future rational catalyst design.

The selective oxidation of sulfides into the corresponding sulfoxides is one of the most significant procedures within both industrial and laboratory settings [1,2]. However, among a large number of traditional oxidants that have been widely used for this selective oxidation, many of them are dangerous to use or toxic [3–5]. Moreover, sulfoxides can often undergo over-oxidation to their corresponding sulfones during relatively harsh or harmful reaction conditions [6,3,7,8], and therefore, to overcome these limitations and considering the eco-sustainability, green chemistry and atom economy, a considerable amount of research has been focused toward the development of new and effective catalytic systems based on the use of aqueous H2O2 as a green oxidant [9–12]. The tetrazole unit continues to arise great attention in both industry and academia. Tetrazoles are of interest in pharmaceuticals, synthetic organic chemistry, coordination chemistry, catalysis technology, the photographic industry, and organometallic chemistry. Also, tetrazoles and their derivatives have been reported as analgesic, antiviral, anti‐inflammatory, anti‐proliferative, antibacterial, potential anti-HIV drug candidate, antifungal, herbicidal and anticancer agents. [13–15] Most notably, the 5-substituted 1H-tetrazole is frequently used as a carboxylic acid isostere in medicinal chemistry, [16,17]. Numerous active pharmaceutical ingredients containing a tetrazole moiety are currently on the pharmaceutical market. Later methods to the synthesis of tetrazoles in the literature generally involve the costly and poisonous metal, suffer from intense water reactivity, or utilization of hydrazoic acid, which is very poisonous, unstable, and flammable. [18–21] Also, there is a growing demand for safe, energetically efficient, and environmentally friendly procedures. Homogeneous catalysts have several advantages such as better performance, high turnover numbers, and high selectivity [22-24,10]. But significant weaknesses are obvious: separating homogeneous catalysts from the reaction medium is tedious and requires several expensive and particular methods [25–27]. Some difficulties frequently related to the homogeneous catalysts can be easily overcome by heterogenization of their counterpart on the surfaces of both organic and inorganic solids [28,29]. Magnetic nanoparticles have attractive physical and chemical properties . Superparamagnetism, high magnetic susceptibility, and low curie temperature are some unique magnetic properties of MNPs [30]. Fe3O4 nanoparticles are a good candidate as a support material for heterogeneous catalysts because of their great properties such as the abundance of unique activities, low toxicity and price, simple synthesis and functionalization, large surface area, and easy separation with magnetic field [31,32]. Fe3O4 has a cubic inverse spinel structure. Magnetic nanocatalysts are used to accelerate various organic reactions [33]. In this paper, we studied the selective oxidation of sulfides and the synthesis of 1H-tetrazole in green media with a green catalyst. The new synthesis proposes a suitable, and eco-friendly synthetic method. The novel Ag-containing creatinine functionalized Fe3O4 catalyst was designed by attaching Ag onto the surface Fe3O4. The catalysts have been used in the selective oxidation of sulfides and synthesis of 1H-tetrazole in green media with the classical and ultrasonic methods. At first, creatinine functionalized Fe3O4 has been prepared through the surface functionalization of Fe3O4 to offer Fe3O4@Creatinine sample, which on the treatment of AgNO3 results in the development of novel Fe3O4@Creatinine@Ag catalyst (Fig. 1 ).Today, many efforts are made to create conditions for the manufacturing of chemicals that meet most of the principles of green chemistry. One way to achieve this goal is to use catalysts that, while taking advantage of their benefits, do not impose any restrictions on the principles of green chemistry. The design and use of eco-friendly catalysts achieve the mutual goals of protecting the environment and promoting economic benefits simultaneously. The objective of this study was to produce a new eco-friendly nanocatalyst with high potential, high mechanical and thermal stability, high contact surface with the selective performance of healthy and eco-friendly materials and solvents, and cost-effective to selective oxidation of sulfides and prepare tetrazole derivatives in a green solvent. Most tetrazoles are obtained in toxic solvents at high temperatures with low efficiencies. Therefore, due to the significant industry demand for these products, materials based on magnetic nanoparticles can be chemically modified to design catalysts containing environmentally friendly active centers. In the synthesis of this catalyst, the environmentally friendly ligand is used to stabilize the eco-friendly active site on the desired support. What distinguishes this catalyst and this research is its high eco-friendliness in several respects. The structure of this catalyst and also, the reaction are based on green chemistry under mild reaction conditions. At a relatively low temperature compared to the reported work, tetrazole derivatives are prepared in water solvent with good efficiency. Other features of this designed catalyst include high activity, stability, easy recovery, and low cost.A mixture of FeCl3•6H2O (5.2 g) and FeCl2•4H2O (2.0 g) was introduced to deoxygenated water (25 mL) containing few drops of conc. HCl. Subsequently, 250 mL of NaOH (1.5 M) solution was added dropwise. The whole mixture was stirred vigorously at 60 °C. Then, brown-colored Fe3O4 NPs were isolated using a magnetic stick. It was rinsed using distilled water and dried at 40 °C.The Fe3O4 (1 g) was decorated by the reaction with 3-chloropropyltrimethoxysilane (CPTMS) (1.5 mL) under refluxing toluene (24 h). In the next step, the obtained Fe3O4 Cl was washed thoroughly with n-hexane and dried at 40 °C (Scheme 1, Supporting information).Creatinine incorporation into Fe3O4 was obtained by the following method: Triethylamine (3 mL) was added to the suspension of 1 g of the Fe3O4 Cl and 2 mmol of creatinine (0.22 g) in 30 mL of toluene. Under the stirring condition, the mixture was refluxed (48 h). The obtained product was separated and rinsed using deionized water. Then, the Fe3O4@Creatinine was dried at 40 °C (Scheme 2, Supporting information).Afterward, the obtained Fe3O4@Creatinine (1 g) was decorated by the reaction with AgNO3 (2.5 mmol) under refluxing ethanol (15 h). In the next step, the obtained Fe3O4@Creatinine@Ag was separated, and rinsed with ethanol, and dried at 40 °C (Scheme 3, Supporting information).Fe3O4@Creatinine@Ag was added (70 mg) to a mixture of nitrile (1 mmol), NaN3 (1.2 mmol), and H2O (3 ml) at 90 °C. At the end of the reaction (checked by TLC), the Fe3O4@Creatinine@Ag was separated using magnetic decantation and washed with EtOAc and distilled water. The organic layer was preserved with 10 mL of 5 N HCl. Then the organic was washed with water and dried over anhydrous Na2SO4 to produce the wanted products (Scheme 4, Supporting information).Fe3O4@Creatinine@Ag was added (60 mg) to a mixture of sulfoxide (1 mmol), H2O2 (0.6 mL) as oxidant, and EtOH (3 ml) at room temperature. After completion of the reaction, checked by TLC, the Fe3O4@Creatinine@Ag was collected using magnetic decantation and rinsed with ethanol and distilled water. Then, the product was extracted with diethyl ether and rinsed using distilled water, and dried over anhydrous Na2SO4. The product was purified by a plate to give the wanted product (Scheme 4, Supporting information).(Table 2, Entry 1) Ethyl phenyl sulfoxide 1HNMR (400 MHz, CDCl3, ppm): δ 3.01 (s, 3H), 3.11–3.17 (m, 2H), 7.29 (s, 2H), 7.57–7.70 (m, 1H), 7.92–7.94 (m, 2H).(Table 2, Entry 8) Dibenzyl Sulfoxide 1HNMR (400 MHz, CDCl3, ppm): δ 4.22 (s, 4H), 7.40–7. 45 (m, 10H).(Table 2, Entry 9) Benzyl phenyl sulfoxide 1HNMR (400 MHz, DMSO, ppm): δ 4.70 (s, 2H), 7.14–7.15 (m, 2H), 7.30–7.34 (m, 3H) 7.56–7.60 (m, 2H) 7.71–7.75 (m, 3H).(Table 5, Entry 1) 5-(3-Nitrophenyl)−1H-tetrazole. 1HNMR (400 MHz, DMSO, ppm): δ 7.92 (t, 2H), 8.40–8.50 (m, 2H), 8.86 (s, 1H.)(Table 5, Entry 2) 5-(4-Nitrophenyl)−1H-tetrazole 1HNMR (400 MHz, DMSO, ppm): δ 8.30–8.33 (d, 2H), 8.45–8.47 (d, 2H).FT-IR spectra of Fe3O4 MNPs (a), Fe3O4 Cl (b), Fe3O4@Creatinine (c), and Fe3O4@Creatinine@Ag were recorded (Fig. 2 ). The band at around 600 cm−1 is the typical band of the vibration of metal-oxygen (Fe-O) in the nanomagnetic compounds. The proper grafting 3-chloropropyltrimethoxysilane linker to pure Fe3O4 is recognized by the presence of the stretching vibrations of the CH2 group at 2958 cm−1 and in stretching vibration modes of Si-O at 1040 cm−1 in the FT-IR spectra of Fe3O4 Cl (Fig. 2b) [34]. In addition, the FTIR spectra for Fe3O4@Creatinine demonstrated new bands at 1680 and 1475 cm−1 which are related to stretching vibration modes of C = O and C = N bonds of creatinine respectively [35], which is an indication that creatinine has been attached successfully onto the Fe3O4. FT-IR spectrum d in Fig. 2 shows the final catalyst after adsorption of silver metal onto [email protected] curves of Fe3O4 and Fe3O4@Creatinine@Ag are shown in Fig. 3 . Three major mass loss processes can be detected for the Fe3O4@Creatinine@Ag sample. The mass loss between 150 and 450 °C (11.87%), 450–600 °C (4.18%), and 600–800 °C (4.22%) is assigned to the decomposition of organic molecules in the Fe3O4@Creatinine hybrid material [36].As seen in Fig. 4 , the XRD analysis patterns of the Fe3O4, Fe3O4@Creatinine@Ag were recorded by X-ray powder diffraction (XRD). XRD pattern of Fe3O4 NPs (Fig. 4a) has been confirmed by the bands at 2Ө (34.9°, 41.20°, 50.24°, 63.10°, 67.40° and 74.27°), corresponding to the (2 2 0), (3 1 1), (4 0 0), (4 2 2), (5 1 1), and (4 4 0) reflections. These spectra reflect the inverse spinel crystal structure of Fe3O4 NPS [37], indicating that a pure and highly crystalline product was obtained. Fig. 4b shows a representative XRD spectrum of the Fe3O4@Creatinine@Ag nanocomposite. The same set of characteristic peaks were observed for Fe3O4@Creatinine@Ag after surface modification, indicating that the functionalization of Fe3O4 did not significantly disturb the Fe3O4 phase. Besides the characteristic diffraction peaks of Fe3O4 reflections, the diffraction peaks at 37.562, 44.546, 64.665, and 76.584 can be indexed to (111), (200), (220), and (311) planes of silver with the face-centered cubic (fcc) structure (space group: Fm3 m), respectively (JCPDS, No. 04–0783) [38,39].Using a VSM (vibrating sample magnetometer), the magnetic properties of the samples Fe3O4 MNPs, and Fe3O4@Creatinine@Ag were investigated (Fig. 5 ). As illustrated in Fig. 5, the saturation magnetization of the Fe3O4 MNPs and Fe3O4@Creatinine@Ag is 62.82, and 32.6 emu/g, respectively. The reduction of saturation magnetization of Fe3O4@Creatinine@Ag nanoparticles is because of the coating of Fe3O4 with the non-magnetic layers (SiO2, Creatinine, Silver). All MNPs reveal superparamagnetic behavior [40].The particle shape and surface texture were evaluated by scanning electron microscopy (SEM). SEM images of Fe3O4, Fe3O4@Creatinine@Ag nanoparticles are shown in Fig 6 a, and 6b, respectively. The shapes of Fe3O4 and Fe3O4@Creatinine@Ag nanoparticles are spherical. There were no considerable changes in the morphology after immobilization of the silver complex on the magnetite surface.The chemical composition of the Fe3O4@Creatinine@Ag nano-catalysts was determined by the EDS analysis (Fig. 7 ). Using EDS spectroscopy, the presence of carbon, iron, nitrogen, oxygen silicon, and silver elements was confirmed in the structure of the catalysts. The exact content of silver that immobilized on the Fe3O4 nanoparticle was measured by inductively coupled plasma optical emission spectrometry method (ICP-OES). This amount was 0.17 mmol g − 1 for Fe3O4@[email protected] size and morphology of the samples were examined by TEM analysis (Fig. 8 ). The TEM and SEM images of as-synthesized Fe3O4@Creatinine@Ag demonstrate that the sample consists of almost spherical nanoparticles with sizes ranging from ca. 20 to 50 nm. Fig. 9 shows the UV–vis spectroscopic results for the Fe3O4@Creatinine@Ag composite spheres catalyst. Surface plasmon resonance was identified and the appearance of distinctive bands in the metal nanocomposite (Fe3O4@Creatinine@Ag) was noted, exhibiting a specific surface plasmon peak around 390 nm depicts the synthesis of silver nanoparticles [41,42].Khan et.al observed a specific surface plasmon peak for Ag/Fe2O3 at 410 nm. The average size of the nanocomposite was to be 25− 35 nm [41].The redshifts of SPR were attributed to the increase in NP size due to p-p interactions between the NPS and dye molecules, while bathochromic SPR shifts were attributed to the modest aggregation of NPs in hybrids, the very low surface curvature, and large particle size, which facilitated orbital overlap to afford a stiffer plasmon surface and induced the damping of electrons of NPs with high-energy resonance [43]. We estimate that the particles size of the Fe3O4@Creatinine@Ag obtained by uv-vis is according to the dimensions obtained from SEM and TEM.We studied the catalytic performance of the Fe3O4@Creatinine@Ag in the oxidation of sulfide at room temperature. In continuation of our studies, we compared ultrasonic and classical methods in the oxidation reaction. To optimize the experimental conditions, the oxidative of methyl phenyl sulfide using H2O2 at room temperature was chosen as a model reaction. The influence of the different experimental parameters like the amount of the catalyst and H2O2, and the nature of the solvent were investigated and the results are shown in Table 1 .The amount of the catalyst was evaluated on the direct synthesis of sulfoxide from sulfide in the range of 0 to 6 mg (Table 1). When the amount of the catalyst is raised to 6 mg, the sulfoxide yield is increased from trace to 95%, resulting probably from the presence of more catalytic active sites [44]. As expected, this factor has a positive effect on the sulfide conversion, since an increase in the catalyst dosage produces an increase in the conversion of sulfide. For further increase in catalyst dose up to 8 mg, conversion of sulfide was increased but was not a significant quantity. It was found that 60 mg of the catalyst is an appropriate amount of catalyst.To gain more understanding, the effect of oxidant (H2O2) was studied. Experiments were performed with different amounts of H2O2 in the range of 0.3 mL to 0.9 mL using model reaction. Table 1 depicts that the conversion of sulfide rose with an increment in the H2O2 amount. In subsequent studies, 0.6 mL of H2O2 was chosen as the appropriate quantity.We next explored the effect of solvents on the catalytic performance by the Fe3O4@Creatinine@Ag and H2O2. Different solvents e.g., ethyl acetate, ETOH, acetonitrile, and solvent-free condition was studied. The results in Table 1 show that the yields of methyl phenyl sulfoxide were markedly increased in ethanol. It should be also noted that it has been demonstrated previously that ethanol binds to the metal of the catalysis as ligands, and one of the effects of the alcohol solvents on the catalyst activity results from the coordination of alcohols as ligands [45].The protic solvent may act by decreasing the negative charge density on oxygen by hydrogen bonding, thus promoting nucleophilic attack by the second sulfide. In aprotic solvents, the intermolecular reaction is less efficient and the intermediate either decomposes, giving ground-state oxygen (favored at room temperature) or rearranges to sulfone (at low temperatures). The effects of Protic solvent suggest that intermediates are stabilized not only by hydrogen bonding with protic solvents but also by coordination with solvents [46].The effect of active metal (mmol/g of Ag on Fe3O4@Creatinine@Ag) on catalytic performance was investigated with different loading ratios of Ag (0.10, 0.17, 0.25, and 0.43 mmol/g) on Fe3O4@Creatinine@Ag performance. It was found that with an increase in active metal, the sulfide conversion was increased i.e., 88%, 95%, 96%, and 97% of sulfoxide yield was obtained. Sulfide oxidation efficiency up to 0.43 mmol/g shows a positive response due to an increase in the availability of active metal species.On having the optimized experimental conditions, we then recognized the scope of an overview of the Fe3O4@Creatinine@Ag to the selective oxidation of a range of sulfides to the related sulfoxide. As shown in Tables 2 and 3, a wide variety of sulfides including aryl, diaryl, alkyl, and dialkyl sulfides furnished the corresponding sulfoxide in excellent yields with classical (Table 2, Supporting information) and ultrasonic (Table 3, Supporting information) methods. The results show that ultrasonic is an appropriate method for the oxidation of sulfides to the related sulfoxide.Based on the previously reported mechanism for the oxidation of sulfides into sulfoxides using hydrogen peroxide in the presence of a catalyst, one explanation for this process is the complex formation between the hydrogen peroxide and the M + (intermediate A). After the formation of [M +−OOH], it may form oxo metal species as the active oxidant (intermediate B). This active species can oxidize organic sulfides by the formation of the oxidant-substrate complex (intermediate C) and the oxygen transfer to the organic substrate (Scheme 5 and 6 ) [47–50].The oxidation of different compounds with H2O2 on the Ag-containing catalyst shows the crucial role of decomposition of hydrogen peroxide. There is no doubt that the decomposition of hydrogen peroxide plays a crucial role in the oxidation of organic compounds.The catalytic performance of the new Fe3O4@Creatinine@Ag was also examined in the synthesis of 5-substituted 1H-tetrazoles. To optimize the experimental conditions, benzyl cyanide with NaN3 was chosen as a representative model. The effect of the different experimental parameters like temperature, solvent, and catalyst amount was studied and the results are shown in Table 4 .To optimize the best solvent needed for the synthesis of 5-substituted 1H-tetrazoles, we have carried out this reaction in the various solvent over Fe3O4@Creatinine@Ag as the catalyst. The results are listed in Table 4. All the reactions have been performed under refluxing conditions considering the different solvents. Ethanol is a polar protic solvent failed to produce the desired 1H-tetrazole in good yield (entries 1). When polar aprotic solvents like DMF and DMSO are used, 43% and 83% yield of products are obtained, respectively in long reaction time (entries 2, 3). It was found that water is the best solvent for reaction yield (entry 5). This effect can be attributed to the strong hydrogen bond interaction at the organic-water interface, which stabilizes the reaction intermediate [51]. Also due to the partial solubility of some azide derivatives in water, it is the best solvent for reaction yield.As expected, an increase in the reaction temperature promotes a faster transformation of nitrile. However, the productivity increased on raising the temperature. The increase of reaction temperature will accelerate the thermal movement of molecules and increase the probability of intermolecular collision; thus, the rate of the reaction is improved. Different research groups established that using high temperature is necessary for cycloaddition reactions [52], so it seems reasonable to think that increasing the reaction temperature would be a good choice for improvement in the yield. The best result for this reaction was obtained at 90 °C and conversion is practically total. (Table 4, entry 5).The effect of catalyst dose is a significant parameter in the synthesis of 5-substituted 1H-tetrazoles. The effect of catalyst dose on the synthesis of 5-substituted 1H-tetrazoles was studied with the dose range of 0–90 mg. As seen in Table 4, the synthesis of 5-substituted 1H-tetrazoles rate rose significantly as catalyst dose increased from 0 mg to 70 mg. For further increase in catalyst dose up to 90 mg, the yield was increased but was not a significant quantity. On increasing the catalyst dose from 0 mg to 70 mg, the synthesis of 5-substituted 1H-tetrazoles rate was enhanced from trace to 96% since higher doses provide more active sites, which in turn provides more chances for nitrile molecules to come into contact with a catalyst. Nevertheless, by further increment in catalyst amount from 70 mg to 90 mg, the synthesis of 5-substituted 1H-tetrazoles enhanced gradually since nitrile and active sites present on the catalyst reaches equilibrium. From this study, the authors conclude that a catalyst dose of 60 mg chosen was the best dose and was used for further work.The effect of active metal (mmol/g of Ag on Fe3O4@Creatinine@Ag) on catalytic performance was investigated with different loading ratios of Ag (0.10, 0.17, 0.25, and 0.43 mmol/g) on Fe3O4@Creatinine@Ag performance. It was found that with an increase in active metal, the nitrile conversion was increased i.e., 78%, 96%, 97%, and 99% of 1H-tetrazole yield was obtained.Based on our attained experimental results, the experimental conditions have been recognized as 90 °C in H2O with 70 mg catalyst. To recognize the scope and overview of Fe3O4@Creatinine@Ag catalyzed synthesis of 1H-tetrazole, different benzonitriles were used as substrates under our optimal conditions, as presented in Table 5 (Supporting information). The 1H-tetrazole was separated in short reaction times and high yields.In the study of azide-nitrile cycloaddition catalyzed by Ag+, the coordination of the nitrile substrate to the Lewis acidic silver is the source of the catalysis in the formation of 1H-tetrazoles. Ag+ coordinated to the nitrile, and this is the dominant factor influencing [2 + 3] cycloaddition. Subsequent neucleophilic attack by azide)intermediate A + B includes mechanism A [53] and B [54]) followed by hydrolysis produces tetrazole as the end product, with Ag+ catalyst being released for the next cycle of reactions (Scheme 7 ) [55].We were also interested in comparing the catalytic performance of our catalyst with that reported in the literature. At this point, selective oxidation of methyl phenyl sulfide and synthesis of 5-(4-chlorophenyl)−1H-tetrazole was chosen (Table 6 ). As can be seen, the catalytic performance of our procedure is superior to other catalysts.The recyclability of Fe3O4@Creatinine@Ag was studied for the synthesis of methyl phenyl sulfoxide and 5-(4-chlorophenyl)−1H-tetrazole under optimized reaction conditions. After completion of the reaction, the catalyst could be easily separated by using an external magnetic field, then it was rinsed with ethyl acetate and subsequently dried and reused directly for the next run. The results are presented in Fig. 10 . We could employ the catalyst for 6 runs for the synthesis of methyl phenyl sulfoxide and 5 runs for the synthesis of 5-(4-chlorophenyl)−1H-tetrazole without an appreciable decrease in catalytic performance. Fig 11 . shows the image of the Fe3O4@Creatinine@Ag nanocatalyst suspension without and with an external magnetic field.In summary, we have reported the synthesis of silver supported on the surface of magnetic Fe3O4 nanoparticles (Fe3O4@Creatinine@Ag) through a facile method. Both the structure and the chemical nature of the catalyst were confirmed using XRD, ICP-OES, FT-IR, SEM, EDX, VSM, and TGA methods. This sustainable magnetic nanocatalyst leads to the efficient oxidation of the sulfides and the synthesis of 5-substituted 1H-tetrazoles with good yields and selectivity under mild conditions. Furthermore, applying ultrasound irradiation in the oxidation of sulfides leads to increase yields, and decrease reaction times. The nanocatalyst could be effectively recovered under a magnetic field without a remarkable decrease in the catalytic performance.The authors report declaration of interest in this work.The authors are deeply grateful to the University of Kurdistan for the financial support and also Erfan Ghadermazi for their help on this research project.Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.apsadv.2021.100192. Image, application 1

The surface of the Fe3O4 has been functionalized with 3-chloropropyltrimethoxysilane, which undergoes an SN2 substitution reaction of the chloro group with the nitrogen of ligand (creatinine), offering the Fe3O4@Creatinine. A new recyclable Ag attached Fe3O4 MNPs (Fe3O4 @Creatinine@Ag) has been produced through a post-synthetic method. FT-IR (Fourier Transforms Infrared), TGA (Thermogravimetric Analysis), XRD (X-ray Diffraction), EDS (Energy-Dispersive X-ray Spectroscopy) and VSM (Vibrating Sample Magnetometer) confirm the effectiveness of the performed chemical modification to synthesize the catalyst. This catalyst displays high catalytic performance in the synthesis of 5-substituted 1H-tetrazoles in water and the selective oxidation of sulfides. In the presence of nano- Fe3O4 @Creatinine@Ag as an efficient heterogeneous nano-catalyst, the corresponding 5-substituted 1H-tetrazoles and sulfoxide were afforded under the mild condition in good to excellent yields. A wide variety of sulfides furnished the corresponding sulfoxide with classical and ultrasonic methods. The results show that ultrasonic is an appropriate method for the oxidation of sulfides to the related sulfoxide. The catalyst can be separated by simple recovery and reused for several periods without any remarkable decrease in the catalysis activity and selectivity. In the synthesis of this catalyst, the environmentally friendly ligand is used to stabilize the eco-friendly active site on the desired support. What distinguishes this catalyst and this research is its high eco-friendliness in several respects. The structure of this catalyst and also, the reaction are based on green chemistry. At a relatively low temperature compared to the reported work, tetrazole derivatives are prepared in water solvent with good efficiency.

Renewable biomass-based feedstocks are hydrogen deficient and often require the use of external hydrogen to generate green fuels/blends that are compatible with the current fossil fuels [1,2]. Aqueous-phase reforming (APR) is a promising catalytic route to generate hydrogen from dilute aqueous streams containing organic molecules [3,4]. Byproduct and waste streams from food industries or biorefineries [5] often contain dissolved organics usually in the range of 5–20 wt.%. One typical example is the aqueous phase of pyrolysis oil which contains a variety of oxygenates such as acids, aldehydes, alcohols, sugars to name a few [2,4].APR is a challenging process for a catalyst due to the drastic hydrothermal conditions (e.g., 225–275 °C and 35–90 bar) used and complex feedstocks utilized, requiring an active and particularly stable catalyst [3]. Typically APR is carried out over supported metal catalysts, e.g., Ni [6] and Pt [7] based catalysts. Critical issues for supports and active metals (e.g., textural properties and phase changes, leaching, and sintering) [8] were often reported for APR catalysts. Recent developments have shown that Pt/C is a promising candidate for the APR of a variety of organic components [9–13].Pt is, however, an expensive noble metal [14] and its loading on catalyst should be minimized for commercial application. This is generally achieved by altering Pt size (e.g. high dispersion [15]) and distribution (e.g., egg-shell [16]) employing different supports. Several effective and controllable means, such as varying Pt loading [17], applying various preparation, calcination, and reduction protocols [18–20], have been reported.Changing Pt size influences catalyst characteristics, which in turn, affects catalytic performance for APR. Lehnert et al. [21] studied APR of glycerol over Pt/Al2O3 catalyst and suggested that C-C cleavage in oxygenates (promoting the formation of C1 species which can be steam reformed to yield H2 [2]) occurs preferentially on face Pt atoms, which increased with Pt particle size. Kirilin et al. observed a similar trend in turnover frequency (TOF) for different carbon-supported Pt catalysts for APR of xylitol [22]. However, Wawrzetz et al. [17] and Barbelli et al. [18] observed only a slightly increased TOF for Pt/Al2O3 catalysts with Pt size increase from 1.1 to 2.6 nm for APR of glycerol, relating it to the enhanced and simultaneous hydro-deoxygenation reactions consuming hydrogen. Ciftci et al. [23] studied Pt size domain of 1.2–4 nm and obtained an optimized performance for Pt size of ca. 2 nm for Pt/C catalysts for APR of glycerol. Chen et al. [24] screened an even wider Pt size range of 1.6–5.7 nm for Pt/Al2O3 catalyst for APR of low boiling point fraction of bio-oil and reported an optimized Pt size of 2.6 nm for H2 production. These results are somehow contradictory. Nevertheless, it needs to be noted that in general, Pt size for the fresh catalysts was applied to correlate with catalyst performance.As compared with the widely-investigated Pt size effect [24], the influence of distribution (uniform and egg-shell) of Pt with varied sizes on APR performance has not been reported yet to the best of our knowledge. In order to comprehensively study the effects of Pt size and also Pt distribution, we have applied different preparation protocols (vide infra) to make a variety of Pt catalysts with distinguishable Pt characteristics. Ethylene glycol (EG) was used as a model reactant to evaluate catalyst performance since it is a simple oxygenate with both carbon atoms connected to OH-groups. This allows to estimate the catalyst preference for CC and CO cleavage [25]. Besides, a carbon material was used as the catalyst support in this study, considering the variety of support materials (e.g., Al2O3, SiO2, ZrO2, and TiO2) that been extensively studied for supported Pt catalysts for APR of EG (Table 1 ). Pt/C catalysts show appreciable turnover frequency for H2 production (TOF-H2) compared with the state-of-the-art catalysts, namely Pt/AlO(OH) and Pt/SiO2 (Table 1), both of which have stability problems. Therefore, the development of Pt/C catalyst is a valid argument and should aim to maximize TOF-H2. Since a Sibunit carbon-supported Pt catalyst showed better H2 productivity than other types of carbon materials supported Pt catalysts for APR of xylitol [22], it was used in this study to prepare the Pt/C catalysts. In total, four representative Pt/C catalysts with distinctive Pt characteristics such as small and agglomerated Pt particles, in a uniform fashion or with concentrated Pt particles on the surface in an egg-shell structure, are reported in this contribution. Moreover, the properties of the spent catalysts after 7-h APR of EG were correlated with the catalytic behavior.Sibunit carbon with a particle size of 100−200 μm was supplied by Boreskov Institute of Catalysis, Russia. H2PtCl6 was supplied by OAO Aurat, Russia. Pt-PVP colloid was prepared by a method published in Ref [33]. Analytical grade Na2CO3, formic acid, and ethylene glycol (EG, >99 %) were supplied by Sigma-Aldrich.Four Pt/C catalysts were prepared by variable methods, which are summarized in Table 2 . Pt/C-IM and Pt/C-OX catalysts were prepared via incipient wetness impregnation with H2PtCl6 followed by drying in air overnight at 100 °C. Afterwards, the dried sample was reduced in H2 at 320 °C for 6 h to produce the Pt/C-IM catalyst. Alternatively, the dried sample was further calcined at 420 °C for 6 h followed by a reduction in H2 at 700 °C for 5 h to make the Pt/C-OX catalyst. Pt/C-PR catalyst was prepared by precipitation of H2PtCl6 with Na2CO3 followed by a reduction in formic acid. After drying in air, the dried sample was reduced in H2 at 700 °C for 5 h. Pt/C-CL catalyst was prepared using a Pt-PVP colloid via wet impregnation. After drying in air, the sample was loaded to the APR reactor (vide infra) and treated in a hot compressed water stream (HCW, 2 mL/min) at 225 °C and 35 bar for 1 h, in order to remove PVP from the catalyst [20]. Afterwards, the sample was unloaded from the reactor and dried in air.Pt loading was semi-quantitatively analyzed by wavelength dispersive X-ray fluorescence (WDXRF) spectroscopy on S8 Tiger (Bruker) with the powder pellet method. An undiluted sample (ca. 0.5 g) was milled and loaded in a 29-mm die. Specific surface area (SBET) was determined from N2 physisorption measurement at -19,615 °C on Tristar 3000 (Micromeritics) according to the Brunauer-Emmett-Teller (BET) method [34]. Pt surface area and dispersion were determined by pulse CO chemisorption on ChemiSorb 2750 (Micromeritics). The catalyst was pretreated in He at 200 °C (5 °C min−1) for 1 h, followed by pulse chemisorption of CO at room temperature. Pt dispersion was calculated by assuming that the adsorbed CO to Pt ratio is 1 [35]. Pt size was measured by high-resolution transmission electron microscopy (TEM) using a CM300ST-FEG (Philips) operated at 300 kV acceleration voltage. The catalyst was ultrasonicated in ethanol, followed by deposition on a carbon-coated copper grid. Approximately 250 particles across 10 spots were counted. The same transmission electron microscope was also used to record TEM images in cross-section (XTEM) images of catalyst grains to measure the size of agglomerated Pt particles. The catalyst particles were embedded in a resin (the details are shown in Supplementary Information, SI), which allows the observation in cross-section in order to locate the Pt nano-particles concentrated on the catalyst surface. Approximately 200 Pt particles for the Pt/C-PR and Pt/C-CL catalysts, and 50 Pt particles for the Pt/C-IM catalyst were counted for analyzing the mean size of the agglomerated Pt particles. Pt content on the catalyst surface was analyzed by X-ray photoelectron spectroscopy (XPS) in a Quantera Scanning X-ray Microprobe (PHI) equipped with an AlKα monochromatic X-ray source (1486.6 eV). The catalysts with two different particle sizes of 100–250 μm (for grains as-prepared) and 20−40 μm (for powder after grinding) were analyzed.Aqueous-phase reforming of ethylene glycol solution (2.5 wt.% in water, feeding rate of 2 mL min−1) over the Pt/C catalysts (loading of 1 g) was carried out on a bench-scale continuous-flow fixed bed reactor setup (Fig. 1 ) at 225 °C and 35 bar for a time on stream (TOS) of 7 h. The details of the experimental setup and procedure, and of the product analyses are given in SI. Catalyst performance was defined and calculated by using Eqs. 1–8. (1) C o n v e r s i o n   o f   E G   % = ( 1 -   m o l   o f   E G   i n   p r o d u c t m o l   o f   E G   i n   f e e d )   ×   100 (2) C a r b o n   y i e l d   o f   p r o d u c t   % = m o l   o f   c a r b o n   i n   l i q u i d   o r   g a s e o u s   p r o d u c t m o l   o f   c a r b o n   i n   i n   f e e d   ×   100 (3) Y i e l d   o f   H 2   % =   m o l   H 2   p r o d u c e d m o l   c a r b o n   c o n v e r t e d   ×   1 R R   ×   X   ( c o n v e r s i o n )   ×   100 (4) E G   c o n v e r s i o n   r a t e   μ m o l E G   A P t - 1   m i n - 1 = μ m o l   o f   E G   c o n v e r t e d s u r f a c e   a r e a   o f   P t   ×   T O S   o f   30   m i n (5) H 2   p r o d u c t i o n   r a t e   μ m o l H 2   A P t - 1   m i n - 1 = μ m o l   o f   H 2   p r o d u c e d s u r f a c e   a r e a   o f   P t   ×   T O S   o f   30   m i n (6) T O F   f o r   H 2   p r o d u c t i o n   m o l H 2   m o l P t - 1   m i n - 1 = m o l   o f   H 2   p r o d u c e d m o l   o f   P t   ×   T O S   o f   30   m i n   (7) S e l e c t i v i t y   f o r   c a r b o n   s p e c i e s = m o l   o f   c a r b o n   i n   l i q u i d   o r   g a s e o u s   p r o d u c t m o l   o f   c a r b o n   i n   i n   f e e d   *   c o n v e r s i o n   o f   E G   ×   100 (8) S e l e c t i v i t y   f o r   H 2 = m o l   H 2   p r o d u c e d m o l   c a r b o n   c o n v e r t e d   ×   1 R R Various methods (Section 2.2 and Table 2) have been applied to prepare the four Pt/C catalysts with distinguishable Pt particle sizes and distributions (viz., with concentrated Pt particles on the surface or in a homogeneous fashion). BET surface areas (Table 2) of fresh Pt/C-IM, Pt/C-OX, and Pt/C-PR catalysts (340 - 372 m2 g−1) are relatively close to that of the Sibunit carbon support (350 m2 g−1). However, a decreased SBET was observed on the fresh Pt/C-CL catalyst (296 m2 g−1), indicating that wet impregnation with the Pt-PVP colloid influenced textural property of the Sibunit carbon though a low amount of Pt (0.7 wt.%, Table 2) was loaded on the Pt/C-CL catalyst. This is most likely related to Pt concentrated on the catalyst surface (vide infra), due to the lower penetration of the Pt colloid into the pores of the support.XPS analyses (Table 2, the corresponding spectra are shown in Figs. S1-S4) of the as-prepared catalyst grains (100–250 μm) and the after-ground powder (20−40 μm) show that the Pt concentration on the outer shell is higher than in the inner core of the Pt/C-IM, Pt/C-PR, and Pt/C-CL catalysts. Particularly, the difference is extremely large for the latter two catalysts, showing that the Pt concentrations on the outer surface are approximately 53 % (Pt/C-PR catalyst) and 4 times (Pt/C-CL catalyst) higher than the inner ones. Comparatively, the Pt/C-OX catalyst, which was prepared by incipient wetness impregnation followed by calcination and reduction at high temperatures, shows a similar Pt concentration on the outer surface and in the inner core. This might indicate that Pt was relatively homogeneously distributed in the Pt/C-OX catalyst, while Pt was more concentrated on the surface of the Pt/C-PR and Pt/C-CL catalysts.The speculation about concentrated Pt particles on the surface of the Pt/C-PR and Pt/C-CL catalysts is further confirmed by XTEM images of the catalysts as prepared, showing that the Pt particles are more visible on the catalyst grain edge than in the core (Fig. 2 C and D). XTEM images of the grain cores (Fig. 2C and D, right) display the fairly even distributed small Pt particles (< 3 nm). Comparatively, more concentrated small Pt particles are observed on the grain edges (Fig. 2C and D, left). Besides, agglomerated Pt particles with mean sizes of 11 nm and 24 nm (Table 2) are also present on the grain edges (with an approximate depth of 500 nm) of Pt/C-PR and Pt/C-CL catalysts. These are totally different from the XTEM images of the Pt/C-OX catalyst (Fig. 2B), showing that small Pt particles were evenly distributed on both the edge and in the core. No agglomerated Pt particles are observed on the Pt/C-OX catalyst, confirming the homogeneity of the Pt particles on the catalyst.The uniformly distributed Pt with a small particle size on the Pt/C-OX catalyst is also evidenced by the sharp Pt particle distribution of 1–3 nm (Fig. 3B) analyzed by TEM. The broader Pt particle size distributions for the Pt/C-PR (1–9 nm, Fig. 3C) and Pt/C-CL (1–6 nm, Fig. 3D) catalysts are most likely related to the larger Pt particles present in the outer shell of the Pt/C-PR and Pt/C-CL catalysts.The mean Pt particle sizes analyzed by CO chemisorption (Table 2) and TEM (Table S2) indicate that the Pt/C-IM and Pt/C-OX catalysts have smaller Pt particle sizes compared with the Pt/C-PR and Pt/C-CL catalysts. It needs to be noted that the Pt particle sizes estimated from CO chemisorption and TEM differ significantly, considering that only limited particles were counted from the TEM images (e.g., 200–400, Fig. 3) and CO chemisorption might over-estimate (Pt−CO stoichiometry) Pt surface area. Nevertheless, the above trends in the Pt particle sizes for the four Pt/C catalysts are similar according to these two analyses.As shown above, the catalysts prepared by the incipient wetness impregnation method, viz Pt/C-IM, and Pt/C-OX catalysts, have uniformly distributed small-size Pt particles. This is different from the catalysts prepared by precipitation (viz., Pt/C-PR catalyst) and wet impregnation with Pt colloid (viz., Pt/C-CL catalyst), which have small and also large Pt particles concentrated on the outer shell of the catalyst grains. According to the semi-quantified Pt content (by XPS, Table 2) and the visual Pt distribution (by XTEM, Fig. 2) on the catalyst edge and in the catalyst core for the four Pt/C catalysts investigated, the Pt/C-CL catalyst has the highest degree of Pt concentration on the catalyst surface, followed by the Pt/C-PR and Pt/C-IM catalysts (Table 2). Pt on the Pt/C-OX catalyst is distributed in a more homogeneous fashion as compared with that on the Pt/C-IM catalyst, indicating that high-temperature calcination and reduction enhanced the homogeneity of Pt on the Pt/C catalysts [36].Aqueous-phase reforming of ethylene glycol over the above four Pt/C catalysts were continuously performed on a fixed bed reactor at 225 °C and 35 bar for 7 h. Catalyst performance over TOS is shown in Fig.4 A in terms of EG conversion. In general, the initial EG conversion is comparable (e.g., 37.5–39.4 %) among the Pt/C catalysts investigated, except for the Pt/C-OX catalyst which shows a relatively lower EG conversion of 26.7 %. All the Pt/C catalysts exhibited excellent stability, evidenced by only a slight drop (ca. 4–5 %) in EG conversion after TOS of 3.5 h. Negligible deactivation occurred afterwards, indicating a steady state of the Pt/C catalysts for EG conversion. Accordingly, the products during TOS of 3.5–7 h were averaged to evaluate the representative products from APR of EG over the Pt/C catalysts.The excellent total carbon balance closures (e.g., 97–101 %) indicate negligible coke formation during APR of EG over the Pt/C catalysts. The selectivity’s to various products are shown in Table 3 . The major carbon-related products are gases, which consist of CO, CO2 and CH4 (Fig. 4B). A very small amount of EG was converted to liquid phase products (Fig. 4C) such as methanol, ethanol, acetic acid, glycolaldehyde and larger polyol (e.g., glycerol).The yield of the most interesting product, viz., H2 (Fig. 4B), differs dramatically with the Pt/C catalysts. Pt/C-PR catalyst has the highest H2 yield (26.4 %), followed by Pt/C-IM (24.1 %), Pt/C-CL (20.4 %) and Pt/C-OX (13.7 %) catalysts. This points to the different characteristics of active Pt sites on the four Pt/C catalysts.As discussed above, the Pt/C catalysts evolved to the steady-state after a TOS of 3.5 h (Fig. 4A). To correlate the catalyst characteristics with the catalytic performance during the 3.5–7 h TOS period (Fig. 4B and C), the used Pt/C catalysts after continuous-flow APR of EG for 7 h were characterized. The four used Pt/C catalysts showed comparable BET areas (326 - 343 m2 g−1, Table 2) with the fresh ones, indicating insignificant changes in catalyst pore structure after 7-h TOS. In addition, the Pt loadings on the fresh and used Pt/C catalysts (Table 2) are similar, showing a negligible loss of Pt under the severe APR reaction conditions.However, Pt particle sizes are larger on the used Pt/C catalysts than on the fresh ones according to both CO chemisorption and TEM analyses (Tables 2 and S2, and Fig. 3). The growth of Pt particles during APR reactions was often observed on supported Pt catalysts, e.g., Pt/C [13] and Pt/Al2O3 [8]. As a consequence, the Pt particle size distributions for the used Pt/C catalysts were broadened (Fig. 3), which is particularly significant for the Pt/C-PR catalyst (Fig. 3C). The mean Pt particle size on the Pt/C-PR catalyst was dramatically increased from 3.2 to 8.3 nm as measured by TEM (Table S2), and from 4.2 to 10.7 nm as measured by CO chemisorption (Table 2). Comparatively, the Pt/C-OX and Pt/C-CL catalysts show smaller changes on the Pt particle size. For the latter catalyst, the high-degree Pt concentration on catalyst surface with agglomerated Pt particles (Section 3.1) might have resistance to a further Pt agglomeration [37], which is reflected by the slightly increased Pt particle size (Table 2) from 24 nm for the fresh Pt/C-CL catalyst (Fig. 2D) to 30 nm for the used one (Fig. 5 C). Besides, the preparation method for the Pt/C-CL catalyst also has influence on the stability, e.g., by hydrothermal treatment to remove PVP and to stabilize the nanoparticles on the support [20]. Whereas for the Pt/C-OX catalyst, the stability of the Pt particle size might be related to the high-temperature calcination and the reduction enhancing Pt and C interaction [36]. As such, a further check of the presence of the agglomerated Pt particles on the used Pt/C-OX catalyst by XTEM was not carried out, considering that no agglomerated Pt particles presented on the fresh catalyst (Fig. 2B) as well.XTEM images of the used Pt/C-PR catalyst (Fig. 5B) show larger Pt particles (e.g., 20–40 nm) on the catalyst edge as compared with the fresh catalyst (Fig. 2C), resulting in a nearly doubled Pt particle size (Table 2). This is in good agreement with the change on mean Pt particle size (by TEM (Table S2) and CO chemisorption (Table 2)) on the Pt/C-PR catalyst after the APR reaction. Similarly, agglomerated Pt particles with a mean size of 17 nm (Table 2) were also formed on the used Pt/C-IM catalyst edge (Fig. 5A), in line with the increased mean Pt particle size from 2.7 nm (for the fresh catalyst, by CO chemisorption, Table 2) to 8.8 nm (for the used catalyst, Table 2). It needs to be noted here that no agglomerated Pt particles are observed in the core of the catalyst (Fig. 5-right), indicating the Pt agglomeration mainly took place on the surface of the Pt/C catalyst under APR conditions.It was demonstrated above that a stable catalytic performance in APR of EG in terms of EG conversion and H2 production was obtained over Pt/C catalysts, which were prepared by a different method in order to alter Pt size and Pt distribution on a Sibunit carbon support. In this contribution, we have used the diluted solution to investigate the relationship between APR performance and catalyst characteristics. For such a diluted stream, the industrial implementation of APR should be further considered, e.g., the economic feature related to the energy consumption for heating the H2O.To recall, the Pt/C catalysts prepared by a general method as incipient wetness impregnation (viz., Pt/C-IM and Pt/C-OX catalysts), have evenly distributed Pt particles with small sizes. A high-temperature treatment, e.g. calcination followed by reduction, was applied to strengthen the interaction between Pt and carbon support. As a consequence, the Pt/C-OX catalyst showed much better stability on Pt size and Pt distribution under the severe APR conditions than the Pt/C-IM catalyst having agglomerated Pt particles on catalyst surface after 7-h APR of EG. Alternatively, the Pt/C catalysts prepared by precipitation method (Pt/C-PR catalyst) and a more novel method of impregnation of pre-prepared Pt colloid (Pt/C-CL catalyst) obtained small Pt particles, as well as agglomerated Pt particles concentrated on the catalyst grain edge. It seems that the inhomogeneous Pt distribution formed on the surface of Pt/C-PR and Pt/C-CL catalysts, and the degree of Pt concentration on the surface of the latter is higher than that of the former. Compared with the Pt/C-CL catalyst, the fresh Pt/C-PR catalyst has more amount of agglomerated Pt particles with a smaller size, resulting in a bigger mean Pt particle size (Table 2). However, these small Pt particles on the Pt/C-PR catalyst grew faster than those large Pt particles on the Pt/C-CL catalyst under the APR reaction conditions. As such, the mean Pt particle size for the Pt/C-PR catalyst increased remarkably, while only a slightly increased mean Pt particle size was observed for the Pt/C-CL catalyst.Since all the Pt/C catalysts were prepared using the same Sibunit carbon support, any difference observed in the chemistry, e.g., product distribution, over different Pt/C catalysts (Fig. 4) should be related to Pt characteristics, e.g., Pt particle size and its distribution. In order to properly correlate the catalyst performance with the catalyst characteristics, the in-situ characterizations of the catalyst during APR is required, e.g., by an in-situ attenuated total reflectance Fourier transform infrared (ATR-IR) technique [26]. However, this is very challenging for Pt/C catalysts, due to the fact that the refractive index of carbon and the internal reflection element (ZnSe) is too similar to obtain ATR-IR spectra for carbon-supported catalysts. The fresh catalyst might change greatly under APR conditions even after a short TOS [38], leading to an inappropriate relationship between initial catalyst performance with fresh catalyst characteristics. Considering that the Pt/C catalysts evolved to a relatively steady state after TOS of 3.5 h (Section 3.2), it might be assumed that Pt/C catalyst characteristics remain stable during TOS of 3.5–7 h. Therefore, the averaged EG conversion (Fig. 4A) and H2 production (Fig. 4B) during TOS of 3.5–7 h, and the characteristics of Pt on the used Pt/C catalysts (Table 2) after TOS of 7 h were used to calculate reaction rates by using Eqs. 4 – 6. In addition, Pt can be taken as metallic Pt during the APR reactions, considering that the pre-reduction of the Pt/C catalysts was performed at temperatures higher than the reduction temperature of PtOx for Sibunit carbon supported Pt catalysts (e.g., Tmax of 125 °C [22]). Even though there might be a very small fraction of PtOx species due to the partial oxidation during the storage and loading to the reactor [22], they would probably be reduced by the H2 formed during APR at a reaction temperature of 225 °C. In order to study the effect of Pt size on catalyst performance, EG conversion and H2 production rates based on the available Pt surface area (μmolEG(or H2) APt −1 min−1) are shown in Fig. 6 A. The mean Pt particle size (Fig. 6A) and the mean size for the agglomerated Pt particles (Fig. 6B) were analyzed by CO chemisorption and XTEM, separately.It is interesting to observe that the rates for both EG conversion and H2 production increased linearly with the increased Pt particle sizes. Comparatively, the sensitivity to Pt particle size for H2 production rate is higher than for an EG conversion rate, as indicated by the slopes of the fitted lines in Fig. 6A. This result is consistent with that reported by Lehnert et al. [21], who also observed a higher H2 production from APR of glycerol over Pt/Al2O3 catalysts with a bigger Pt particle size. This is most likely related to the enhanced C-C cleavage of oxygenates on more Pt surface forming H2, in turn competing for C-O cleavage reaction yielding low hydrocarbons [2]. The extremely low yields of C1 and C2 hydrocarbons (Fig. 4B) also confirm this.It needs to be highlighted here that the mean Pt particle sizes on the Pt/C catalysts in this study are quite big (e.g., 3–11 nm in Fig. 6A), related to the presence of large Pt particles. The correlations between Pt particle size and the rates for EG conversion and H2 production (Fig. 6-B) suggest that a mean size for agglomerated Pt particles of ca. 20.7 nm is the most suitable. There is a trade-off of the size of the agglomerated Pt particles for an optimal H2 production rate over Pt/C catalysts, due to the fact that the number of exposed surface Pt atoms continues to decrease as the size of agglomerated Pt particles increases. As a consequence, the Pt/C-PR catalyst, which has a number of Pt particles with small size concentrated on the catalyst grain edge, has the highest TOF for H2 production of 248 molH2 molPt −1 min−1 (Fig. 7 ). Comparatively, the Pt/C-CL catalyst of which the level of Pt concentration on the surface is the highest has a much lower TOF-H2 (100 molH2 molPt −1 min−1, Fig. 7), due to the presence of Pt particles with a large size. For the Pt/C-IM catalyst, which has the Pt particles relatively homogeneously distributed both on the grain edge and in the core of the catalyst, has a much higher TOF-H2 (78 molH2 molPt −1 min−1, Fig. 7) than the Pt/C-OX catalyst (18 molH2 molPt −1 min−1, Fig. 7). This is obviously related to the bigger mean Pt particle size for the Pt/C-IM catalyst compared with the Pt/C-OX catalyst.Of great interest is that the highest TOF-H2 (248 molH2 molPt −1 min−1) obtained on the Pt/C-PR catalyst in this study is much higher than those reported Pt/C catalysts (Table 1) by Shabaker et al. (7.5 molH2 molPt −1 min−1) [28] and Kim et al. (103 molH2 molPt −1 min−1) [32], representing the best performance of Pt/C for APR of EG to bio-H2. Furthermore, what is significant is that the TOF-H2 of Pt/C-PR is close to the top two catalysts (viz., Pt/AlO(OH) catalyst with a TOF of 300 molH2 molPt −1 min−1 [26] and Pt/SiO2 catalyst with a TOF of 275 molH2 molPt −1 min−1 [27]) developed for APR of EG so far.Stability of Pt/SiO2 catalyst, related to the leaching of silica under APR conditions, is a critical issue for a long-term practical application [39]. Pt/AlO(OH) catalyst might have a high tendency for coke formation during APR [40] due to the acidity of AlO(OH) support [41]. It has been demonstrated in this study that the Sibunit carbon is stable under APR conditions and coking on Pt/C catalyst is negligible during a continuous 7-h APR of EG (Section 3.2). Besides, using carbon as a carrier for supported Pt catalysts ensures that it is easy to harvest Pt for recycling after usage by burning [42]. Having a high intrinsic activity for hydrogen production and an excellent stability for long-term operation, Pt/C catalyst could definitely be an excellent catalyst for APR of oxygenates for bio-H2 production.Pt concentrated on the catalyst surface with small-size Pt particles on Pt/C catalyst is advantageous for APR of a small molecule (viz., EG), which might also be significant for larger oxygenates. Further exploitation of Pt/C catalyst for APR of the aqueous phase of pyrolysis oil or other waste aqueous oxygenate streams is thus recommended. On the other hand, a large amount of CO (e.g., carbon yield of 18–25 %, Fig. 4B) were also formed during APR of EG over Pt/C catalysts. This indicates an inefficient water-gas shift (WGS, CO + H2O → H2 + CO2) reaction, in line with the low yield of CO2 (Fig. 4B). Therefore, bio-H2 production over Pt/C catalysts via APR could be further improved, e.g., by adding a second metal such as Ni to enhance WGS reaction (e.g., Pt-Ni/Al2O3 catalyst for APR of EG [43]).Catalyst preparation protocols, including the incorporation of the metal precursor (e.g., incipient wetness impregnation, precipitation, and impregnation of Pt colloid) and further treatment (e.g., high-temperature calcination and reduction), affect Pt size and Pt distribution (homogeneous Pt distribution and with concentrated Pt particles on the surface).Pt/C catalysts showed excellent H2 yields (up to 24.1 %) for aqueous-phase reforming of ethylene glycol and excellent catalyst stabilities with a slight drop (ca. 4–5 %) in EG conversion (ca. 37.5–39.4 %) after 3.5-h TOS. The characteristics of the used catalysts after 7-h APR of EG, which were in a steady-state, were used to correlate the catalyst performance. The linear relationships between mean Pt particle size (in a range of 3–11 nm investigated) and the rates for EG conversion and H2 production were observed.Pt/C-PR catalyst, which was prepared by the precipitation method, had small Pt particles distributed in the catalyst as well as large Pt particles concentrated on the catalyst grain edge after TOS of 7 h. Pt/C-PR catalyst showed the highest turnover frequency for H2 production (TOF-H2 of 248 molH2 molPt −1 min−1) among the four Pt/C catalysts investigated. This was attributed to the preferred Pt particles concentrated on the catalyst surface with the biggest mean Pt particle size (ca. 10.7 nm) and the appropriate mean size of agglomerated Pt particles (ca. 21 nm). This superb TOF-H2 and the excellent stability of the Pt/C catalyst make it promising for APR of EG as compared with the state-of-the-art Pt catalysts, viz., Pt/AlO(OH) (TOF-H2 of 300 molH2 molPt −1 min−1) and Pt/SiO2 (TOF-H2 of 275 molH2 molPt −1 min−1) catalysts. Pt/C catalysts are therefore recommended for APR of other model oxygenates (e.g., hydroxyacetone) present in waste streams and also APR of real waste streams (e.g., the aqueous phase of pyrolysis oil) to make renewable and green H2.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. A.K.K. Vikla: Investigation, Conceptualization, Methodology, Validation, Writing - original draft, Writing - review & editing. I. Simakova: Resources, Validation, Supervision, Writing - review & editing. Y. Demidova: Investigation, Resources, Validation. E.G. Keim: Investigation, Resources, Validation, Writing - review & editing. L. Calvo: Investigation, Resources, Validation. M.A. Gilarranz: Resources, Supervision, Writing - review & editing. Songbo He: Conceptualization, Writing - original draft, Supervision, Writing - review & editing. K. Seshan: Supervision, Writing - review & editing, Funding acquisition.The research was funded by European Union Seventh Framework Programme (FP7/2007-2013) within the project SusFuelCat under grant agreement No. 310490. Dr. Vikla would like to thank Ing. B. Geerdink for his technical and emotional support when carrying out the APR experiments, and also Ing. Benno Knaken for his expertise in maintaining the high-pressure reactors. Prof. L. Lefferts is thanked for preliminary discussions. Besides, Mrs. K. Altena-Schildkamp is thanked for BET and CO chemisorption measurements. Mr. Tom Velthuizen is thanked for XRF characterization, and Mr. Gerard Kip for XPS analysis at MESA + NanoLab. IS acknowledges with support from Ministry of Science and Higher Education of the Russian Federation.Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.apcata.2020.117963.The following are Supplementary data to this article:

Pt/C catalysts with varied Pt sizes and distributions were investigated for aqueous-phase reforming (APR) of ethylene glycol (EG) to H2. APR experiments were performed on a continuous-flow fixed bed reactor with a catalyst loading of 1 g and EG feeding of 120 mL h−1 at 225 °C and 35 bar for 7 h. The fresh and used Pt/C catalysts were characterized by XRF, BET, CO chemisorption, TEM, XTEM, and XPS. Catalyst preparation protocols changed Pt characteristics on Pt/C catalysts, leading to a distinguishable H2 production. The rates for EG conversion and H2 production increased linearly with mean Pt size (3–11 nm), while having a volcano relationship with the mean size of agglomerated Pt particles (17–30 nm). Pt with concentrated Pt particles on surface of Pt/C catalysts was more preferable for APR of EG than the homogeneously distributed in catalysts. Optimal performance was obtained over a Pt/C-PR catalyst, which was prepared by precipitation method, showing a superb turnover frequency of 248 molH2 molPt −1 min−1 for H2 production from EG in APR. Besides, Pt/C catalysts also showed excellent stability. These results have shown the promise of Pt/C catalyst for APR of EG, which can be extended for bio-H2 production via APR of biomass-derived oxygenates in waste streams.

Chemical looping gasification (CLG) is recognized as a novel technology that has advantages in energy utilization and syngas quality improvement [1,2]. Compared with conventional gasification, CLG replaced gaseous oxygen with an oxygen carrier (OC). The OCcommonly consists of a metal oxide, thus, the product, syngas, will not be diluted in a nitrogen source from air. On the other hand, reduced metals, such as Ni and Fe, have function in tar cracking catalysis. A lower reaction temperature also reduces the generation of NOx from SO2 [3]. Therefore, CLG is considered more appropriate for organic solid waste management or other raw materials with complicated components [4].Wast pulp (WP) is the main solid waste from the paper industry, and the reuse of WP is an important problem for the optimization of the whole industry chain of the paper industry. The main constituent of WP is cellulose, it possesses rich resource and energy potential. However, the plentiful volatiles cause considerable tar generation in the gasification process. Therefore, CLG is a flexible and efficient technology for WP management.In the CLG process, the OC transfers oxygen and heat while catalysing the pyrolysis of raw materials. An ideal OC should have good activity in redox reactions, good mechanical strength and high thermodynamic stability. Cost is also the important factor to be considered. Owing to the complicated components of WP or other organic solid waste, the inactivation of the OC source from sintering and inert component generation caused by ash and other pollutants must be considered seriously. On the other hand, cheap ore and solid waste, such as slag and sludge, are rich in transition metal oxides and hence have the potential for use as low-cost OCs in CLG processes [5,6]. Guo et al. [7] tested the performance of natural copper ore and haematite as an OC for biomass CLG, and the results indicated that the natural ore has better mechanical behaviour and thermal stability than pure CuO or Fe2O3. Schmitz et al. [8,9] used manganese ore as an OC to assist the oxidation of biomass char in the chemical looping process. The results showed that manganese ore has excellent redox behaviour, but the mechanical strength needs to be strengthened. Deng et al. [10] explored the behaviour of copper slag in municipal sludge CLG and found that alkali and alkaline earth elements such as K and Ca can improve CLG activity. Chen et al. [11] investigated bauxite residual for H2-rich syngas production via the CLG process and proved that the bauxite residual could be a high-reactivity OC in the CLG process. Moghtaderi et al. [12] compared concrete waste and CaO behaviours in syngas upgrading in CLG of biomass and revealed that concrete waste has a longer cycle life than pure CaO. The research of Yang et al. [13] revealed the inactivity mechanism of phosphogypsum OC in the CLG process, in which the generation of transition metal silicate gradually reduced the activity of the OC. Although the studies listed above showed the potential of natural minerals in CLG, the low redox activity and cycle performance restrict the practical application of low-cost natural minerals. Introducing other ingredients to modify natural minerals has been recognized as an efficiency solution. Ryden et al. [14] used NiO as an additive to modify waste products from the steel industry and proved that only a small amount of NiO can improve the activity for Fe-based OCs. Sun et al. [15] explored the behaviour of manganese-modified ilmenite as an OC in biomass CLG and found that the generation of Fe-Mn oxide would improve the activity of the OC. The research of Bao et al. [16,17] revealed that heterogeneous ions such as K+ and Ca2+ promoted the redox reaction of iron ore. Tian et al. [5] used cement to modify copper ore, which not only promoted the gasification activity of copper ore but also improved the thermal stability and mechanical behaviour. Otomo et al. [18] explored the effect of calcium nitrate melt infiltration for biomass chemical looping process and found that the calcium salt improve the behaviour of ilmenite obviously. The research of Yan et al. [19] revealed that heterogeneous ions enhance fuel conversion in chemical looping process with red mud as OC. In summary, an economic effect and high performance can be achieved by introducing a small amount of artificial ingredients.Aiming at cellulose solid waste, we developed multimetal composite OCs [20–22]. The results revealed that the Ca-based material has an obvious catalytic function in the reforming of H2O and tar and char decomposition. Base on Ni-based material was active in tar cracking catalysis, the Ca-Fe OC modified by NiO achieved complete conversion of cellulose to syngas at 850 °C [21]. Therefore, the complexes of Ca, Fe and Ni should be considered as potential OCs for WP management. Ni-containing electroplating sludge (NES) is an important solid waste for the electroplating industry. The abundant heavy metal content gives it a greater potential to become an high-efficiency OC. Our previous research [23] explored the reactivity of NES as an OC in dyeing sludge CLG. A carbon conversion efficiency of 81 % and 0.37 Nm3/kg syngas could be obtained at 850 °C. Considering that NES is rich in Ni, Fe, and Ca, NES should be a potential OC for WP gasification. However, the relatively lower carbon conversion efficiency and H2 yield revealed that NES might be incompetent for tar cracking and inappropriate for raw materials with high volatility, such as cellulose or WP. The Ni in NES is combined with other metals and forming a complexed compound such as NixFeyOz [24,25]. It has more reduced steps and makes it more difficult for the Ni element to form as Ni0 to exert its catalytic function. Hence, NiO was introduced to modify NES to promote tar cracking and syngas generation in this study. Ni-based materials should be placed on the outside of OC particles to easily contact reactants and performed as catalysts for tar cracking and H2 release [26,27]. The immersion method has been proven to be an effective method to realize the above assumptions [21]. Thereby, the NES modified by NiO via the immersion method, i.e., NiO-modified NES (NNES), was used as the OC in WP CLG in this research. The reactivity was explored under different working conditions in a fixed bed, and a few characterizations were conducted on the OC to reveal the reaction characteristics of the OC.WP was collected from a paper factory in Guangzhou, and the received WP was predried. The sample was dried at 105 °C for 20 h to obtain a usable material. Table 1 shows the proximate and ultimate analyses of the usable material. The high content of volatiles and the molar ratio of C/H/O (∼6:11:5.5) were similar tocellulose (C6H10O5), itcorroborated that the main composition of WP was cellulose. Under ideal circumstances, all C should be formed as CO and H should be formed as H2 after gasification, therefore the ideal syngas yield should be 1.51 Nm3/kg. Notably, small amounts of ash and N could be found in WP, it reveals that other compositions should exist and may influence the pyrolysis and gasification behaviour of WP, resulting in different characteristics from pure cellulose. Compared with pure cellulose [20], the fixed carbon increase from 5 % to 7% to about 10 %, and the content of H2 in thermolysis gas was much higher than pure cellulose. Noticeably, the small amount of nitrogen may come from the biomass components such as proteins. In CLG process, the raw material is mainly under reducing atmosphere, and the generation of pollution such as NOx could be avoided.NES was collected from an electroplating factory in Guangzhou. The raw NES was dried at 105 °C for more than 48 h, calcined at 950 °C for 6 h, ground and sieved to 100 mesh as a useable sample. Table 2 shows the element composition of the NES (measured by X-ray fluorescence (XRF)). It can be seen that the main composition of NES is oxides of Fe, Ni and Ca, and the mole ratio of Fe/Ni/Ca is approximately 2.8:2:1. The content of S is approximately 3.42 wt%, and it may derved from SO4 2-. Notably, the content of Mg is also higher than 1 wt%. The existence of Mg2+ will promote the generation of syngas and reduce the activation energy of pyrolysis [28], and benefit for WP gasification.NNES was synthesised by the immersion method as shown in Fig. 1 . Briefly, NES was placed into a nickel nitrate (Ni(NO3)2·6H2O, AR) solution at the set concentration (in completed NNES, the mass ratio of NiO and NES was 0, 0.1, 0.5, and 1.0) with vigorous stirring, and the as-formed samples were aged for 24 h, dried at 105 °C for 20 h and calcined at 950 °C for 6 h to obtain NNES.The WP CLG test was conducted in a fixed bed experimental device as shown in Fig. 2 . The fixed bed experimental device consisted of a quartz tube (length: 800 mm and inner diameter: 17 mm), temperature control device, gas control device, water injection pump, tar/water filter devices and gas collection bag. The tar/water filter devices were three scrubbers in series. The first two scrubbers were filled with ethylene glycol (C2H6O2, AR) and placed in an ice water bath. The third scrubber was filled with allochroic silica gel. In a typical experiment, the mixed OC and WP were placed in a quartz hanging basket (height: 50 mm and inner diameter: 3 mm) and hung in a nonheated area in advance. Then, 20 mL/min nitrogen was introduced into the quartz tube, and the heated area was heated to the designated temperature. Deionized water was injected into the quartz tube early to create a steam atmosphere. With the start of the experiment, the hanging basket was sent to the centre of the heat area of the quartz tube. Meanwhile, gas bags were used as collection devices to collect the gas flowing through the fixed bed device. Each experiment lasted 1 h. In the cycle performance experiment after the gasification process, 100 mL/min air was introduced into the reactor, which lasted 30 min.The gas was analysed by refinery gas chromatography (Agilent 7890A), and the yield of gaseous products was calculated as follows: (1) Y i = 60 q N 2 C i C N 2 m C E where Y i represents the yield of the gaseous product (Nm3/kg), i represents gaseous products such as CO, CO2, H2, CH4, and C2H4, C i represents the concentration of gaseous products, C N 2 represents the concentration of nitrogen, q N 2 represents the flow of nitrogen (Nm3/min), and m C E represents the mass of CE (kg).The total syngas yield was calculated as follows: (2) Y t o t a l = ∑ Y i where Y t o t a l represents the total gas yield (Nm3/kg).The effective syngas yield is the yield of flammable gas and is calculated as follows: (3) Y e f f e c t i v e = Y t o t a l - Y C O 2 where Y e f f e c t i v e represents the effective syngas yield and Y C O 2 represents the CO2 yield.The carbon conversion efficiency was calculated as follows: (4) η c = 12 ∑ j i Y i V m m carbon m CE · 100 % where η c represents the carbon conversion efficiency, j i represents the number of carbon atoms in molecule i , V m is the gas molar volume (Nm3/mol), m c a r b o n represents the mass of carbon in CE (kg), and m C E represents the mass of cellulose.The lower heating value (LHV, MJ/Nm3) was calculated as follows: (5) LHV = 10.82 Y H 2 + 12.54 Y C O + 35.88 Y C H 4 + 59.44 Y C 2 H 4 Y t o t a l The OC was characterized with X-ray diffraction (XRD, Panalytical X’pert Pro diffractometer) and scanning electron microscopy (SEM, Hitachi SU70 instrument) with energy dispersive spectroscopy to reveal the change in NNES in the CLG process.Considering the environmental toxicity and cost, the addition of NiO should be as little as possible. Based on this, we explored the effect of the addition contents of NiO in the WP CLG process, and the results are shown in Fig. 1. With the addition of NiO, the H2 yield was obviously improved, which caused the growth of the total and effective syngas yields and H2/CO. Especially for the sample with mN/mNES = 0.1, the carbon conversion efficiency achieved is approximately 85 % at 850 °C, the total syngas yield reached 1.19 Nm3/kg (79 % of ideal syngas yield) and H2/CO was 1.2 at 850 °C without steam addition. Compared with the NES, the total syngas yield increased by approximately 29 %, and the H2 yield increased by approximately 112 %, which shows the excellent activity of NiO addition on H2 release and tar or char pyrolysis. However, the carbon conversion efficiency did not increase with NiO addition. This phenomenon indicates that NES already has excellent tar and char gasification activity, and the independent NiO shows no obvious enhancement in tar or char cracking compared to the Ni-Fe complex oxide [29–33]. Noticeably, as mN/m NES increased to 0.5 or higher, the syngas yield, carbon conversion efficiency and H2/CO markedly decreased. The results revealed that NiO might cover the surface of OC grains and hinder the synergy of NiO and NES. Therefore, the small amount of NiO addition is also based noon the consideration of OC activity maximization. According to Fig. 3 , when the best syngas yield and H2/CO were obtained, the negative impact of NiO on carbon conversion efficiency can be ignored. Hence, mN/mNES = 0.1 was used as the standard in further experiments, and NNES refers to samples with mN/mNES = 0.1 in the following unless otherwise specified.The effect of the added amount of OC is shown in Fig. 4 . Compared with WP pyrolysis in N2 atmosphere, the addition of NNES effectively promotes gasification. The total syngas yield improved by approximately 60 % from 0.75 Nm3/kg (50 % of ideal syngas yield) to 1.19 Nm3/kg, the effective syngas yield increased approximately 42 % from 0.68 Nm3/kg (45 % of ideal syngas yield) to 0.96 Nm3/kg (64 % of ideal syngas yield), and the carbon conversion efficiency increased from 72 % to 84 %. The increase was mainly from the increase in CO2 and H2, which proved the activity of NNES on oxygen release and volatile reforming assistance. Markably, when mOC/mWP was higher than 1.0, the syngas yield decreased obviously. It can be due to the decrease in the H2 and CO yield. However, the CO2 yield increased slightly. The result reveals that the excessive addition of NNES would cause deeper oxidation of syngas and H2 and CO would be further oxidized to H2O and CO2, causing a decline in the syngas yield. Therefore, mOC/mWP = 1 should be the appropriate ratio for WP CLG.Steam is universally used as an assistant gasification agent in gasification processes [34] to improve the carbon conversion efficiency and H2 yield. In the CLG process, the synergy of steam and OC is the focus for researchers. With steam addition, the process of organic compound gasification can follow the following path: (R1) C n H m O l + H 2 O → C O k + H 2 (R2) C n H m O l + O C ( o x i d i z e d ) → C O k + O C ( r e d u c e d ) (R3) C n H m O l → a C O k + b H 2 + C n - a H m - 2 b O l - a k (R4) H 2 O + C O → H 2 + C O 2 (R5) H 2 O + O C ( r e d u c e d ) → H 2 + O C ( o x i d i z e d ) (R6) C O + O C o x i d i z e d → C O 2 + O C ( r e d u c e d ) where a, b, and l = 0,1,2,3,……; n and m = 1,2,3,……; C n H m O l refers to the raw material, i.e., tar or char; and k = 1,2.As shown in R1-R6, the addition of H2O improves the generation of CO2 and H2. The phenomenon is exhibited in Fig. 5 . The yield of CO2 and H2 increased obviously with the amount of added water; but the CO yield decreased, which corresponds to R4-R6. The carbon conversion efficiency increased from 84 % to 90 % with steam addition, which indicates that the synergy of H2O and OC promotes the gasification of tar or char. Notably, H2/CO increased from 1.2 to 3.6 gradually when the water flow increased from 0 to 2.4 mL/g(WP) but remained at 3.6 when the water addition flow was higher than 2.4 mL/g(WP). The results reveal that H2/CO of produced syngas can be adjusted by injecting water at below 2.4 mL/g(WP); however, excess steam addition has no noticeable effect on the products. Considering the energy and resource savings, 2.4 mL/g(WP) should be the proper water addition flow for the generation of hydrogen-rich syngas. Fig. 6 shows the effect of temperature. Fig. 6 (a) and (b) show that when the temperature increased from 800 °C to 850 °C, the total and effective syngas yields and carbon conversion efficiency increased sharply, which might be due to oxygen release from the Fe-based material [35]. When the temperature was higher than 850 °C, the temperature has little effect on the carbon conversion efficiency. However, the H2 yield decreased sharply when the temperature exceeded 850 °C, which might be due to further oxidization of H2. Noticeably, the variation of H2/CO at high temperature indicates that the temperature should be the essential factor for syngas ingredient adjustment.The performance of the NNES in a steam atmosphere was different from that in a dry atmosphere. According to Fig. 6 (c) and (d), the total and effective syngas yields increased slightly with increasing temperature. The carbon conversion efficiency was similar at different temperatures in a steam atmosphere. However, a higher temperature is beneficial to H2O reforming (R4-R6). Thereby, H2/CO improved obviously as the temperature increased. Based on product quality and energy savings, 850 °C should be the appropriate temperature for WP CLG with NNES as the OC.The redox cycling performance of the NNES in a steam atmosphere at 850 °C is shown in Fig. 7 . According to the curves of the total and effective syngas yields, the total syngas yield fluctuates around approximately 1.45 Nm3/kg, and the effective syngas yield fluctuates approximately 1.20 Nm3/kg. The activity of NNES for WP gasification does not decline after 10 redox cycles, which reveals the excellent cyclic behaviour of NNES, especially for H2O reforming. However, the carbon conversion efficiency declined by approximately 10 % from approximately 95 % to 85 %, which indicates that the tar cracking activity of NNES decreased gradually after the redox cycles. The inactivation of NNES might be due to agglomeration and subsidence of surface-active components, the OC characterization presented in Section 3.2 will corroborate it. Notably, the decline in carbon conversion efficiency indicates a decrease in CO yield, and results in the growth of H2/CO after redox cycles.Overall, a small amount of NiO will effectively improve the gasification activity of NES by enhancing H2 generation and tar cracking. For WP CLG, 850 °C and mOC/mWP = 1 should be the appropriate conditions for syngas generation; with steam injection, the H2 yield will be increased sharply, and H2/CO could be adjusted from 1.2 to 3.6 with NNES as the OC. After 10 redox cycles, NNES maintains a favourable activity on syngas generation and volatile reforming. NNES should be a potentially low-cost OC for cellulose solid waste CLG.To further explore the reaction process of the NNES, XRD and SEM combined with EDS were conducted, and the results are shown in Fig. 6 and Fig. 7. Fig. 6 shows the XRD patterns of the OCs. It can be found that in NES, Ni and Fe were comparable to NiFe2O4, an efficient OC for biomass CLG [36]. Ca is formed as Ca2Fe2O5, which is a high-activity compound for H2O reforming and H2 generation [37,38]. The results revealed that NES is a natural OC with high activity for WP CLG. The crystal composition of NNES is similar to that of NES. However, in the process of NNES preparation, the samples underwent multiple calcination cycles, and the same elements were more prone to agglomerate and cause crystal phase separation; therefore, Fe2O3 and SiO2 could be observed in the XRD pattern of NNES. After reaction with WP at 850 °C with 2.4 mL/g(WP) water injection, although the main composition of NNES was still NiFe2O4, the characteristic peaks of Ni0 or Fe0 can be observed in the XRD pattern. The results indicates the good oxygen releasing activity of NNES. On the other hand, the appearance of Ni0/Fe0 would promote tar cracking and H2 generation. Notably, Ca2Fe2O5 can be found in the XRD pattern more obviously, which reveals that with the reduction of Ni, Ca tended to combine with Fe and form Ca2Fe2O5. According to previous studies [21,39], Ca2Fe2O5 is an excellent intermediary to realize the reforming of H2O and reducing substances by an inert redox cycle as follows: (R7) 3 H 2 O + 2 C a O + 2 F e → 3 H 2 + C a 2 F e 2 O 5 (R8) C n H m O l + C a 2 F e 2 O 5 → 2 C a O + 2 F e + m 2 H 2 + k C O + ( n - k ) C O 2 The generation of Ca2Fe2O5 explained the higher H2 yield with steam addition. NNES supplied a platform for H2O absorption and reforming [20], thereby maintaining a high effective syngas yield with water injection.The XRD pattern of regenerated NNES shows the crystal composition of NNES after a redox cycle. This result was similar to that of Fresh-NNES, which shows the excellent regeneration performance of NNES. However, the peaks of SiO2 are more obvious, showing that crystal phase separation still appeared after the redox reaction. Noticeably, CaSO4, a potential effective OC for biomass CLG [40], can be found in the regenerated OC. The sulfate in NNES is also an active ingredient and it enhance the oxygen-carrying capacity of NNES. In fresh OC, sulfate might exist as other forms and can’t be observed in XRD pattern. In the redox cycles, Ca would combine with other anions such as SO4 and form as CaSO4. However, CaSO4 would release oxygen and might not be regenerated in such complexed compound as NNES. Therefore, small amount of SO2 may be generation in multiple redox cycles. The further study of migration of hazardous elements in NNES is necessary.After 10 redox cycles, though NiFe2O4 is still main crystal composition of NNES, the characteristic peak of silicate is manifest in XRD pattern. Inert compound such as CaFeSi2O6 can be found in NNES, it reveals that multiple redox reaction would cause active cation combining with SiO2 and reduce the active ingredient. The generation of silicate results in the declining of oxygen-carrying capacity of NNES, and explains the decrease of carbon conversion efficiency after 10 redox cycles.The morphological changes are provided by SEM images shown in Fig. 9. As the Fig. 9 (a) shown, the NNES was composited by grains of tens to hundreds of nm in diameter. However, with the reduced of active composition, the size of grains increased obviously in Fig. 9 (b), reveals the agglomeration happened in gasification reaction. The sintering can be also observed by comparing Fig. 9 (a) and Fig. 9 (c). The regenerated NNES was consisted of grains whose diameter was over 250 nm, which much bigger than that in fresh NNES. After 10 redox cycles, there’s no obviously grains can be observed on surface of NNES, it indicates the serious agglomeration has been happened. The agglomeration decrease the surface area of NNES, impede the contact of active composition and reactant, thereby cause the inactivity of NNES.The EDS results are shown in Table 3 . The important elements are listed separately in the table, and the elements with less than 1 % content, i.e., Mg, Al, Mn, P, etc., were aggregated in others. Compared with the results of the XRF analysis of the NES in Table. 2, it can be summarized that Fe and Ni tend to distribute on the surface in the NNES. The loading of Ni by the immersion method is prone to distribute Ni on the outside of the NES grains and attracts Fe atoms to form NiFe2O4 on the surface of the NNES. After reduction, Ni0/Fe0 will be prone to stay on the surface and exert a better catalytic effect. For the reduced OC, the O and Ca contents on the surface increased sharply, which maybe because the reduction of NiFe2O4 exposed inert oxides such as CaO, and Ni/Fe tended to agglomerate. The content of Ni, Fe or Cr decreased in regenerated NNES compared with fresh NNES, which proves the agglomeration of transition metal elements in NNES. Notably, the S content increased in the regenerated sample, which corresponds to the generation of sulfate, as observed in Fig. 8 . In the sample after 10 redox cycles, the content of metal elements decreased further. However, might be due to the generation of inert silicate, the content of Si increased substantially. Noticeably, the content of S dramatically decreases, it confirms the presence and consumption of active sulfates such as CaSO4.In total, the characterization of the NNES explained the inactivity of the OC after multiple redox cycles. The agglomeration of the active element hinders the contact of the catalyst and reactant. Crystal phase separation due to agglomeration inhibits the synergy of different compositions [21]. Meanwhile, redox reactions promote the combination of metal positive ions with Si and the generation of inert silicate, and reduces the oxygen carrying capacity and reaction activity of NNES.A small amount of NiO was introduced to modify the NES, and greatly improved activity for WP CLG. 10 wt% NiO was considered the most appropriate addition amount for NES modification. At 850 °C, when mOC/mWP = 1 with 2.4 mL/g(WP) water injection, 1.73 Nm3/kg syngas with an LHV of 11.9 MJ/Nm3 and H2/CO of 3.63 was obtained from WP CLG and the carbon conversion efficiency was over 90 %. The above reaction conditions are considered the most appropriate work conditions for H2-rich syngas produced. H2/CO can be adjusted easily by steam addition and temperature control. After 10 redox cycles with the above conditions, the syngas yield shows no obvious decline. However, the decreased carbon conversion efficiency reveals the slow inactivation of NNES.XRD and SEM combined with EDS was conducted to reveal the reaction mode and inactivation mechanism of NNES. The results indicate that the redox reaction promotes the agglomeration and weakens the synergy of different compositions and catalyst contact areas. Multiple redox cycles also cause inert silicates generation and reduce the oxygen carrying capacity and reaction activity of NNES.Overall, NNES should be a potential OC with high efficiency and low cost for cellulose solid waste CLG such as WP CLG. Genyang Tang: Methodology, Investigation, Writing – original draft. Jing Gu: Conceptualization, Data curation, Writing – review & editing. Guoqiang Wei: Data curation, Writing – review & editing. Haoran Yuan: Resources, Funding acquisition, Writing – review & editing. Yong Chen: Supervision.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This work was supported by the Key-Area Research and Development Program of GuangDong Province [2020B1111380001]，Guangdong Basic and Applied Basic Research Foundation [2021B1515020068].

Waste pulp (WP) is a typical byproduct of paper industry, and Chemical-looping gasification (CLG) as a recently developed technology is highly suited to dispose high-volatile wastes like WP. In order to make a high-efficiency oxygen carrier (OC) for CLG of WP, the Ni-containing electroplating sludge (NES) was used as the matrix and NiO modification was performed to enhance the hydrogen production in CLG. These resulted in a potentially high-efficiency OC denoted as NNES. Testing CLG of WP was in a fixed-bed reactor at 850 °C by adopting NNES as the OC, injecting 2.4 mL/g(WP) water, and keeping a mass ratio of 1.0 between OC to WP. It produced 1.73 Nm3/kg syngas that has an LHV of 11.9 MJ/Nm3 and a H2/CO ratio of 3.63. In 10 redox cycles, the syngas yield did not have obvious decrease, but a certain reduction in the activity of NNES was observed. Characterization of the spent NNES revealed that it is the Ni agglomeration and inert silicate generation which reduced the activity of NNES.

Hydrogen (H2), with the merits of high enthalpy, zero-emission utilization and abundant natural resources, has been considered as renewable and clean energy carrier allowing a replacement of conventional fossil fuels [1,2]. Water electrolysis as a promising environment-friendly technology for producing hydrogen, has attracted enormous attention in recent years [3,4]. In water splitting, Hydrogen evolution reaction (HER) heavily depends on state-of-the-art Pt or Pd/Rh based catalysts owing to their lower overpotential and small values of Tafel slope [5,6]. Most notably, the high-cost and low-abundance of precious metals severely limited their wide applications. Up to now, numerous efforts have been made towards development of earth-abundant and low-cost electrocatalysts for high-efficiency and stable hydrogen evolution reaction. In this respect, transition metals (i.e., Ni, Co and Fe) based electrocatalysts, including metal oxides/hydroxides [7–9], chalcogenides [10–13], nitrides [14,15], borides [16,17], phosphides [18–20], and alloys [21,22], have attracted most attention as the substitutes for noble metal catalysts.Among them, transition metal phosphides (TMPs) are among the most promising electrocatalysts as the negatively charged P in TMPs can provide superior activity by capturing positively charged protons in HER process [23,24]. Furthermore, the amorphous materials endow excellent catalytic activity due to their unique structure and abundant unsaturated sites. Recently, Xu and co-workers reported that the fabrication of amorphous phosphides can regulate the P content of phosphides by breaking the limitation of ordered stoichiometric crystal structures [24]. Therefore, it is expected that exploiting amorphous phosphides with the merits of amorphous materials would overcome high catalytic reaction barriers and reach higher activities for HER. The Ni–Fe–P and Co–P materials meet the above criteria and deserve further design and development. Moreover, recent reports demonstrated that the Ni–Fe–P materials exhibit superior catalyst activity owing to its bimetallic system [25,26]. Some reports found that microstructures of Co–P materials determine its catalyst activity [27,28].In an electrocatalytic process, the catalyst activity is strongly related to the following three key steps [29,30]: (i) mass diffusion; (ii) electron transfer; (iii) surface/interface reactions. It has been widely recognized that electronic structure adjustment and surface/interface optimization are the reliable strategies. So, higher catalytic reactivity has been achieved though judiciously engineering electrocatalysts through the use of electronic and structural engineering [31,32], strain engineering [33] and defects engineering [34]. However, most of strategies utilize energy-intensive, time-consuming and complicated preparation processes, which largely hampered their extensive commercial applications. Combining different catalyst interfaces to get more efficient electrocatalysts is a simple and effective strategy [35], such as Ni3S2/Co9S8 [10], Pt-NC/Ni-MOF [36], CoO/CoP [37].The electroless plating is a controllable autocatalytic chemical reduction process for metal deposition. In the process of electroless plating, non-metallic atoms (i.e., P, B, S) are co-deposited with metal atoms, resulting in the formation of amorphous alloys. Therefore, transition metals (i.e., Ni, Co, Fe, Mo) with electrocatalytic activity could be deposited easily as electrocatalysts. Moreover, since the alloys prepared by industrialized electroless plating are deposited directly on the substrate without using additional binder, electroless plating is quite suitable for the fabrication of electrocatalytic electrodes suitable applications [38,39].Inspired by these aspects, we constructed layer-by-layer alternately stacked Ni–Fe–P and Co–P films deposited on nickel foam (marked as Co–P/Ni–Fe–P/NF), via facile electroless plating and de-alloying process. Notably, there is a significant interaction between different layers, instead of a simple stacking of the Ni–Fe–P and Co–P films. Using scanning electron microscope (SEM) study, we demonstrated that Ni–Fe–P films induced and promoted the specific growth of following films. Furthermore, charge transfer between metal and P atom of different layers was observed using X-ray photoelectron spectroscopy (XPS) characterization, implying the regulation of interfacial electronic structures. Aiming at these phenomena, in-depth and detailed research has been carried out in this paper. Most importantly, the as-prepared Ni–Fe–P/Co–P/NF electrocatalysts exhibit remarkable HER performance (a low overpotential of 43.4 mV at 10 mA cm−2), as well as a lower Tafel slope of 56.5 mV dec−1 and outstanding durability throughout 72 h in an alkaline medium. The innovative strategy of this work would promote the meticulous design and will create a better understanding of the electrocatalysts.Commercial nickel foam (1 × 3 cm2) was placed in 200 mL beaker containing sodium carbonate anhydrous (3 g), sodium hydroxide (2 g), trisodium phosphate dodecahydrate (0.5 g), and deionized water (100 mL). Boil for 3 min to remove greasy dirt from the surface of nickel foam (NF). Then, after deionized water washing, NF was cleaned thoroughly with 3 mol L−1 HCl 30 min for eliminating surface oxides.The chemical deposition of Co–P alloys on NF was performed in the solution (denoted as solution A) at 90 °C, containing 25 g L−1 cobalt sulfate heptahydrate, 20 g L−1 trisodium citrate dihydrate as complexing agent, 30 g L−1 ammonium fluoride as stabilizer, 40 g L−1 sodium hypophosphite monohydrate as reducing agent. The pH values were regulated to 9 through adding appropriate amount of ammonia. Bubbles were slowly generated around NF during electroless plating. After 30 min of deposition, the Co–P/NF electrode cleaned with deionized water and dried for 8 h at 60 °C.The chemical deposition of Ni–Fe–P alloys on NF was performed in the solution (denoted as solution B) at 90 °C, containing 7.5 g L−1 nickel sulfate, 17.5 g L−1 ammonium ferrous sulfate, 20 g L−1 trisodium citrate dihydrate as complexing agent, 30 g L−1 ammonium fluoride as stabilizer, 40 g L−1 sodium hypophosphite monohydrate as reducing agent. The pH of the electroless plating solution was adjusted and controlled to 9 with ammonia water. Bubbles were slowly generated around NF during electroless plating. After 60 min of deposition, Ni–Fe–P/NF electrodes were taken out from the solution, cleaned with deionized water. Then, the sample was dealloyed in the 5% HCl solution for 30 s at RT, after repeated cleaning dried for 8 h at 60 °C.The preparation process of Co–P/Ni–Fe–P/NF combined two steps mentioned above. Step one, Co–P films were deposited on NF for 10 min in solution A. Step two, Ni–Fe–P films were deposited for 30 min in solution B based on the samples obtained from the previous step; then, the samples were dealloyed in the 5% HCl solution for 30 s at RT. Generally, the multilayered Co–P/Ni–Fe–P/NF electrodes were obtained by cyclic execution of steps one and two, marked as 2 L (2 Layers), 2 L’, 3 L, 4 L, 5 L and so on, wherein the 2 L’ represents the electrodes obtained by changing the order of steps one and two. Finally, the last samples were dried at 60 °C for 8 h. Furthermore, to investigate the effect of dealloying time, different dealloying time (0, 30, 60 and 90 s) was performed for preparing 5 L samples. According to the controllability of deposition time to coating thickness, different deposition time of Co–P (5, 10 and 15 min) and Ni–Fe–P (15, 30 and 45 min) in 5 L samples were performed to investigate the effect of coating thickness, marked as X/Y, wherein the X and Y represent the deposition time of Co–P and Ni–Fe–P, respectively.The electrochemical measurements were performed in a standard three-electrode system on an electrochemical work station with a CHI 760 E (Shanghai Chenhua, China) in 1 mol L−1 KOH solution. For the HER tests, the as-prepared electrodes (1 × 1 cm2), a carbon rod and a Hg/HgO (1 mol L−1 KOH) electrode were used as the working electrode, the counter electrode and the reference electrode, respectively. The linear sweep voltammetry (LSV) was carried out at a sweeping rate 1 mV s−1 with 90% iR-compensation. The LSV was run five times for obtaining more accurate and stable polarization curves. Electrochemical impedance spectroscopy (EIS) was performed to measure the solution resistance (Rs) and charge transfer resistance (Rct). Electrochemical impedance spectroscopy (EIS) was performed at −0.1 V (vs. RHE) in the frequency range from 100 kHz to 0.01 Hz. And the equivalent circuit of the EIS data was fitted with Z-View software. The electrochemical active surface area (ECSA) was evaluated from the electro-chemical double-layer capacitance (Cdl), through collecting cyclic voltammograms (CVs) in the potential range non-faradaic processes at various scan rates from 2 to 10 mV s−1 in the potential range from −0.9 to −0.98 V versus Hg/HgO. The potential was converted to the reversible hydrogen electrode (RHE) according to the Nernst equation: E(RHE) = E(vs. Hg/HgO) + 0.9268 V. Chronopotentiometry (CP, 10 mA cm−2) was performed for 72 h in 1.0 mol L−1 KOH solution to evaluate the long-term stability of the Co–P/Ni–Fe–P/NF electrodes. Multi-current steps test from 20 to 200 mA cm−2 with an increment of 20 mA cm−2 per step for Co–P/Ni–Fe–P/NF electrodes. The chronopotentiometry and multi-current process were carried out without iR-compensation.The fabrication process of Co–P/Ni–Fe–P/NF electrodes with hierarchical structure was illustrated in Fig. 1 , including repeated deposition Co–P and Ni–Fe–P coating and the de-alloying process for Ni–Fe–P alloys. The nickel foam with unique 3D network and tortuous skeleton structure was used as both a current collector and substrate.The phase structures of different samples were studied by X-ray diffraction (XRD) patterns. As shown in Fig. 2 , the inconspicuous diffraction peaks of Co–P/NF can be assigned to pure hexagonal-phase metallic Co (JCPDS card 05–0727) and the Co–P/NF exhibit a broad peak of an amorphous alloy around 25°, which indicate the samples mixed amorphous and nano-crystalline structure features. Moreover, the obvious peaks in XRD patterns of Co–P/Ni–Fe–P/NF and Ni–Fe–P/NF attribute to Ni foam (JCPDS card 04–0850), as the incomplete separation of the coatings and nickel foam. The broad peaks at 25° and 45° of Co–P/Ni–Fe–P/NF and Ni–Fe–P/NF suggest the existence form of amorphous phase of the alloys, indicating that the crystal phase of Co–P alloys transformed into amorphous phase after interface engineering.To detect the structure morphology of different electrodes, scanning electron microscope (SEM) has been employed to explore the details. Cross-sectional SEM images (Fig. 3 a, Fig. S3a and S3c in the Supporting Information), elemental mapping images (Fig. 3b and Fig. S3b), and Line-scan SEM-EDS elemental distribution curves (Fig. S3d) unambiguously showed that Co–P/Ni–Fe–P/NF-5L electrodes have multilayer structure comprised of Co–P and Ni–Fe–P films. There are varied surface morphologies in different layers of Co–P/Ni–Fe–P/NF electrodes. As Fig. 3a and Fig, S1 displayed, the Co–P alloys that first deposited are densely grown on the Ni foam. The Co–P alloy is distributed in blocks on the nickel foam (Fig. 4 a), which is observed from the surface image. The Ni–Fe–P alloys grown on the Co–P film or Ni foam are both uniformly and regularly deposited on the.Substrate (Fig. 3a and Fig. S2). After 30 s de-alloying process, the Ni–Fe–P film surface has distinct nano-micropores structure (Fig. 4b). As revealed in Figs. 3c, 3d and 4c, Co–P alloys in third layer of Co–P/Ni–Fe–P/NF electrodes are nanosheets structure, which is different with the Co–P alloys grown on Ni foam. Moreover, unlike the layer two and Ni–Fe–P/NF, the Ni–Fe–P alloys in fourth layer have cracked structure (Figs. 3e and 4d). The Co–P alloys in layer five show another new structure, which is randomly distributed on the layer four in particles form (Fig. 4e and f). It is because of combining these surfaces with various structures that the active sites are increased tremendously as well as the diffusion and mass transfer are accelerated greatly [40].In order to investigate the reason and mechanism for the above phenomenon, a series of experiments were carried out. In the course of experiment, whether de-alloying treatment for Ni–Fe–P film was performed or not, which had a significant influence on depositing Co–P and the following films. A large number of bubbles were rapid produced in intense reaction when Co–P film was deposited on Ni–Fe–P film that had been de-alloying treated. The experimental steps are as follows: First, one layer Ni–Fe–P film was deposited on the Ni foam. Then, regarding the de-alloying step performed or not as a variable. At last, one layer Co–P film was deposited on Ni–Fe–P film. As shown in Fig. S5, discrete lamellar structure was formed in the Co–P coating sections. When the Ni–Fe–P coatings were not performed de-alloying process, there is no discrete layered structure in Co–P film (Fig. S6). It is observed from the Fig.S6c that the Ni–Fe–P coatings without performing de-alloying process have smoother surface (Fig. 4b) compared to the coatings after de-alloying process. In alkaline solution, the electroless plating including two steps: the oxidation of hypophosphite and the co-deposition of phosphorus with the metals (Ni & Fe or Co) [41], respectively as: H2PO2 - + 2OH- = H2PO3 - + H2O + 2e (1) (2) 4H2PO2 − = 2OH− + H2 + 2P + 2H2PO3 − The electroless plating initiates a metal deposition process by means of auto-catalytically active centers on the surface of the substrate. In combination with the experimental phenomena, the rapid release of bubbles means that step (2) reacts violently. The de-alloying process activated the Ni–Fe–P film, which maybe form abundant reaction sites of auto-catalytically on the surface of the Ni–Fe–P film. And it was also further activated Co–P film. In brief, dispersed growth Co–P film is formed due to the co-deposition of phosphorus with the metals is accompanied by rapid release of bubbles. In addition, the fracture of Ni–Fe–P film in layer four owe to the dispersed structure of Co–P film. As for the various dispersed morphologies of Co–P in different layers and samples (Figs. 3d, 3e, 4c, 4e, and Fig. S5c), which attributes to the concentration of solution A decreases. In summary, the de-alloying process activated the coatings, which endowed the electroless plating process more abundant autocatalytic active center and thus further promoted the dispersive growth of these films.The surface chemical composition of Co–P/Ni–Fe–P/NF-5L was characterized by x-ray photoelectron spectroscopy (XPS). In Fig. 5 a, the fitting peaks of Ni 2p at 852.25 and 869.45 eV coincide with Ni 2p3/2 and Ni 2p1/2 for Ni0 [42,43]. The other two peaks emerged at 855.95 and 873.65 eV should be assigned to Ni 2p3/2 and Ni 2p1/2 according to the spin–orbit characteristics of Ni2+, accompanying with two palpable satellite peaks (denoted as “sat.”) [44,45]. The fitted broad Fe 2p3/2 envelope in Fig. 5b is resolved into two peaks at 712.6 eV and 716.6 eV (sat.), which corresponds to FeOX species when exposed to air [46,47]. It is also observed that a very weak peak with the binding energies of ∼706 eV in Fe 2p3/2 is in agreement with.The characteristics of Fe0 [43], indicating the existence of relatively little metallic Fe in the catalysts. For Co 2p in Fig. 5c, the peaks emerged at 2p1/2 781.15 eV and 2p3/2 796.95 eV originate from high valence state cobalt and metallic cobalt [48–50]. Note that the peak located at 777.60 eV is in agreement with the Co0 signal [51], demonstrating that the presence of metallic cobalt in the catalysts. The metallic characteristics have a series of advantages including superior electron transportation capacity, high hardness and high tensile strength, which is eligible to be efficient and robust electrocatalys. Moreover, compared to red phosphorus (130.0 eV), P 2p has a lower binding energies of 129.60 eV (Fig. 5d), which is assigned to reduced phosphorus in the form of metal phosphides [52]. This demonstrates that electron density transfer from the metallic (mainly Ni and Co) state with positive charge (δ+) to P with negative charge (δ-), which illustrates the P with the electronegativity effect of trapping positively charged protons in electrocatalys [50,52]. Another broad peak located at 133.20 eV is attributed to phosphate species (P5+) and the P species arising from superficial oxidation of the alloys exposed to air [53,54]. In addition, Supplementary Table S1 compares the XPS data of our previous work Co–P/NF, Ni–Fe–P/NF and of this work Co–P/Ni–Fe–P/NF in detail. Note that the Ni 2p XPS region of Co–P/Ni–Fe–P/NF shows slightly negative shift than that of Ni–Fe–P/NF, with a negative shift of 0.15 eV at Ni 2p peaks. Similar with Ni 2p, the Co 2p XPS region of Co–P/Ni–Fe–P/NF is also observed the slightly negative shift of 0.15 eV than that of Co–P/NF at Co 2p peaks. Interestingly, the P 2p XPS region of Co–P/Ni–Fe–P/NF exactly shows positive shift of about 0.30 eV than that of Ni–Fe–P/NF and Co–P/NF. This indicates that the electron transfer from P to Co and Ni in the multilayer films. The transfer of electrons proves that the electronic structure of the layered materials is tuned when the interface is constructed [55]. Research shows that there are strong electronic interactions between deposited Co–P alloys and Ni–Fe–P alloys, after performing interface engineering, which created ample metal active sites and provided more electron transfer access to promote the electrocatalytic in the alloys.To evaluate HER activity, the Co–P/Ni–Fe–P/NF electrodes were subjected as an cathode to three electrode chemical cell in 1 mol L−1 KOH electrolyte. For comparison, Ni–Fe–P/NF, Co–P/NF, bare NF and 20% Pt/C coated on NF (Pt/C on NF, 20% Pt/C loading: 5 mg cm−2) were also measured at the same conditions. As expected, the Co–P/Ni–Fe–P/NF electrodes after multilayer interface engineering exhibit prominent catalytic activity for HER with a lower overpotential of 43.4 mV at 10 mA cm−2 (Fig. 6 a). Even when compared with the benchmarking Pt/C, the Co–P/Ni–Fe–P/NF only requires an additional 26 mV to reach 10 mA cm−2. It obviously outperforms the corresponding single-layer electrodes of Co–P/NF (η10 = 177.5 mV) and Ni–Fe–P/NF (η10 = 106.9 mV). Even at high current density, the Co–P/Ni–Fe–P/NF electrodes still keep comparatively outstanding performance (Fig. 6b). Moreover, according the loading mass table in Fig. 6a, the Co–P/Ni–Fe–P catalyst displays a high mass activity of 7.311 mA mg−1 (Fig. S7) at an overpotential of 200 mV by normalization with the loading mass, which is significantly higher than the catalysts of Co–P (0.395 mA mg−1) and Ni–Fe–P (2.029 mA mg−1). Calculation details of mass activity (mA g−1) are seen in the Supporting Information. This indicates that the superior performance of multilayer Co–P/Ni–Fe–P/NF is originated from the rational regulation of coating interface by interface engineering and the synergistic effect of Co–P and Ni–Fe–P alloys, instead of the simple increaseing of loading mass. To further investigate the effect of layer number, dealloying time, and coating thickness, the different Co–P/Ni–Fe–P/NF samples were also fabricated by changing deposition layers number, dealloying time and deposition time, as shown in Figs. S9, S10 and S11. It is observed from the Fig. S9a that there is an increasing trend towards HER activity for all the multilayer Co–P/Ni–Fe–P/NF samples with increasing layers number till five layers, demonstrating that the superposition of coating layer could accelerate the hydrogen evolution of electrocatalyst. Furthermore, the de-alloying process has immense impact on the HER catalytic activity. The Co–P/Ni–Fe–P/NF sample without performing de-alloying process shows poor performance, in that the unactivated Ni–Fe–P and Co–P alloys can't form the favorable nanostructures. However, the prolonged de-alloying treatment led to severe corrosive for the alloys in acidic media, and thus the metal atom ratio and the stable structure were destroyed. Consequently, the Co–P/Ni–Fe–P/NF sample fabricated at 30 s de-alloying time displays excellent electrocatalytic activity than at de-alloying time (Fig. S10a). Similarly, the appropriate deposition thickness of Co–P films and Ni–Fe–P films could sufficiently exert electrocatalytic effect. The result reveals that the sample shows an optimal HER catalytic activity when the deposition time of Co–P and Ni–Fe–P are 10 min and 30 min, respectively (Fig. S11a). As observed in SEM image (Fig. 3a), the thickness of Co–P films and Ni–Fe–P films are about 1 μm and 2 μm, respectively. These results demonstrate that more layers of deposition, suitable dealloying time and appropriate deposition thickness would enhance the HER activity.The evaluation of the amount of active sites for HER obtained through calculating the electrochemical active surface area (ECSA). There is a positively correlation between ECSA and the electrochemical double-layer capacitance (Cdl), according to the equation ECSA = Cdl/Cs where Cs is the specific capacitance with a fixed value under identical electrolyte conditions [56]. And the electrochemical double-layer capacitance (Cdl) value is measured via cyclic voltammetry (CV) test (Fig. S8). Evidently, the multilayer Co–P/Ni–Fe–P/NF shows a higher Cdl (445.4 mF cm−2) than Co–P/NF (17.7 mF cm−2) and Ni–Fe–P/NF (390.2 mF cm−2), suggesting more catalytically active sites on the multilayered alloys. Notably, combining with the XRD patterns (Fig. 2), it is apparent that the Ni–Fe–P/NF and Co–P/Ni–Fe–P/NF with amorphous phase have higher Cdl value than Co–P/NF with crystal phase. This result indicates that the enhanced performance of Co–P/Ni–Fe–P/NF more likely resulted from the phase transformation of Co–P alloys, meaning that amorphous electrocatalysts could provide more active sites in HER process, which is consist with reported previously [21,57].To figure out the reaction kinetics, linear fitting Tafel plots were further conducted to further characterize the electrocatalysts. As widely accepted, the various values of Tafel slopes correspond to the different determining rate steps, including Volmer, Heyrovsky and Tafel steps. In this work, the multilayer Co–P/Ni–Fe–P/NF sample has a low Tafel slope of 56.5 mV dec−1 (Fig. 6c) which is lower than that of Co–P/NF (83.6 mV dec−1) and Ni–Fe–P/NF (78.3 mV dec−1), indicating that the Heyrovsky step is the mainly rate-determining step [58]. Furthermore, the Co–P/Ni–Fe–P/NF-3L has a small Tafel slope of 54.4 mV dec−1 (Fig. S9b), which is even smaller than that of Co–P/Ni–Fe–P/NF-5L (56.5 mV dec−1), revealing the nanosheets Co–P films in layer three endow the electrochemical HER with favorable kinetics. In addition, the Tafel slope of Co–P/Ni–Fe–P/NF samples fabricated at de-alloying time of 0 s and 30 s have significantly difference, confirming that nanostructures alloys formed after activation have rapid HER reaction kinetics (Fig. S10b).The electrode–electrolyte interfaces reaction kinetics were studied by electrochemical impedance spectroscopy (EIS) analysis. The charge transfer resistance (Rct) of the interface between the electrolyte and the catalysts is usually considered to be determined by the semicircle diameter of the Nyquist plots [45]. And the Rct is closely related to the reaction situation between the electrolyte and the catalysts. Fig. 6d exhibits the Nyquist plots of Co–P/NF, Ni–Fe–P/NF and Co–P/Ni–Fe–P/NF, implying a faster electron transfer for the multilayer Co–P/Ni–Fe–P/NF. Besides, the values of the Rct and the solution resistance (Rs) extracted from Nyquist plots decrease with the increasing of the layer number of films (Fig. S9d), indicating that the interface engineering effectively promoted the charge transfer efficiency for HER. In addition, the as-prepared samples after de-alloying process exhibit small impedance (Fig. S10c), which reveals that the favorable structures of the multilayer alloys after activation provided abundant and efficient transfer pathways for electrons and ions in the catalysts.Finally, in order to further evaluate the long-term stability of the Co–P/Ni–Fe–P/NF electrode, the Co–P/Ni–Fe–P/NF was tested in 1 mol L−1 KOH solution employing a three-electrode system. There was no obvious increasing for overpotential of Co–P/Ni–Fe–P/NF electrode at the current density of 10 mV cm−2 after running at least 72 h (Fig. 7 a). Moreover, multi-current steps test from 20 to 200 mA cm−2 with an increment of 20 mA cm−2 per step was performed to assess the electrocatalytic durability of the Co–P/Ni–Fe–P/NF electrode. The potential immediately reach a stable state at the initial current of 20 mA cm−2 and keep stable for 1000s, and continue to remain stable at the later steps (Fig. 7b). In addition, after long-term stability test, the XRD pattern (Fig. S12a), SEM images (Fig. S12b) and elemental mapping (Fig. S12c) of the post-HER Co–P/Ni–Fe–P/NF sample have no obvious change compared with fresh samples. These results reveal that the Co–P/Ni–Fe–P/NF electrode has superior long-term stability, mechanical robustness, conductivity and mass transportation property [50,59], which is promising and available electrocatalysts for HER.Combining with the above research results, the reasons for the excellent performance of Co–P/Ni–Fe–P/NF electrode are explained from the following three points: (i) source of activity, (ii) hydrogen evolution reaction kinetics, and (iii) electronic transfer and mass transportation capability. Based on the results of the SEM images (Figs. 3, 4 and Fig. S4), LSV curves (Fig. 6a, S9a, S10a and S11a), mass activity data (Fig. S7) and XPS analysis, after rational design and preparation processes, the Co–P/Ni–Fe–P/NF electrode with favorable multilayer structures and prominent electrocatalytic performance for HER was successfully fabricated via the interface engineering. The Co–P/Ni–Fe–P/NF electrode with best hydrogen evolution performance has five layers films alternately stacked by Co–P and Ni–Fe–P films, and has a lower overpotential of 43 mV at 10 mA cm−2. Compared with recent advanced electrocatalysts for HER in 1 mol L−1 KOH solution, the Co–P/Ni–Fe–P/NF still maintains a higher advantage (shown in Table S2). These results powerfully certified the success of multilayer strategy in this study. (i) Source of activity. According to the previous research, Yang et al. believed that the crystalline materials, like Co–P alloys in our work, have limited active sites existing on the edges or surface of crystalline materials. As for amorphous materials, like Ni–Fe–P and Co–P/Ni–Fe–P alloys in our work, their short-range ordered and long-range disordered structures provide rich defect sites to act as active centers in the HER process. Therefore, the whole amorphous electrocatalysts could adequately provide active sites to boost the interfacial catalytic reactions [21,60]. In addition, the P element has important function in electrocatalysts for HER. Liu and Rodriguez [61] discovered that the Ni2P-(001) behaves somewhat like the hydrogenase, and the P atoms with electronegativity attract the electrons from metal atoms. In HER process, the negatively charged P atoms in phosphides can trap proton to enhance the catalytic efficiency. In this work, according to the results of XPS, there are similar situation in Co–P/Ni–Fe–P/NF. Moreover, the amorphous structure of Co–P/Ni–Fe–P/NF could break through the limitation of crystal lattice and thus facilitate the catalytic reaction. Our work proves the superiority of amorphous materials in the field of electrocatalysis once again. (ii) Hydrogen evolution reaction kinetics. In alkaline electrolytes, the HER proceed with the following steps: Water dissociation and H∗ (adsorbed hydrogen) generation step (120 mV dec−1) H2O + e- → H∗ + OH- (Volmer step) (1) Source of activity. According to the previous research, Yang et al. believed that the crystalline materials, like Co–P alloys in our work, have limited active sites existing on the edges or surface of crystalline materials. As for amorphous materials, like Ni–Fe–P and Co–P/Ni–Fe–P alloys in our work, their short-range ordered and long-range disordered structures provide rich defect sites to act as active centers in the HER process. Therefore, the whole amorphous electrocatalysts could adequately provide active sites to boost the interfacial catalytic reactions [21,60]. In addition, the P element has important function in electrocatalysts for HER. Liu and Rodriguez [61] discovered that the Ni2P-(001) behaves somewhat like the hydrogenase, and the P atoms with electronegativity attract the electrons from metal atoms. In HER process, the negatively charged P atoms in phosphides can trap proton to enhance the catalytic efficiency. In this work, according to the results of XPS, there are similar situation in Co–P/Ni–Fe–P/NF. Moreover, the amorphous structure of Co–P/Ni–Fe–P/NF could break through the limitation of crystal lattice and thus facilitate the catalytic reaction. Our work proves the superiority of amorphous materials in the field of electrocatalysis once again.Hydrogen evolution reaction kinetics. In alkaline electrolytes, the HER proceed with the following steps: Water dissociation and H∗ (adsorbed hydrogen) generation step (120 mV dec−1)Electrochemical desorption step (40 mV dec−1) H∗ + H2O + e- → H2 + OH- (Heyrovsky step) (2) Chemical desorption step (30 mV dec−1). 2H∗ → H2 (tafel step) (3) The HER process in alkaline solution follows the Volmer–Heyrovsky process or the Volmer–Tafel process, depending on the Tafel slope given in above equations [62]. In this work, as revealed in Tafel plots (Fig. 6c), the Tafel slope value of Co–P/NF, Ni–Fe–P/NF and Co–P/Ni–Fe–P/NF are 83.6 mV dec−1, 78.3 mV dec−1 and 56.5 mV dec−1, respectively. Accordingly, the Volmer–Heyrovsky process determined HER process on the Co–P/Ni–Fe–P/NF. Moreover, different from the HER in acidic electrolytes, H∗ generation in alkaline medium requires to break covalent H–O–H bond, which requires extra energy [63]. According to the changed Tafel slope (Fig. 6c) after interface engineering, this reveals that the water dissociation and H∗ generation kinetics are significantly improved. The XPS analysis (shown in Table S1) after interface engineering reveals the amelioration of electronic structures of Co–P/Ni–Fe–P/NF, which probably decreased Gibbs free energy of intermediates generation and benefited the electrochemical desorption step. (iii) Electronic transfer and mass transportation capability. On the one hand, the catalysts were directly deposited on the nickel foam with 3D porous structure via electroless plating and didn't require extra binder, which is conducive to sufficient contact between the substrate and the catalysts. And the intrinsic metallic properties of alloys endow the electrocatalysts with higher electrical conductivity. On the other hand, multilayer strategy effectively reduced the charge transfer resistance and the solution resistance on electrode (Fig. S9d), demonstrating the efficient interface reaction between solution and electrocatalysts. Furthermore, various morphologies alloys (shown in Figs. 3, 4 and Fig. S3) in Co–P/Ni–Fe–P/NF with hierarchical and highly interconnected structure are not only quite beneficial for the diffusion of ions, but also meet the requirements of different reaction situation in solution. Electronic transfer and mass transportation capability. On the one hand, the catalysts were directly deposited on the nickel foam with 3D porous structure via electroless plating and didn't require extra binder, which is conducive to sufficient contact between the substrate and the catalysts. And the intrinsic metallic properties of alloys endow the electrocatalysts with higher electrical conductivity. On the other hand, multilayer strategy effectively reduced the charge transfer resistance and the solution resistance on electrode (Fig. S9d), demonstrating the efficient interface reaction between solution and electrocatalysts. Furthermore, various morphologies alloys (shown in Figs. 3, 4 and Fig. S3) in Co–P/Ni–Fe–P/NF with hierarchical and highly interconnected structure are not only quite beneficial for the diffusion of ions, but also meet the requirements of different reaction situation in solution.To summarize, we have successfully constructed the interface engineering based on the Ni–Fe–P and Co–P films by utilizing the characteristics of autocatalytic reaction in electroless plating process. The de-alloying process activated the Ni–Fe–P films to induce subsequent films dispersion growth, which formed various morphologies alloys. Notably, the as-prepared Co–P/Ni–Fe–P/NF electrode shows impressive catalytic performance and long-term durability in 1 mol L−1 KOM solution, which only requires an ultralow overpotential of 43.4 mV and 113.6 mV at the current density of 10 mA cm−2 and 100 mA cm−2, respectively. And the Co–P/Ni–Fe–P/NF electrode also has a low Tafel slope of 56.5 mV dec−1. Such superior catalytic activity should attribute to: (1) the amorphous Co–P/Ni–Fe–P alloys with unique chemical structure supplied rich active sites. (2) The interface engineering improved the kinetics of interfacial reaction. (3) Multilayer strategy effectively reduced the charge transfer resistance and the solution resistance. (4) The various morphologies alloys enhanced the ion transportation capability. (5) The interface coupling of layered alloys led to the improvement of electronic structure. The innovative strategy of this research may play a guidance role in design and development of catalysts for HER.There are no conflicts to declare.This work was financially supported by the Taishan scholar foundation of Shandong (ts201712046) and the National Natural Science Foundation of China (Grant No. 51672145).The following is the Supplementary data to this article: Supplementary data to this article can be found online at https://doi.org/10.1016/j.gee.2020.07.023.

Judiciously engineering the electrocatalysts is attractive and challenging to exploit materials with high electrocatalytic performance for hydrogen evolution reaction. Herein, we successfully perform the interface engineering by alternately depositing Co–P and Ni–Fe–P films on nickel foam, via facile electroless plating and de-alloying process. This work shows that there is a significant effect of de-alloying process on alloy growth. The electronic structure of layered alloys is improved by interface engineering. The multilayer strategy significantly promotes the charge transfer. Importantly, the Co–P/Ni–Fe–P/NF electrode fabricated by interface engineering exhibits excellent electrocatalytic hydrogen evolution activity with an overpotential of 43.4 mV at 10 mA cm−2 and long-term durability for 72 h in alkaline medium (1 mol L−1 KOH). The innovative strategy of this work may aid further development of commercial electrocatalysts.

An imminent decarbonization of the energy sector is needed in order to reduce the environmental damage caused by fossil fuel. As a result, clean, sustainable and renewable energy sources appear to be potential alternatives. In fact, the development of hydrogen-based technologies can help to reduce or eliminate greenhouse gas emission. In this sense, hydrogen (H2) possesses the highest specific energy content among all conventional fuels and, it can be used as a green energy carrier using fuel cells and internal combustion engines by releasing only nontoxic by-products such as water [1]. However, the main drawback related to this compound is its low volumetric energy density which increases storage and transport costs [2]. An alternative to remove these issues is the use of hydrogen carrier compounds.In this respect, liquid fuels generated from hydrogen (ammonia, methanol, metal amine salts, etc.) might be easily stored and transported to be in situ decomposed to produce clean hydrogen through suitable conversion processes [3,4]. Ammonia (NH3) is a promising hydrogen carrier because of its high volumetric energy density and high hydrogen content, well-known technology for production and distribution and relatively low cost [5]. Moreover, its decomposition only produces hydrogen and nitrogen. Therefore, ammonia is an exceptional carbon-free hydrogen vector. However, fuel cells which are very sensitive to ammonia concentration (<1 ppm), demand high purity hydrogen [6]. Thus, almost complete ammonia conversion is required at relatively low temperature (<500 °C). For that purpose, hydrogen production from ammonia decomposition requires efficient and low-cost catalysts to reduce the reaction temperature and the energy cost of the process.Promising results of ammonia decomposition at low temperatures are achieved with ruthenium (Ru) catalysts [6–8]. Nevertheless, catalytic activity is very affected by other factors such as the size of metal particles since it is a structure-sensitive reaction [7,8]. Consequently, hydrogen production from ammonia has been widely studied using different types of active phases (Ni, Co, Rh, Pd, Pt, etc.) and supports such as Al2O3, SiC, ZrO2, CeO2, La2O3, MgO and carbonaceous materials [6–13]. Moreover, different types of promoters have been investigated to enhance the catalytic activity by adding alkaline species to the active phase [11–14]. In fact, these promoters increase the electron-donation to the active metals and stabilize the binding energy between metal and N atoms, favouring ammonia decomposition reaction.In recent years, a novel route for adding promoters to heterogeneous catalyst has been developed through the phenomenon of Electrochemical Promotion of Catalysis (EPOC). This phenomenon discovered by Stoukides and Vayenas in 1981 [15] is a promising alternative way to explore the in-situ addition of electronic promoters to a heterogeneous catalyst and hence, to enhance catalytic reaction rates [16–18]. This phenomenon is based on the electrochemical supply of promoter ions from a solid electrolyte material (support) to a metal catalyst (working electrode) by the application of electrical currents or potentials. This electrochemical activation of the catalyst by using a solid electrolyte support allows the in-situ electrochemical addition or removal of a wide variety of promoters, anionic (O2−) or cationic (Na+, K+, H+) to different kinds of catalysts in a wide range of catalytic reactions [16,17,19–22].The EPOC phenomenon has been widely investigated for different kinds of hydrogen production reactions, e.g., catalytic reforming of methane [20], water gas shift reaction [23–25] and reforming or partial oxidation of alcohols such as methanol and ethanol [16,26,27]. However, one can find in literature a unique previous study of EPOC in the catalytic decomposition of ammonia [17]. In this previous study an iron catalyst film deposited on both K2YZr(PO4)3 (K+ ionic conductor) and, CaZr0.9In0.1O3-α (H+ ionic conductor material), were electrochemically activated. Although very promising results were obtained, a high temperature range (500–600 °C) was explored probably due to the requirements for ionic conductivity of the solid electrolyte used.In this work it has been explored for the first time in the literature, the effect of the electrochemical promotion for low temperature catalytic decomposition of ammonia (250–350 °C). For that purpose, a ruthenium catalyst and an alkaline solid electrolyte (Na-βAl2O3 and K-βAl2O3) have been used on the catalytic reaction. Very promising results have been obtained in a lower temperature range (250–350 °C) which have been discussed in terms of the EPOC rules and the mechanism of ammonia decomposition reaction. Hence, relevant findings are reported of great interest for the general catalysis field which could serve for the design of future novel catalyst formulations.Na-βAl2O3 and K-βAl2O3 (20 mm diameter and 1 mm thickness from Ionotec company) solid electrolyte discs were used as supports. First, thin coatings of gold paste (Gwent Electronic Materials) were deposited on the one side of the solid electrolyte disk as Au counter (CE) and reference (RE) electrodes, followed by calcination steps at 300 °C for 1 h (5 °C·min−1) and 800 °C for 2 h (5 °C·min−1). Blank experiments demonstrated the catalytically inactive properties of the prepared gold counter and reference electrodes for ammonia catalytic decomposition reaction. Then, an electrically continuous ruthenium catalyst film-working electrode (WE) (geometric area of 2.01 cm2) was deposited on the other side of the disk as schematically shown in Figure S1 of the supporting information. An impregnation method described in detail elsewhere was used [28] by using a RuCl3 solution. The precursor salt, RuCl3 ·3H2O (Sigma Aldrich) was dissolved in 1:1 (volume) water: 2-propanol (Sigma Aldrich, 99.9% purity) solution, followed by a calcination step at 500 °C for 1 h (5 °C·min−1). Both obtained electrochemical catalysts (Ru/Na-βAl2O3 and Ru/K-βAl2O3) showed a similar final metal loading of 1.3 mgRu ·cm−2. Before the catalytic activity measurements, the metal catalyst film was reduced under 5 v/v% H2/He gas mixture (100 mL·min−1) at 400 °C for 1 h (10 °C·min−1), in order to ensure the complete reduction of the active ruthenium particles, achieving a metal catalyst film with an in plane electrical resistance around 30  Ω.For the ex-situ characterization of the ruthenium catalyst film, X-Ray Diffraction (XRD) and Scanning Electron Microscopy (SEM) techniques were used. XRD diffractograms were obtained on a Philips X'Pert MPD with co-filtered Cu-Ka radiation (λ=1.54056 Å), after reduction under 5 v/v% H2/Ar and for used catalyst. The XRD pattern were recorded from 20<2θ<80° with a scan rate of 0.02° step size and acquisition time of 4 s per step. The crystal size was determined by the Debye-Scherrer Eq. (1): (1) d = K · λ β · cos θ where d is the average particles size (nm), assuming particles are spherical, K = 0.9, λ=1.54056 Å, β is the full width at half the diffracted peak and θ is the Bragg angle.SEM images of the ruthenium films were performed by ZEISS GeminiSEM 500 FE-SEM with a PIN-diode BSE detector. This instrument was equipped with an energy-dispersive X-ray spectroscopy (EDX) analyzer to verify the composition of the samples.Catalytic tests were carried out in a single chamber solid electrolyte cell reactor configuration (Figure S2 of the supporting information). The electrochemical catalyst was suspended in the reactor by using gold wires (Alfa Aesar, 99.95% purity) which also allow the electrical connections of the three electrodes (Counter (CE), Reference (RE) and Working electrode (WE)) with the potentiostat-galvanostat (Voltalab PGZ 301, Radiometer Analytical). The EPOC phenomenon was investigated by applying different electrical potentials between the WE and CE and measured between the WE and RE electrodes (VWR), according to the technique generally used in conventional three-electrode electrochemical cells [29]. Reaction gasses (Air Liquide) were certified standards of helium (99.999% purity) and ammonia (5011.50 ppm) and the gas flows were controlled by a set of calibrated mass flow meters (Brooks 5850 E).The ammonia decomposition reaction was carried out at atmospheric pressure operating at a temperature range from 220 to 350 °C, using an inlet composition of 1250 ppm ammonia with an overall gas flow rate of 200 mL·min−1 (He balance). Gas effluents were analysed on-line with a dispersive IR Rosemount X-STREAM Enhanced XEGP continuous gas analyzer (EMERSON) for ammonia detection. The hydrogen formation rate (mmol H2 ·min−1 ·gcat −1) was calculated from balance of the ammonia content in the outgas stream, while conversion of ammonia ( X N H 3 ) was calculated as follows: (2) X N H 3 ( % ) = F N H 3 in − F N H 3 out F N H 3 in . 100 Where, F N H 3 in and F N H 3 out referred to the inlet and outlet ammonia molar flows (mmol gas·min−1), respectively. Furthermore, the apparent activation energy of the synthesized catalysts was calculated from the Arrhenius plot at low conversion values (<10%) in order to operate into differential conditions.The electrochemical catalyst was also in-situ characterized in the single chamber cell reactor by cyclic voltammetry experiments. A cyclic voltammetry experiment with simultaneous ammonia recording was performed under ammonia catalytic decomposition reaction (1250 ppm ammonia and a gas flow of 200 mL·min−1) at 280 °C and a scan rate of 5 mV·s− 1.Firstly, the catalytic activity of the ruthenium catalyst film was tested from 200 to 450 °C in the ammonia decomposition reaction under open circuit conditions (O.C.) for Ru/Na-βAl2O3 electrocatalyst (Figure S3 of the supporting information). This initial experiment demonstrated that the prepared metal catalyst film was active for the ammonia catalytic decomposition reaction under the explored reaction conditions. It shows a typical trend of ammonia conversion increasing with the reaction temperature due to the increase in the reaction kinetics [3]. However, it can be observed that full conversion of ammonia was never reached (maximum conversion of 74% at 450 °C) due to certain bypass of the gas reactants flow to the ruthenium catalyst film. This behavior is typically observed with those kinds of solid electrolyte cells reactors and have been reported in previous studies [30].The crystalline structure of the ruthenium catalyst film in the Ru/Na-βAl2O3 electrocatalyst after hydrogen reduction and after ammonia decomposition reaction was examined by XRD (Fig. 1 ). The reduced electrocatalyst showed the main diffraction peaks corresponding to hexagonal phase of Na-βAl2O3 (JCPDS: 19–1173) associated with the electrolyte support [31] and the hexagonal phase of metallic ruthenium (JCPDS: 06–0663) [7]. After catalytic activity measurements, the post-reaction ruthenium catalyst film showed the same crystalline phases than the reduced sample (metallic ruthenium and Na-βAl2O3) not showing crystalline phases derived from the EPOC experiments. Using the Scherrer equation at peak 2θ=44°, the average particles size of ruthenium for both electrocatalyst, reduced and post reaction, was 15.7 and 21.4 nm, respectively. These results demonstrated certain sintering of the metal particles in the catalyst film during the catalytic activity measurements. This kind of sintering of metal particles are typical for conductive catalyst films and have been reported in previous studies of EPOC [28]. Fig. 2 shows SEM micrographs and EDX analysis of the metal film on reduced catalyst (a, b and c) and after ammonia decomposition reaction (d, e and f). Typically, the ruthenium film after reduction treatment was porous and continuous, which suggested that the preparation technique was adequate to synthesize ruthenium films for EPOC experiments [32]. It was observed from the EDX of the reduced sample (Fig. 2b), large concentrations of sodium (in blue) which were attributed to the thermal migration of alkaline ions from the solid electrolyte to the ruthenium catalyst film during the catalyst preparation procedure and reduction step [16,21]. From the SEM analysis of the post reaction sample, in agreement with the XRD, a sintering of the metal particles in the catalyst film could be observed. Furthermore, some cracks in the ruthenium catalyst film after catalysis were observed at microscopy level. However, these cracks did not affect the electrical conductivity of the ruthenium catalyst film as verified along the surface with a multimeter. In any case, this sintering process of the metal particles shown in the sample after catalytic reaction probably occurred at the beginning of the experimental tests (initial experiment under open circuit conditions shown in Figure S3), since reproducible and stable catalytic activity values were obtained during EPOC experiments (as will be shown below). Fig. 3 shows the variation of the hydrogen production rate vs. time at different applied potentials between 2 and −2 V at 300 °C. Each polarization was applied for 30 min in order to achieve a steady state catalytic behavior.In agreement with previous EPOC studies and considering that the initial polarization at VWR= 2 V allowed to obtain a metal catalyst film free of any sodium ions (reference state, i.e., un-promoted catalytic activity), the subsequent decrease in the applied potential led to the electrochemical supply of sodium ions to the catalyst surface. It was confirmed by the obtained measured negative electrical currents (not shown here) of higher magnitude as the applied potential decreased to more negative values, corresponding to a higher electrochemical supply of electropositive alkali sodium ions. This direct correlation has been recently confirmed by an Operando Near Ambient Pressure Photoemission Spectroscopy performed on Ni/K-βAl2O3 in a previous work of the research team [33]. Concerning the variation of the catalytic activity with the different applied polarizations, according to previous studies [22,24,25,34–37], one may rationalize the observed variation in the catalytic activity on the basis of the catalyst-electrode work function modification upon applied potential. A decrease in the catalyst potential below the open circuit conditions, and therefore, in the catalyst work function, led to an increase of the electronic density of ruthenium catalyst with a concomitant spillover of Na+ ions from the electrolyte onto the metal catalyst surface. This decreased work function, weakened the ruthenium chemical bond with electron-donor adsorbates and strengthened those with electron acceptors ones. In this work, the strong promotional effect, observed when lowering the catalyst potential below the open circuit conditions (50 mV), can be explained on the basis of a strengthening of the chemisorptive bond of weakly adsorbed N surface species, which stabilizes N on the catalyst surface. It facilitates the ammonia decomposition reaction, in good agreement with a previous EPOC study in the catalytic ammonia decomposition reaction on iron deposited on K2YZr(PO4)3, a K+ conductor solid electrolyte [17]. This experimental observation is also in good agreement with the mechanism proposed in the literature for the catalytic ammonia decomposition reaction, which follows these six consecutive steps [38]: (3) Step 1 : N H 3 ( g ) ↔ N H 3 ( a ) (4) Step 2 : N H 3 ( a ) + s ↔ N H 2 ( a ) + H ( a ) (5) Step 3 : N H 2 ( a ) + s ↔ NH ( a ) + H ( a ) (6) Step 4 : NH ( a ) + s ↔ N ( a ) + H ( a ) (7) Step 5 : 2 N ( a ) → N 2 ( g ) + 2 s (8) Step 6 : 2 H ( a ) ↔ H 2 ( g ) + 2 s where s represents a vacant site of the catalyst surface, (g) stands for gas and (a) stands for adsorbed molecules. Hence, a previous detailed kinetics analysis [8] has demonstrated that for the case of a ruthenium based catalyst under similar reaction conditions than our work, step 2 is the main rate determining step in the overall reaction mechanism. Then, the stabilization of NH2(a) adsorbed species on the catalyst surface when the potential is decreased, might increase the kinetics of the dehydrogenation reactions (steps 2–4), leading to an overall activation of the process. On the other hand, the increase in the value of the applied potential above that of the open circuit conditions (50 mV), enhanced the binding strength of the electron donor ammonia molecules (facilitating the ammonia adsorption, step 1 of the mechanism) [17]. It led to a slight increase in the catalytic activity, as can be observed under application of 2 V in Fig. 3, but much less important considering that a higher effect is produced when the electronic effect increases the kinetics of the rate determining step. Finally, it can also be observed that the final application of a potential of 2 V allowed to recover the initial un-promoted catalytic activity achieved at the beginning of the experiment, leading to a reversible EPOC phenomenon [16,25]. It implied that the same amount of sodium ions, initially transferred from the solid electrolyte to the catalyst during the previous polarizations, were returned from the catalyst to the solid electrolyte after the final application of 2 V. It also demonstrated the stability of the catalyst film under EPOC reaction conditions. Therefore, the sintering of the metal particles observed in Figs. 1 and 2 might occurred during the preliminary reaction experiment under O.C. (Figure S3), stabilizing the ruthenium catalyst film for the subsequent EPOC experiments.In order to study the influence of the reaction temperature, the same experiment performed at 300 °C was repeated at other reaction temperatures, i.e. 250 °C, 320 °C and, 350 °C. The steady state variation of the ammonia conversion vs. the applied polarization at the different reaction temperatures is shown in Fig. 4 . Also, to quantify the magnitude of the EPOC phenomenon, Fig. 4 shows the maximum value calculated of the rate enhancement ratio (ρ), obtained by the following equation and typically used in EPOC studies [24,39]: (9) ρ = r r 0 where r0 is the un-promoted catalytic reaction rate (VWR=2 V) and r is the promoted catalytic reaction rate at the explored potential.Firstly, it could be observed a similar overall EPOC behavior, at the explored reaction temperatures, to the one already described in Fig. 3, with an increase in the catalytic reaction rate as the applied potential decreases from the open circuit potential values (typically around 50 mV). As typically found in previous EPOC studies with alkaline ion conductors [40], the highest value of the rate enhancement ratio was achieved at an intermediate reaction temperature, i.e. 300 °C, increasing the ammonia conversion 1.4 times vs. the un-promoted conditions (2 V). Hence, at low temperatures (i.e. 250 °C) the EPOC phenomenon is limited by the low ionic conductivity of the solid electrolyte which decreases the amount of electrochemically supplied ions. However, above certain reaction temperatures (e.g. 320 °C), the relative increase induced by the EPOC phenomenon, measured by ρ, is limited by the initial higher value of the un-promoted catalytic activity (i.e. 30% at 320 °C). On the other hand, it is also interesting to note that, a poisoning effect can be observed at higher temperatures (above 320 °C), which led to an optimal value of the applied potential which maximized the catalytic activity at VWR= −1 V at 320 °C and VWR= −0,5 V at 350 °C. This behavior typically found in previous studies of alkaline EPOC can be attributed to an excess of promoting species on the catalyst surface which block the catalytic active sites [41]. In the present study, as already mentioned, the application of negative potentials decreased the relative coverage of electron donor molecules (ammonia) (step 1), causing a poisoning effect observed above certain temperatures and below certain applied potentials. Hence, above certain temperatures and above certain alkali coverage on the catalyst, the mechanism could be limited by the ammonia adsorption step, which may also justify the higher inhibiting effect observed at 350 °C vs. 320 °C at high negative potential (close to −2 V). On the other hand, it could be observed that higher optimal applied potential values (which maximized the catalytic activity) were obtained at higher reaction temperatures. It can be attributed to an increase in the solid electrolyte ionic conductivity with temperature leading to a higher supply of sodium ions for the same potential. Obtained results clearly demonstrated the interest of the EPOC phenomenon to electrochemically supply the optimal promoter amount at different reaction conditions (e.g. temperature), which is not possible with conventional heterogeneous catalyst doped with alkali, where a fix amount of promoter is added to the catalyst during the preparation step. This is one of the most interesting findings of EPOC which has been analysed in detail in a previous reviewed manuscript [42].In order to evaluate the electro-promotional effect of other type of alkali ions, the EPOC phenomenon was investigated on a Ru/K-βAl2O3 electrochemical catalyst. Fig. 5 shows the variation of the hydrogen production rate vs. time at different applied potentials (between VWR= 2 V and −2 V) and reaction temperatures (300 and 320 °C). Each polarization was again applied for 30 min until a steady state catalytic behavior is achieved.In good agreement with the results obtained with Ru/Na-βAl2O3, a strong activation effect in the catalytic activity was observed under conditions of electrochemically supply of alkali ions (K+). The observed promotional effect can be explained considering the influence of the alkali ions on the binding strength of chemisorbed reactants and intermediate molecules, as previously discussed for Ru/Na-βAl2O3. The electrochemical supply of potassium ions enhanced the chemisorption of electron acceptor molecules (weakly adsorbed N surface species) facilitating ammonia decomposition reactions, increasing the rate of the three consecutive steps 2, 3 and 4, on the previously mentioned reaction mechanism. In this case, it can be observed that an optimum applied potential of VWR= 0 V and VWR=0.5 V was obtained at 300 °C and 320 °C, respectively. These optimal VWR values were higher than the ones obtained for Ru/Na-βAl2O3, which can be justified considering the higher electronic effect on the catalyst of K+vs. Na+ions, attributed to the different ionic size of both cations (Na+= 0.10 nm and K+=0.13 nm). In addition, some authors suggested that the higher dipole moment of potassium ion (~14 Debye) could increase the promotional effect of K+ vs. Na+ (whose dipole moment was ~6 Debye) and H+ ions into the CO2 hydrogenation reaction on Ru catalysts [43]. Hence, Lang et al. [44] clearly showed that the larger the alkali cation was, the greater the electric field suffered by co-adsorbed species located at an adjacent site. These higher promotional effect of K+ vs. Na+ ions have been observed in different chemical reactions through Electrochemical Promotion [35,45,46]. On the other hand, in a previous study of a conventional heterogeneous Ru catalyst doping with K+ and Na+ led to an increase in the ammonia conversion with respect un-doped catalyst at lower temperature, being the Ru-K the highest ammonia conversion due to the lower electronegative of potassium [47].Thus, K+ should perturb the Ru-NH3 bond more strongly than Na+, leading to a higher activation effect. In fact, considering the lower ionic conductivity of K-βAl2O3 (5.5 (ohm.m)−1) vs. Na-βAl2O3 (23.8 (ohm.m)−1) at 300 °C provided by the supplier (Ionotec), lower values of potassium vs. sodium ions coverages on the metal catalyst surface would be attained for the same explored potential range (at the same temperature). Then, the higher effect on the reactants and intermediates chemisorption induced by K+ vs. Na+ can also explain the lower optimal potential values obtained, which maximized the hydrogen production rate. As already mentioned, the presence of an optimal VWR value, and hence, an optimal alkali coverage on the catalyst, is probably due to an excessive decrease on the adsorption of electron acceptor ammonia molecules induced by potassium, which limits step 1 in the overall reaction mechanism. Results are in good agreement with a previous work of EPOC on catalytic ammonia decomposition reaction, in which moderate potassium coverages electrochemically supplied at VWR=1.3 V were found to optimize the catalytic activity of an iron catalyst [17]. On the other hand, it is also interesting to note that a slight permanent EPOC effect was observed at 300 °C (the initial un-promoted catalytic rate was not reached after the final polarization of 2 V). This permanent effect is clearer at 320 °C in which larger differences between the initial and final polarization at VWR=2 V were obtained. This kind of permanent EPOC effect has been also observed in previous studies above certain reaction temperatures working with K+ ions conductor solid electrolytes [35]. It might be explained considering the possible formation of different kinds of promoting phases with higher stability on the catalyst surface, which cannot be electrochemical decomposed under the explored reaction conditions (temperature and electrical potential) [26]. Likely, an applied potential value higher than 2 V would be required for the decomposition of such promotional phases in order to reach the initial un-promoted state. At this point, it should be mentioned that under the explored reaction conditions, K+ ions may also form different kinds of surface species (promotional phases) as a result of their reaction with the chemisorbed reactant and intermediates molecules. These species such as ammonia and nitrogen adsorbed molecules react with K+ ions by distinct charge transfer electrochemical reactions (e.g., potassium nitrites, nitrates or nitrides among others) occurring at the different applied VWR values. In fact, various kinds of promotional phases have been already shown in other EPOC studies with potassium ions conductor electrolyte, verified by ex-situ FTIR [35] and XPS measurements [48]. This in-situ formation of different kinds of adsorbed species will be analysed with more detail in the next section by cyclic voltammetry measurements. Thus, the presence of distinct kind of promotional phases could also confirmed the presence of two local maximum on the hydrogen production rates observed at 320 °C at VWR=0.5 V and −1.5 V. The formation and nature of different kind of promotional phases causing two optimal applied potential values have been reported in a previous work of hydrogen production via partial oxidation of methanol on platinum catalyst film deposited on K-βAl2O3 [26]. Thus, the catalytic rates would be affected by the formation of several kinds of promoter phases as will be also lately confirmed by cyclic voltammetry. The higher promotional effect of potassium vs. sodium ions can be clearly observed also on Figure S4 of the supporting information, which compares the potentiostatic variation of ρ vs. VWR at the two common explored reaction temperatures (300 and 320 °C). It can be observed that in both cases higher ρ values were obtained for Ru/K-βAl2O3 vs. Ru/Na-βAl2O3 at lower VWR values. Hence, in the case of Ru/K-βAl2O3 the catalytic hydrogen production rate via ammonia decomposition under optimal promotor coverage is multiplied by a factor close to 231.8% at 300 °C, which is one of the most important findings of the present work. It led to an overall higher ammonia conversion on the Ru/K-βAl2O3 electrochemical catalyst vs. Ru/Na-βAl2O3 as can be observed on Figure S5 of the supporting information. Considering the similar nature of both kinds of ruthenium catalyst films deposited on both, Na-βAl2O3 and K-βAl2O3 (same preparation procedure and same geometric area of the catalyst working-electrode), the observed difference in the ammonia conversion is clearly due to a higher promotional effect of potassium. Therefore, considering that the most interesting results were obtained by the Ru/K-βAl2O3 electrochemical catalyst, this sample was selected for further reaction and characterization experiments as will be shown below. Fig. 6 shows the ammonia conversion on Ru/K-βAl2O3 electrochemical catalyst through temperature programmed reaction experiments (2 °C·min−1) under application of three different potentials (VWR=2, 0 and −1 V).In first place it can be observed that under the three explored VWR values, ammonia conversion increased with the reaction temperature due to an increase in the reaction kinetics and the endothermic nature of the reaction. In good agreement with previous results, it can be observed that the application of a mild potential value (VWR= 0 V) enhanced ammonia conversion at the whole explored temperature range vs. the un-promoted potential value (2 V). Under these conditions, a moderate amount of potassium ions was supplied to the catalyst leading to an electrochemical activation of the metal catalyst film. This catalyst potential (VWR= 0 V) was selected from the results obtained in Fig. 5 in order to improve the catalytic activity in a wide temperature range (220–320 °C). However, it can be observed again that, a decrease in the potential to VWR=−1 V decreased ammonia conversion in the whole temperature range. It was related to a strong increase on the K+ coverage, which weakens the binding strength of ammonia, decreasing the reaction rate. Moreover, the excess of promoter over catalyst surface, induced by a negative potential, led to the blockage of ruthenium active sites causing electrochemical poisoning, similarly to a previous EPOC study on ammonia decomposition using an Fe electrochemical catalyst under potassium coverages [17]. This kind of poisoning behavior has been also observed in previous studies of catalytic ammonia decomposition on ruthenium catalysts chemically doped with potassium [47] and cesium [49].In any case, results shown in Fig. 6 demonstrate the interest of EPOC phenomenon to activate the catalyst at lower reaction temperatures, which may be of great interest for energy saving, especially for endothermic reactions as the one explored here. In this sense, Figure S6 of the supporting information shows the reaction temperature necessary to achieve a specific ammonia conversion (5 and 10%) as a function of the three explored potentials (VWR=2, 0 and −1 V) used in the experiment of Fig. 6. It can be observed an activation by means of a reduction of 20 °C to achieve 5% of ammonia conversion and 30 °C for 10% ammonia conversion induced by EPOC under VWR= 0 V. At this point it should be mentioned that a further optimization could be performed (out of the scope of the present study), applying the optimal potential value at each explored temperature to maximize the hydrogen production rate.The apparent activation energy values (Ea) were calculated from these experiments via the Arrhenius plot at each applied potentials (VWR=2, 0 and −1 V) as shown in Figure S7 of the supporting information. The obtained values, which were in the range of previously obtained ones with ruthenium catalysts [7,50], decrease from Ea = 80.9 kJ·mol−1 at VWR= 2 V to Ea = 74.6 kJ·mol−1 at VWR= 0 V. This decrease in the Ea induced under electrochemical activation conditions is in good agreement with the promotional mechanism explained above. Hence, the optimal potassium coverages achieved at VWR=0 V allowed an activation of the chemisorption of electron acceptor molecules (N adsorbed molecules), which promotes the rate determining step of ammonia decomposition reaction, decreasing the Ea [17]. In addition, a similar decrease in the apparent activation energy of 5–20 kJ·mol−1 was observed for ruthenium catalysts chemically doped with potassium species on ammonia decomposition reaction [47]. Fig. 7 shows the cyclic voltammetry experiment (CV) of Ru/K-βAl2O3 with the simultaneous recording of ammonia signal under ammonia catalytic decomposition reaction at 280 °C, from VWR=2 V to −2 V and a scan rate of 5 mV·s− 1. Before the cyclic voltammetry experiment, the electrochemical catalysts were kept at 2 V for 30 min in order to define an initial reference state.Staring from 2 V, during the initial forward scan from 2 V to −2 V, it can be observed negative current values attributed to the electrochemical supply of K+ ions from the solid electrolyte to the metal catalyst film with the simultaneous formation of surface promotional phases between the K+ ions and the chemisorbed reactant and product species [51]. According to the obtained CV curve one could envisage different kinds of charge transfer reactions evidenced by the different obtained cathodic peaks vs. other CV curves obtained in EPOC systems where single cathodic peaks were typically observed [51]. It demonstrates the formation of different kinds of promotional phases between the K+ ions and chemisorbed molecules, in good agreement with the different observed promotional effects already discussed in Fig. 5. During the forward CV scan, an increase in the hydrogen production rate was also observed due to the electrochemical supply of K+ ions. In good agreement with the previous results shown in Fig. 5, an optimal potential value was attained corresponding to an optimal alkali coverage on the catalyst surface. During the positive scan from −2 V to 2 V (backward scan) the different promotional phases previously formed on the metal catalyst film were electrochemically decomposed, returning the K+ ions to the solid electrolyte and leading to the appearance of different anodic current peaks. It is interesting to note that during this positive scan a new promotional state is again reached at VWR= −0.5 V maximizing the hydrogen production rate. In addition, during the backward scan a second local maximum in the hydrogen production rate was also observed at VWR= 1 V. Considering the dynamic character of the experiment, these results again confirmed the presence of different kinds of promotional phases which locally optimize the hydrogen production rate at different potentials. Some of these species were likely formed from the partial electrochemical decomposition of the promotional phases initially formed during the forward scan. The existence of a permanent EPOC effect is also clear at the end of the cyclic voltammetry experiment at 2 V which showed a higher hydrogen production rate in the backward vs. the forward scan. At this point (VWR= 2 V), positive current values were obtained showing that certain promotional phases were still present on the catalyst surface and were not completely removed (via electrochemistry). It supports the origin of the permanent EPOC phenomenon observed in Fig. 5 and previously discussed in the case of the Ru/K-βAl2O3 electrochemical catalyst.The electrochemical supply of alkaline ions (Na+ and K+) to a ruthenium catalyst film activates the hydrogen production rate via catalytic ammonia decomposition reaction. The promotional effect is due to the strengthening of the chemisorptive bond of weakly adsorbed N surface molecules which stabilizes N adsorbed species on the ruthenium catalyst surface. This stabilization increases the kinetics of the ammonia decomposition reaction leading to the overall activation of the process.A higher promotional effect of K+ ions was found vs. Na+ ions. It is attributed to the higher electronic effect induced by K+which allows to increase the hydrogen production rate above 230% under optimal potential conditions. However, a large amount of alkali ions supplied to the catalyst at very negative potentials, led to a poisoning effect on the catalytic activity. It is associated with a strong decrease in the chemisorption of the electron donor molecules (ammonia).Temperature programmed reaction experiments show the interest of EPOC for the electrochemical activation of the catalyst at lower temperatures, which may contribute to decrease on the overall energy requirements of the process.A permanent EPOC effect was also found for the case of Ru/K-βAl2O3 which led to a permanent activation of the catalyst under the explored conditions. It was attributed to a higher stability of some promotional phases formed during the negative polarization as supported by cyclic voltammetry. M. Pinzón: Investigation, Methodology, Validation, Visualization, Writing – original draft. E. Ruiz-López: Investigation, Methodology. A. Romero: Conceptualization, Visualization, Writing – review & editing, Supervision, Funding acquisition. A.R. de la Osa: Conceptualization, Visualization, Writing – review & editing, Supervision, Funding acquisition. P. Sánchez: Conceptualization, Visualization, Writing – review & editing, Supervision, Funding acquisition. A. de Lucas-Consuegra: Conceptualization, Visualization, Writing – review & editing, Supervision, Funding acquisition.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This work was supported by the Regional Government of Castilla-La Mancha and the European Union [FEDER funds SBPLY/180501/000281].Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.mcat.2021.111721. Image, image 1

This study reports the electrochemical activation (EPOC) of ruthenium catalyst film with alkaline ion conductors for hydrogen production via catalytic decomposition of ammonia. Two electrocatalysts, Ru/Na-βAl2O3 and Ru/K-βAl2O3 have been prepared, characterized, and tested under low temperature reaction conditions (250–350 °C). The electrochemically supply of moderate amounts of alkaline ions (Na+ and K+) from the solid electrolyte support to the ruthenium catalyst film, activated the hydrogen production rate. The promotional effect has been attributed to a strengthening of the chemisorptive bond of weakly adsorbed N surface species, which stabilizes N adsorbed molecules on the ruthenium catalyst surface and thus facilitating the ammonia decomposition reaction. Among the two alkali ions, the effect of potassium was stronger, increasing the hydrogen production rate above 230% at 300 °C under optimally conditions. Temperature programmed reaction experiments also confirms the interest of EPOC for the activation of the catalyst at low temperatures.

The intensive consumption of fossil fuels along with excessive emission of carbon dioxide (CO2) acceleratingly exacerbate global environmental problems, which severely limit the potential of a sustainable progress of human civilization. 1,2 Developing clean energy conversion technologies becomes extremely urgent to circumvent these challenges. Electrochemical CO2 reduction reaction (CO2RR) under ambient conditions, coupled with renewable electricity sources, represents a promising approach to curb CO2 emissions while generating value-added fuels and chemicals. 3–13 In a variety of CO2RR pathways such as C1 (CO, formate, methane, etc.), 14–21 C2 (ethylene, ethanol, etc.), 22–28 or C3 (n-propanol, etc.), 29,30 the reduction of CO2 to CO is currently one of the most promising practices due to its relatively high selectivity and large current density, as well as the facile separation of gas product from liquid water. More importantly, CO as a fundamental chemical feedstock such as the component of syngas, holds a large market compatibility and a wide range of applications in bulk chemicals manufacturing, medicine, and so on. Despite recent breakthroughs on exploiting various selective catalysts for reduction of CO2 to CO, the ultimate practical viability of this technology, however, is contingent upon the scaling up of CO2RR process, which is still in its infancy with challenges in catalyst cost, product selectivity, scalable activity, as well as long-term stability.On the way of scaling up CO2RR for practical CO2 electrolysis, mass production of high-performance catalysts with cost efficiency is the cornerstone and first step. However, there are only a few known catalysts to date, including Au and Ag noble metals, 31–34 developed to deliver a significant selectivity toward CO evolution. As a cost-effective substitute and for the continuous efforts in our group, 35,36 earth-abundant single-atom catalysts (SACs) provide an intriguing paradigm for CO2-to-CO conversion, with projected high atomic efficiency, superior activity, and selectivity. 37–41 The Ni single atoms coordinated in graphene vacancies, with/without neighboring N coordination, have been demonstrated to be highly selective to CO. 42–45 Nevertheless, the commonly pursued strategies for preparing SACs, 46 e.g., core-shell strategy, confined pyrolysis strategy, and polymer encapsulation strategy are not as straightforward to scale up, and sometimes lack general applicability: most of the carbon precursors, including graphene oxides, 36,45 carbon nanotubes, 47 and metal organic frameworks (MOFs), 43 are either not economically viable for large-scale production, or involve relatively complicated preparation steps; in addition, some of the carbon matrix with nanosheet structures suffer from gas diffusion limit when piled up layer by layer on the electrode, greatly hindering the reduction current density for practical implementation. In this sense, developing a facile process for massive production of SACs becomes an important stepping-stone for practical CO2 electrolysis.Another critical challenge that goes beyond the nature of the electrocatalysts revolves around the low current density needed to maintain a high CO selectivity. In a traditional H-cell device where the catalysts were immerged in liquid water, the maximal CO evolution current was limited by the following two factors: (1) the solubility of CO2 in water is relatively low, and beyond some point the CO2RR current density will be dominated not by the reaction kinetics but by the mass diffusion limitation, and (2) due to the concentrated water molecules around the catalyst surface, once the overpotential is gradually increased for larger current density, the hydrogen evolution side reaction (HER) can take off and eventually dominate the reaction as observed in previous studies. 43,44,48 Fuel cell technology emerges as a platform for maximizing the throughput of CO2RR as reflected in the current and selectivity boost, via preventing the catalyst from direct contact with liquid water, as well as facilitating CO2 gas diffusion. 36,49–54 In addition, the compact design of cell and membrane electrode assembly (MEA) can further boost a practical CO2 electrolyzer system with scalable stacks and gas flow system. 55 Herein, we report the synthesis of high-performance Ni SACs with commercial carbon black particles as the support via a simple and scalable method. The Ni single-atomic sites exhibit excellent performance for CO2RR in a traditional H-cell, with a CO faradic efficiency (FECO) of ∼99% at −0.681 V in 0.5 M KHCO3 aqueous electrolyte. More importantly, large current densities above 100 mA cm−2 with nearly 100% CO generation, which are ∼10-fold higher than the current densities in H-cell, were demonstrated on an anion MEA. An ultra-high CO/H2 ratio of 114, which we define as the “relative selectivity” when the CO selectivity is close to 100% and H2 below 1% by gas chromatography (GC), was achieved while maintaining a significant current of 74 mA cm−2. In addition, after 20 hr continuous operation with an average current density of ∼85 mA cm−2, the CO formation FE was still maintained around 100%, while H2 below 1%. When the Ni SACs were further integrated into a 10 × 10-cm2 modular cell, the CO evolution current in one unit cell can be scaled up to as high as 8.3 A with an FECO of 98.4%, representing a large CO generation rate of 3.34 L hr−1 per unit cell.Instead of starting with well-defined graphene matrix or precursors such as polymers or MOFs, 35,36,43,56 we used commercially available carbon blacks with activated surface to trap Ni single atoms and thus form a similar coordination environment and active sites for CO2-to CO-conversion. Compared with that of graphene nanosheets where the layer-by-layer stacking could block the gas diffusion pathways, 36 the nanoparticulate morphology of the carbon black support further facilitates the CO2 diffusions across the gas diffusion layer to ensure a high local concentration of reactants. An illustration of the synthetic process for the catalyst is shown in Figure 1 . In a typical preparation (see Experimental Procedures), 1 g of activated carbon blacks was well dispersed in water, followed with drop-by-drop addition of Ni2+ solution under vigorous stirring. Due to the presence of defects and oxygen-containing functional groups on the surface as well as the high surface areas, the activated carbon black possesses a high adsorption capacity to metal cations in aqueous solution. To ensure a full, but not excess, adsorption of Ni2+ on the carbon black, the solution was stirred overnight and then centrifuged to collect the products denoted as Ni2+-adsorbed carbon black (Ni2+-CB). Subsequently, the Ni2+-CB was mixed with certain amount of urea as the N source and annealed at elevated temperatures (800°C) in Ar for 1 hr, with gram scale catalysts (denoted as Ni-NCB) produced.The high-resolution transmission electron microscopy (HRTEM) image of Ni-NCB in Figure 2 A shows the onion-like, defective graphene layers in CB particles, which can serve well as the coordination matrix for Ni single atoms. The corresponding aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image reveals the individually and uniformly dispersed Ni atoms as bright spots on the CB nanoparticle (Figure 2B). The individual Ni atoms were well separated from each other and were relatively stable under electron beam irradiation, suggesting strong anchoring (Figure 2C). In supplementation, a large area TEM image confirms that no Ni nanoparticles or clusters were formed on the CBs (Figure S1). Elemental mapping by energy-dispersive X-ray spectroscopy (EDS) demonstrates that Ni and N species are homogeneously distributed throughout the carbon framework (Figure S2). The mass loading of Ni was determined to be ∼0.27 wt % by inductively coupled plasma atomic emission spectrometry (Experimental Procedures). X-ray photoelectron spectroscopy (XPS) characterization of Ni-NCB was performed to further elucidate the profile of elemental composition and related chemical states (Figure S3). The Ni 2p spectrum of Ni-NCB shows a positive Ni 2p3/2 binding energy (854.9 eV) relative to Ni metal (852.6 eV), indicating the positive oxidation states of Ni single atoms (Figure 2D). The XPS N 1s spectrum deconvoluted into pyridinic (∼398.3 eV), Ni-N (∼399.5 eV), pyrrolic (∼400.5 eV), quaternary (∼401.3 eV), and oxidized (∼403.0 eV)-like N species (Figure S4). 45,57 The atomic concentrations of Ni and N in Ni-NCB determined by XPS is 0.28 at % and 1.81 at %, respectively. Synchrotron-based X-ray absorption near-edge spectroscopy (XANES) and extended X-ray fine structure (EXAFS) were used to determine the electronic and local coordination of the single-atomic sites in Ni-NCB (Experimental Procedures). The Ni K-edge XANES profiles in Figure 2E indicate that Ni species in Ni-NCB were in a higher oxidation state than Ni foil and lower than NiO, according to the near-edge position, which is consistent with the XPS results. As shown in the EXAFS results in R space (Figure 2F), Ni-NCB exhibits prominent peaks at 1.4 and 1.9 Å arising from the first shell Ni-N or Ni-C coordination. 58 No other typical peaks for Ni-Ni contribution at longer distances (2.2 Å) were observed. Thus, Ni atoms were atomically dispersed throughout the N-doped carbon blacks. Although different Ni-N and Ni-C structures have been proposed in literatures, 35,36,59 the explicit coordination environment of Ni is still not clear and awaits further exploring.The CO2 electrocatalytic reduction activity and selectivity of Ni-NCB were first evaluated in a standard three-electrode H-cell configuration with CO2-saturated 0.5 M KHCO3 as the electrolyte. In control, an N-doped carbon black (denoted as N-CB) and a Ni-doped carbon black (denoted as Ni-CB) were also prepared for comparison (Figures S5 and S6). As revealed by linear sweep voltammetry in Figure S7, Ni-NCB shows a much higher current density in CO2-saturated electrolyte than that of N2, indicating the participation of CO2 gas in the reaction. Steady-state chronoamperometry of CO2 electrolysis was recorded under different potentials between −0.3 and −1 V versus reversible hydrogen electrode (vsRHE). The FE of gas products were analyzed by online GC (Figures 3 A and S8; Supplemental Information). 35,60 In CO2-saturated 0.5 M KHCO3, Ni-NCB exhibits current densities significantly higher than those of Ni-CB and N-CB (Figure S8). H2 and CO are the major gas products in all these three samples. For Ni-NCB, CO signals were detectable at −0.41 V vs RHE, suggesting that the onset overpotential of CO2 to CO is at least lower than 290 mV. It is noted that the overall FE under this potential is far less than 100%, which is possibly due to the instrumental detection limit. As the potential becomes more negative, the FE of CO increases, while that of H2 decreases correspondingly. A high plateau of CO FEs over 95% was retained under a broad potential range from −0.6 to −0.84 V vs RHE, with a maximum CO selectivity of above 99% at −0.68 V vs RHE while the competitive HER suppressed to 2%. No other liquid products were detected by 1H nuclear magnetic resonance (NMR) (Figure S9). In sharp contrast, NCB exhibits a faint activity for CO generation, indicating that Ni single atoms play a critical role in activating CO2 to produce CO (Figure S8). In addition, Ni-CB only shows a maximum FECO of 29%, which is presumably attributed to the poor dispersion of Ni atoms on the CBs in the absence of nitrogen, as demonstrated in our previous study. 35,36 The partial current shown in Figure 3B demonstrates that the activity of the Ni-NCB is better than, or comparable with, most of the noble-metal-based catalysts reported to date. 32,33,61 Moreover, Ni-NCB exhibits a high intrinsic CO2 reduction activity, reaching a specific CO current of 111 A g−1. Besides, a CO2-to-CO Tafel slope of 101 mV/decade on Ni-NCB (Figure S10) suggests that the first electron transfer process generating surface adsorbed *COOH species is possibly the rate-determining step for CO evolution. 45 To further testify the intrinsic activity of Ni-NCB for CO2 reduction, the CO production turnover frequency (TOF) per Ni single-atomic site is calculated based on the total mass loaded on the electrode, as a minimum value of estimation, as well as the electrochemical double layer capacity (EDLC), as the effective surface area normalization (Figure S11). As shown in Figure 3C, the TOF of Ni-NCB normalized by the mass and electrochemical active surface area (ECSA) for CO production was calculated to be 3.67 and 9.66 s−1, respectively, at an overpotential of 0.56 V, which is better than, or comparable with, those of metal porphyrins or noble metal catalysts in aqueous solutions. 7,8,31,62 Furthermore, to elucidate the influence of Ni content, N doping, as well as annealing temperature, a series of control samples were prepared and tested for CO2 reduction (Figures S12–S18). It shows that both the partial current density and CO FE of Ni-NCB annealed under NH3 atmosphere are slightly lower than those with urea as N precursor. This could be due to the different vapor pressures of the N dopants, or different radicals from N2H4 and NH3 under high temperature. It also reveals that increasing the Ni loading leads to the generation of Ni clusters, which impairs the overall performance for CO2RR (Figure S14). Besides, appropriate temperature and amount of N doping are required to gain the optimal performance of Ni-NCB. More importantly, the CO2-to-CO reduction performance of Ni-NCB is extremely stable, retaining 99% of the initial current for CO formation (∼23 mA cm−2) after 24 hr of continuous operation, with FECO remaining above 95%. Post-catalysis HAADF-STEM imaging and EXAFS (Figure S19) show that those Ni species still maintain the feature of well-dispersed single atoms, reiterating the excellent chemical stability of the Ni atomic sites in Ni-NCB.The scaling up of CO generation rate in a traditional H-cell is limited by the following two factors: (1) a larger overpotential is usually required to deliver a higher kinetic current, which, however, can promote strong HER competition due to the contact between catalyst and liquid water, and (2) the reduced CO2 gas reactant in an H-cell configuration is that dissolved in liquid water, therefore the reaction rate beyond a certain point is limited by CO2 mass diffusion. To circumvent this issue and inspired by fuel cell reaction mechanisms, an anion MEA was adopted in a gas-phase electrochemical reactor to greatly boost the current density while maintaining high CO selectivity (Experimental Procedures). 36 On the cathode side, humidified CO2 gas was supplied. This high concentration of CO2 and low concentration of H2O vapor can block the direct contact between catalyst and liquid water and prevent limiting of reactant diffusion. On the anode side, 0.1 M KHCO3 solution was circulated whereby the water oxidation is taking place (Figure S20). As shown in Figure 3E, the CO2 conversion increases rapidly above 2.1 V cell voltage and reaches a significantly high current density of 130 mA cm−2 at only 2.7 V without iR compensations. Notably, the catalyst maintains nearly 100% FE for CO formation across a broad range of current densities from 30 to 130 mA cm−2, while the FE of H2 was suppressed to a minimum of 0.9% (Figures 3F and S21). It is important to mention here that, due to the experimental errors introduced by GC detection, the measured CO selectivity could sometimes be slightly higher than 100%, especially when H2 was suppressed to below 1%. In this case, we propose to define the CO/H2 ratio, which we denote as relative selectivity, as an additional criterion to more accurately evaluate the high selectivity toward CO evolution. As shown in Figure 3G, with the gradual increase of cell voltage, the CO/H2 ratio increases accordingly and reaches a maximum value of 113.8, with a high CO2RR current density of 74 mA cm−2. This is to our knowledge the highest ratio of CO/H2 under a significant current density compared with the most active catalysts reported to date (Table S1). An impressive stability of the catalyst in this gas-phase electrochemical reactor is also presented in Figure 3H, with an average current density of 85 mA cm−2 over 20 hr continuous electrolysis, while maintaining CO formation FEs ∼100% and H2 below 1%. The slight degradation of the current density was probably attributed to several factors including the deactivation of catalysts, the corrosion of the gas diffusion layer, as well as membrane degradation (Figures S22–S24). Overall, this high performance of the Ni-NCB catalyst in the gas-phase electrochemical reactor opens up great opportunities in scaling up highly selective CO2 reduction.Motivated by the superior activity of Ni-NCB and its facile synthesis process, it is expected that, by increasing the catalyst loading, extending the size of the gas diffusion layer, as well as alternatively stacking anodes and cathodes, the Ni-NCB integrated gas-phase electrochemical reactor can be further scaled up to produce large CO generation currents for potential practical applications. Here we customized one unit cell with a 10 × 10-cm2 anion MEA as a preliminary demonstration to justify this application possibility in the future (Figures 4A–4C). Considering the high CO2 flow rate needed to ensure sufficient reactants, gas collecting bags were used to collect the gas products which were later analyzed by GC under different cell voltages (Experimental Procedures). As shown in Figures 4D–4F, a record-high CO2RR current of 8.3 A was achieved with a high CO selectivity approximating to 99% and H2 about 1%. Delivering an average current of ∼8 A for stability test, our device maintained a stable CO selectivity of more than 90% for over 6 hr continuous electrolysis with a total volume of 20.4 L CO generated (Figure 4G). This represents a CO generation rate of 3.42 L hr−1 or 0.14 mol hr−1 and a conversion rate of 11.33%.It is anticipated that this work can be further pushed forward toward commercialized CO2 electrolysis by optimizing several technological aspects according to industry standards. First, stability is one of the major concerns, which suffers from the corrosion of both anode and cathode, as well as membrane and electrode decay, which require tremendous efforts to overcome. In addition, CO2 feed recycling can be set up by separating CO2 from gas products to achieve a sustained CO2 supply. The cost of the anode should also be taken into consideration, which can be greatly reduced by replacing IrO2 with efficient transition metal-based materials. Meanwhile, one circumstance should be paid attention to, where the metal leaching happens from the anode to be deposited onto the cathode, which will hamper the CO2RR by encouraging competitive HER.In conclusion, a highly efficient transition metal-based SAC was synthesized via an economic and scalable protocol, and applied in CO2 electrolysis for large-scale production of CO. The results demonstrate that it is promising to replace noble metal catalysts, such as Au or Ag, with earth-abundant materials with remarkable CO evolution performance approaching practical expectations, which opens an avenue for future renewable energy infrastructures and achieves a significant progress in closing the anthropogenic carbon cycle for global sustainability.The carbon blacks were activated by dispersing 2 g carbon blacks in 100 mL of 9 M nitric acid solution followed with refluxing at 90°C for 3 hr. The Ni-NCB catalyst was prepared via a facile ion adsorption process followed with further pyrolysis. Typically, a 3-mg/mL nickel nitrate stock solution was first prepared by dissolving Ni(NO3)2⋅6H2O (Puriss, Sigma-Aldrich) into Millipore water (18.2 MW⋅cm). A carbon black suspension was prepared by mixing 1 g activated carbon blacks (Vulcan XC-72, purchased from Fuel Cell Store and activated in acid bath) with 400 mL of Millipore water, and tip sonicated (Branson Digital Sonifier) for 30 min until a homogeneous dispersion was achieved. Then 40 mL of Ni2+ solution was dropwise added into carbon black solution under vigorous stirring overnight and then centrifuged to collect the products (Ni2+-CB). The as-prepared Ni2+-CB powder was mixed with urea with a mass ratio of 1:10, and then heated up in a tube furnace to 800°C under a gas flow of 80 standard cubic centimeters per minute (sccm) Ar (UHP, Airgas) and maintained for 1 hr, obtaining the final products. NCB and Ni-CB were prepared in a similar way but with the absence of Ni precursor and urea, respectively. Ni-NCB-NH3 was prepared by annealing the as-prepared Ni2+-CB powder at 800°C under a gas flow of 80 sccm NH3. Ni-NCB-1:5 and Ni-NCB-1:20 were prepared in the same way as Ni-NCB, except by varying the mass ratio of Ni2+-CB powder and urea to 1:5 and 1:20. Ni-NCB-600 and Ni-NCB-1000 were prepared by just varying the annealing temperature to 600°C and 1,000°C. Ni excess was prepared with a modified strategy reported before. 36 The electrochemical measurements were run at 25°C in a customized gastight H-type glass cell separated by Nafion 117 membrane (Fuel Cell Store). A BioLogic VMP3 work station was employed to record the electrochemical response. The set-up of the three-electrode test system can be found in our earlier reports. 35,36 Typically, 5 mg of as-prepared catalyst was mixed with 1 mL of ethanol and 100 μL of Nafion 117 solution (5%, Sigma-Aldrich), and sonicated for 20 min to get a homogeneous catalyst ink. Ink (80 μL) was pipetted onto a 2-cm2 glassy carbon surface (0.2 mg/cm2 mass loading). For the stability test, 500 μL of the ink was air-brushed onto a carbon fiber paper gas diffusion layer toward a mass loading of 1.25 mg/cm2, and then vacuum dried prior to use. All potentials measured against a saturated calomel electrode were converted to the RHE scale in this work using E (vs RHE) = E (vs SCE) + 0.244 V + 0.0591*pH, where pH values of electrolytes were determined by an Orion 320 PerpHecT LogR Meter (Thermo Scientific). Solution resistance (Ru) was determined by potentiostatic electrochemical impedance spectroscopy at frequencies ranging from 0.1 Hz to 200 kHz, and manually compensated as E (iR corrected versus RHE) = E (vs RHE) − Ru *I (amps of average current).For the anion MEA test (or scale-up fuel cell test), 1.25 mg/cm2 Ni-NG and IrO2 was air-brushed onto two 2 × 2-cm2 (or 10 × 10-cm2) Sigracet 35 BC gas diffusion layer electrodes as a CO2RR cathode and an oxygen evolution reaction anode, respectively. A PSMIM anion-exchange membrane (Dioxide Materials) was sandwiched by the two gas diffusion layer electrodes to separate the chambers. On the cathode side, a titanium gas flow channel supplied 50 sccm (or 500 sccm) humidified CO2 while the anode was circulated with 0.1 M KHCO3 electrolyte at 2 mL min−1 (or 10 mL min−1) flow rate. The cell voltages in Figures 3E–3H were recorded without iR correction. The 10 × 10-cm2 MEA response was recorded by a Sorensen DCS 33-33 power supply and is shown in Figure 4 without iR correction.During electrolysis, CO2 gas (99.995%, Airgas) was delivered into the cathodic compartment containing CO2-saturated electrolyte at a rate of 50.0 sccm (monitored by an Alicat Scientific mass flow controller) and vented into a Shimadzu GC-2014 GC equipped with a combination of molecular sieve 5A, Hayesep Q, Hayesep T, and Hayesep N columns. 35,60 A thermal conductivity detector was mainly used to quantify H2 concentration, and a flame ionization detector with a methanizer was used to quantitative analysis CO content and/or any other alkane species. The detectors are calibrated by three different concentrations (H2: 100, 1,042, and 49,830 ppm; CO: 100, 496.7, and 9,754 ppm) of standard gases. The gas products were sampled after a continuous electrolysis of ∼15 min under each potential. The partial current density for a given gas product was calculated as below: j i = x i × v × n i F P 0 RT × ( e l e c t r o d e a r e a ) − 1 where x i  is the volume fraction of certain product determined by online GC referenced to calibration curves from three standard gas samples, v is the flow rate, n i is the number of electrons involved, p 0 = 101.3 kPa, F is the Faraday constant, and R is the gas constant. The corresponding FE at each potential is calculated by FE = j i j total × 100 % For a 10 × 10-cm2 MEA, the FEs of H2 and CO were tested ex situ and calculated based on the concentration normalization.1D 1H NMR spectra were collected on an Agilent DD2 600 MHz spectrometer to test if any liquid products present during the CO2 reduction (Figure S9). Typically, 600 μL of electrolyte after electrolysis was mixed with 100 μL of D2O (Sigma-Aldrich, 99.9 at % D) and 0.05 μL DMSO(Sigma-Aldrich, 99.9%) as internal standard.Calculation of TOF by mass loading normalization: catalyst loading on glass carbon electrode is 0.2 mg cm−2. The content of Ni in Ni-NCB is 0.27 wt %. The moles of active sites per cm2: N = 0.2 × 10 − 3 × 0.27 × 10 − 2 58 = 9.31 × 10 − 9 m o l e c m − 2 TOF ( s − 1 ) = J × F E CO × 0.965 2 × 96485.3 × 9.31 × 10 − 9 Calculation of TOF by ECSA normalization: according to the reported EDLC value of graphene ∼21 μF/cm2(36), the electrochemical surface area of graphene layers in Ni-NCB was calculated to be 390.5 cm2, given the 8.2 mF/cm2 EDLC value of Ni-NCB. The moles of carbon atoms on the electrochemical surface can be calculated to be 390.5 × 10−4/2,600 × 12 = 1.25 × 10−6 mol, where 2,600 m2 g−1 is the theoretical specific surface area of graphene. Taken together the Ni atomic content in Ni-NG was determined to be 0.28% by XPS (Figure S1), and the number of Ni sites in the surface was N = 3.5 × 10−9 mol. Accordingly, TOF ( s − 1 ) = J × F E CO × 0.965 2 × 96485.3 × 3.5 × 10 − 9 .The STEM characterization in Figure 1A was carried out using a JEOL ARM200F aberration-corrected scanning transmission electron microscope at 200 kV with an image resolution of ∼0.08 nm. All other TEM images were obtained by using a JEOL 2100 transmission electron microscope operated under 200 kV. EDS analysis was performed at 300 kV using Super-X EDS system in a Probe-corrected FEI Titan Themis 300 S/TEM. Drift correction was applied during acquisition. XPS was obtained with a Thermo Scientific K-Alpha ESCA spectrometer, using a monochromatic Al Kα radiation (1,486.6 eV) and a low energy flood gun as neutralizer. The binding energy of the C 1s peak at 284.6 eV was used as reference. Thermo Avantage V5 program was employed for surface componential content analysis as well as peaks fitting for selected elemental scans. XAS spectra on Ni K-edge was acquired using the SXRMB beamline of Canadian Light Source. The SXRMB beamline used an Si(111) double-crystal monochromator to cover an energy range of 2–10 keV with a resolving power of 10,000. The XAS measurement was performed in fluorescence mode using a four-element Si(Li) drift detector in a vacuum chamber. The powder sample was spread onto double-sided, conducting carbon tape. Ni foil was used to calibrate the beamline energy.This work was supported by the Rowland Fellows Program at Rowland Institute, Harvard University. The Center for Nanoscale Systems (CNS) is part of Harvard University. This research used resources of the Canadian Light Source, which is supported by NSERC, the National Research Council Canada, the Canadian Institutes of Health Research, the Province of Saskatchewan, Western Economic Diversification Canada, and the University of Saskatchewan. J.L. and N.T. were supported by the National Science Foundation under CHE-1465057, and gratefully acknowledge the use of facilities within the John M. Cowley Center for High Resolution Electron Microscopy at Arizona State University. T.Z. and N.T. acknowledge funding from the China Scholarship Council (CSC) (201706340152 and 201704910441, respectively). J.Z. acknowledges support from MOST of China (2014CB932700) and NSFC (21573206). This work was performed in part at the CNS, a member of the National Nanotechnology Infrastructure Network (NNIN), which is supported by the National Science Foundation under NSF award no. ECS-0335765. H.W. acknowledges support from Rice University.H.W. designed the studies. T.Z. conducted the synthesis and catalytic tests of catalysts. K.J. performed the characterization of catalysts. N.T. and J.L. conducted HRTEM characterization. Y.H. performed XAFS measurements. J.Z. provided suggestions on the work. T.Z. and H.W. wrote the manuscript. All authors discussed the results and commented on the manuscript.H.W. has submitted a patent application (US 62/486,148, 2017) regarding the transition single-atom catalyzed carbon dioxide conversion technology.Supplemental Information includes 24 figures and 1 table and can be found with this article online at https://doi.org/10.1016/j.joule.2018.10.015. Document S1. Figures S1–S24 and Table S1 Document S2. Article plus Supplemental Information

The scaling up of electrocatalytic CO2 reduction for practical applications is still hindered by a few challenges: low selectivity, small current density to maintain a reasonable selectivity, and the cost of the catalytic materials. Here we report a facile synthesis of earth-abundant Ni single-atom catalysts on commercial carbon black, which were further employed in a gas-phase electrocatalytic reactor under ambient conditions. As a result, those single-atomic sites exhibit an extraordinary performance in reducing CO2 to CO, yielding a current density above 100 mA cm−2, with nearly 100% selectivity for CO and around 1% toward the hydrogen evolution side reaction. By further scaling up the electrode into a 10 × 10-cm2 modular cell, the overall current in one unit cell can easily ramp up to more than 8 A while maintaining an exclusive CO evolution with a generation rate of 3.34 L hr−1 per unit cell.

With recent changes in the energy-consuming sector's advancements and depleting fossil fuels, significant attention has been drawn to transform renewable resources into transportation fuels and fine chemicals [1,2]. Selective chemical conversion of lignocellulosic biomass, especially cellulose and hemicellulose, provides platform chemicals such as 5-hydroxymethylfurfural, levulinic acid (LA), and furfural, which can be further down streamed to versatile alkyl levulinates [3,4]. Alkyl levulinates possess high lubricity, low toxicity, stable flashpoint, and moderate flow properties, making them suitable gasoline and diesel blends. Moreover, alkyl levulinates have numerous applications as a solvent, plasticizer, and precursor to obtaining γ-valerolacotone [5]. Among the alkyl levulinates, butyl levulinate (BL), an oxygenate fuel additive with high octane number, rich oxygen content, and low solubility in water makes it a better fuel blend than ethyl levulinate and an alternative for the water-soluble carcinogen methyl tert-butyl ether (MTBE) [6].The esterification of LA with alcohols and alcoholysis of furfuryl alcohol (FAL) in the acid-catalyzed environments are the general methods to produce the alkyl levulinates. However, the production of butyl levulinate using LA as a raw material is an expensive process, with product water inhibit the reaction to progress effectively [7,8]. The other route via FAL's alcoholysis has gained much attention due to its smooth and cost-effective pathway for the BL synthesis. The BL production was reported using various homogeneous acid catalysts such as H2SO4, AlCl3 [9], and a double SO3-H functionalized ionic liquids, [10], which surmises that the butanolysis of FAL is a function of strong acidity. To avoid the major drawbacks of homogeneous systems comprehended by potential reactor corrosion, recovery, and recyclability, several heterogeneous catalyst systems such as ion exchange resins [11], zeolites, SBA-16 [6], metal oxides [12,13], mesoporous aluminosilicates [14], and zinc exchanged heteropoly acids [15] catalysts have been successfully employed for the alkyl levulinates production.To accommodate butanolysis reaction, a facile, highly stable, and easily regenerable tungstated zirconia (WZr) catalyst [16] was first employed to test its activity. Hino and Arata first reported the WZr catalyst's synthesis to cope with the challenges posed by the desorption of active SO4 2- ions from SO4 2-/ZrO2 catalysts [17]. W-O-Zr bonds' presence makes it strongly stable, which overcomes the detachment of active WOx species from ZrO2 support. WZr catalysts efficiently enhanced the alkanes isomerization [18], hydrogenolysis [19], dehydration [20], aqueous phase hydrolysis [21], selective catalytic NOx reduction [22], cyclohexane hydration [23], liquid phase Beckmann rearrangement [24], esterification and transesterification reactions [25]. Besides, LA's esterification with butanol was carried out to obtain BL with 97% selectivity [26]. Owing to the excellent catalytic activity of WZr, this study was interested in reporting the WZr activity for the butanolysis of FAL with lower initial butanol to FAL mole ratios.Upon the WZr catalyst activity for butanolysis of FAL, with barely around 28% of the BL yield, the reaction was carried out using sulfonated carbon catalyst. Carbon-based catalysts with low synthesis cost, high catalytic activity, and abundant carbon sources make them suitable heterogeneous catalytic systems [27]. The partially pyrolyzed carbon source is rich in oxygen content composed of several surface functional groups such as carboxylic acid, aromatic hydrocarbons, and phenolic hydroxyl groups [28]. Thus, obtained carbon black can easily be functionalized with -SO3H groups by sulfonation with various -SO3H group sources, for instance, conc. H2SO4, fuming H2SO4, and 4-Benzenediazoniumsulfonate. Literature reports were available to find the excellent catalytic activity of sulfonated carbon catalysts for a wide range of reactions like fatty acids (FFA) esterification [29,30], cellulose hydrolysis [31], transesterification of vegetable oils [32], glycerol conversion [33], and esterification reactions [34–36].Most of the studies reported for the butanolysis of FAL were with high initial 1-butanol to FAL molar ratios to avoid the formation of FAL polymers. Bringué et al. performed the butanolysis of FAL over sulfonated polystyrene-divinylbenzene (PS-DVB) resins at low initial 1-butanol to FAL molar ratios (8:1) with a maximum of 63% BL yield at 110 °C. This study also reported that 2-BMF was still unconverted to BL at the end of the reaction time owing to the reaction temperature. The Amberlyst catalysts are thermally challengeable above 120 °C and the lower reaction temperature influenced to obtain the maximum yield [11]. Similarly, Enamula et al. also employed a high initial mole ratio of 16 at high reaction temperature (180 °C) to obtain 63 mol% of BL yield using Al2O3/SBA-15 catalyst. The same catalyst at 110 °C resulted in 91 mol% of BL yield but at the initial mole ratio of 65 [13]. Yang et al. reported the activity of a magnetic carbonaceous solid acid (SMWP) catalyst with 91 mol% of BL yield, operating at an initial mole ratio of 40 (BtOH: FAL) [8]. Thus, the present study focuses on increasing the BL yield to a maximum level at a possible lower initial molar (BtOH: FAL) ratio.Pluronic P-123, zirconium (IV) butoxide solution (80 wt.% in 1-butanol), ammonium metatungstate hydrate, Conc. HNO3 (65%), n-Butanol (purity > 99.0%), carbon tetrachloride (CCl4), and methanol were acquired from Sigma-Aldrich. Ethanol, furfuryl alcohol, sucrose, and sulfuric acid were obtained from Fuji film Wako chemicals Ltd. n-Butyl levulinate and dibutyl ether were purchased from TCI chemicals. All the chemicals were used as received.The catalyst synthesis was carried out using the evaporation-induced self-assembly (EISA) method mentioned elsewhere to obtain a mesoporous structured catalyst [37]. Ammonium meta tungstate hydrate (0.05074 mmol) and zirconium (IV) butoxide solution (80 wt.% in 1-butanol) (68.982 mmol) were used as precursors for WO3 and ZrO2, and pluronic P123 (1.724 mmol) was used as a structure directive agent. In a typical synthesis, Pluronic P123, tungsten, and zirconia precursors were dissolved in 250 ml of ethanol. To this solution, 2 ml of H2O and 8 ml of HNO3 were added to promote the condensation during the synthesis and to maintain the pH of the solution below the electrostatic point of the tungsten and zirconia. The solution was aged for 12 hours with continuous stirring of 300 rpm. After that, the homogeneous mixture was dried in a hot air oven at 40 °C for 48 hours to facilitate the slow evaporation and then completely dried further at 70 °C for 12 hours. Finally, thus obtained solid was calcined in air at 800 °C for 6 hours with a ramping rate of 1 °C/min (25 → 200 °C, 1-hour stay at 200 °C, 200 → 400 °C, 1-hour stay at 400 °C, 400 → 800 °C, 6 hours stay 800 °C). The prepared catalysts were termed xWZrT (where × -wt.% of WO3 & T-calcination temperature, ⁰C). The promoted catalyst was prepared by the wet impregnation method with noble and novel metals and further calcined at respective reduction temperatures of the metals. The notation yM15WZr800 (M- Metal, y-wt.% of the metal) was used to represent the promoted catalysts.The catalyst was synthesized by incomplete carbonization of the sucrose, followed by sulfonation at the designated temperature for a particular time under inert conditions [38–40]. The catalyst preparation was done in two steps. In the first step, 5 g of sucrose was partially carbonized at 400 °C for 15 hours under constant N2 flow (200 ml/min) to obtain a black solid. Thus, obtained carbon black was grounded to a fine powder. In the second step, 2 g of carbon black was sulfonated using 40 ml of conc. H2SO4(>98%) at 80 °C for 10 hours under inert conditions. After the sulfonation, the black mixture was washed with hot distilled water and then vacuum filtered until the neutral pH of the water was observed. Finally, the catalyst was dried at 100 °C for 12 hours before the direct use. The prepared catalyst was denoted by SO3H_C80S (Carbon black sulfonated at 80 °C) and the spent catalyst after 3 cycles by SO3H_C80S Spent.The specific surface area and pore size distribution of the prepared catalysts were analyzed using nitrogen adsorption/desorption isotherms data obtained at 77 K using the BELSORP-MiniX analyzer. The catalyst samples were first degassed under vacuum (10−5 torr) conditions at 573 K (Tungstated zirconia catalysts) and at 393 K (carbon catalysts) for 4 hours to remove the surface adsorbed species and moisture on the catalysts. The specific surface area was determined by the adsorption isotherm of nitrogen in the relative pressure range of 0.05 < p/p0 < 0.3 using the BET equation. The pore size distribution and pore volume were determined by the BJH desorption method using desorption isotherm.Fourier-transform infrared spectroscopy (FTIR) was done using a Bruker ALPHA II. The samples were mixed with spectroscopic grade potassium bromide (KBr, 100 mg) and pressed to acquire a circular transparent disk with a hydraulic press. The spectra were collected from 4000 to 400 cm−1 with a resolution of 4 cm−1 for 16 scans using KBr disks.The pyridine probed FTIR was performed to distinguish the acidic sites present on the catalyst. Pyridine was added to the catalyst samples and allowed pyridine to adsorb on the samples' surface for 2 hours at room temperature. The unadsorbed pyridine was removed by keeping the samples in the oven at 383 K for 2 hours. 100 mg of KBr was added to the samples, and the transparent pellets were made using the hydraulic press. The IR spectra were recorded against the KBr background.The acidic properties of the tungstaed zirconia and metal promoted catalysts were estimated by the NH3 adsorption and temperature-programmed desorption (NH3-TPD) technique using BEL-CAT (MicrotracBEL corp.) automated chemisorption analyzer with a TCD detector. The samples were first pretreated with helium gas (50 ml/min) at 250 °C for 1 hour and cooled down to 100 °C. After that, ammonia adsorption was carried out using a 5% NH3 gas mixed with helium (95%) for 30 minutes at 100 °C. After completing ammonia adsorption, the samples were purged with pure He for 30 minutes and allowed the TCD stabilization. Finally, the ammonia desorption spectra were obtained by gradually increasing the temperature with 10 °C/min until the final temperature, followed by the calibration with a 5% NH3-He mixture. The amount of ammonia adsorbed in mmol/g was automatically calculated by the ChemMaster software using the calibration curve and the amount of ammonia taken. The ammonia adsorption technique was not performed for carbon catalysts as the detachment of -SO3H progressed at higher temperatures (>230 °C) during the analysis (TGA analysis, SFig.2.).The acidic sites of the carbon catalysts were measured by the acid-base back titrations using an aqueous ion-exchange method using NaHCO3 base solution followed by the titration against aqueous HCl solution [41]. In a typical process, 30 mg of catalyst was dispersed in 0.005 N NaHCO3 solution and continuously stirred for 24 hours. After that, the resulting mixture was filtered, and the filtrate was titrated against 0.005 N HCl solution using a methyl orange indicator. The quantification of the acid sites on the catalysts was calculated by the amount of NaHCO3 consumed.The catalytic activity was tested for the butanolysis of FAL in a 100 ml high-pressure batch reactor (Parr Instruments). In a typical experiment, specified amounts of reactants and catalyst were charged into the reactor. The reactor was purged with nitrogen gas several times to ensure the inert environment and then pressurized. The reaction temperature and agitation speed were fixed, and the reaction was carried for the specified times after reaching the set temperature. After completing the reaction, the reactor was cooled down to room temperature, and the sample was collected. The liquid samples were centrifuged to remove the catalyst traces and diluted with internal calibration solvent carbon tetrachloride and dilutant methanol before analyzing with GC-FID (flame ionization detector) and GC–MS (mass spectroscopy). The following equations were used for the quantification calculations [33]. C o n v e r s i o n ( m o l % ) = I n i t i a l m o l e s - F i n a l m o l e s of F A L I n i t i a l m o l e s of F A L × 100 S e l e c t i v i t y ( m o l % ) = m o l e s o f a p r o d u c t o b t a i n e d t o t a l m o l e s o f the p r o d u c t × 100 Y i e l d ( m o l % ) = m o l e s o f B L f o r m e d t h e o r e t i c a l m o l e s o f B L e x p e c t e d × 100 The reuse test was conducted to determine the number of cycles that the catalyst can be used without requiring regeneration. Thus, three reaction cycles of the best performing catalyst were performed for the spent analysis. After completing every reaction, the catalyst was separated from the reaction mixture by vacuum filtration and several times washings with methanol followed by ethanol to remove the adsorbed organic compounds. After that, the catalyst was dried at 100 ⁰C for 12 hours before use.The N2 physisorption illustrates the physical properties of the prepared WZr catalysts. The typical isotherms correspond to type IV isotherm represents the mesoporous nature of the catalyst. Fig. 1 a and 1b further showed that for a 15 wt.% WO3 loading, the incipient of the hysteresis loop shifted towards higher relative pressure with the increase in calcination temperature. On the contrary, for a fixed calcination temperature of 800 °C, the beginning of the hysteresis loop was observed at P/P0 of 0.55 for 10WZr800 and shifted towards higher relative pressure with the increase in the WO3 content in tungstated zirconia catalysts. These results exhibited the expansion and contraction of the pores with the increase in calcination temperature for fixed WO3 content and WO3 loading for a fixed calcination temperature, respectively. The pure zirconia's surface area was also quite low, representing its non-porous nature at such a high calcination temperature. The pure zirconia and pure WO3 have a poor surface area (Table 1 , entries 1,2), which explains the integration of W-O-Zr bonds needed to stabilize this catalyst, which is responsible for its strong thermal and mechanical strength [42]. An increase in calcination temperature at a constant metal oxide loading increases the surface density of the metal on the support oxide, which triggers the mobility of support metal atoms [43]. In the case of tungstated zirconia catalyst, calcination at higher temperature promoted Zr atoms mobility, which triggered the sintering of Zr atoms along with augmentation of pore size [26]. These phenomena were observed for 15WZr900. The specific surface area of the catalyst was reduced to 17.5 m2/g with widened pores (Table 1, entry 5). The surface density of the W dispersion on the ZrO2 support was calculated by the following equation, which is a measure of the tungsten monolayer coverage [44]. Surface density of W = ( WO 3 Loading ( wt . % ) / 100 ) × 6 . 023 × 10 23 231 . 8 ( formula weight of WO 3 ) × BET Surface area ( m 2 g - 1 ) × 10 18 An increase in WO3 loading increased the surface area to 79 m2/g for 15 wt.% calcined at 800 °C (Table 1, entry 4) for which the surface density was around 4.9 W-atom/nm2, which is in the range for a typical value of the surface monolayer coverage [45,46]. The pore size distribution measured by the BJH desorption method was also represented in Fig. 1b. The 15WZr800 catalyst has a pore diameter of 10.6 nm compared to 8.3 nm of 10WZr800, and the depth of the pores for the 15 Wt.% is longer than 10 Wt.%, which was clearly shown by the increased pore volume for the 15 Wt.% loading (Table 1, entries 3,4). The further increase in the calcination temperature to 900 °C profoundly affected the pore structure of the catalyst, resulting in the widening of the pores caused by the sintering of the Zr atoms (Table 1, entry 5) [16].The specific surface area of the prepared carbon-based catalysts measured by the N2 adsorption technique at 77 K was reported in the following Table 2 . The specific surface area of the catalysts is relatively influenced by the carbonation temperature and sulfonation temperatures [47]. The carbon catalysts have the specific surface area < 1 m2/g (Table 2, entries 1–3), mostly because of the formation of amorphous carbon by incomplete carbonization of the sucrose, which resulted in the dispersion of large phenol hydrophilic and carboxylic acid functional groups [29,36]. There was no significant change in the surface area of the carbon black after the sulfonation with conc. H2SO4 indicates that the –SO3H groups were incorporated into the carbon structure by bonding with the existing functional groups. The pore volume and pore diameter of the catalysts were very low, caused by no surface area development as well as the pore structure. N2 adsorption isotherms and micropore analysis are shown in the Supporting information (SFig.8a &8b).NH3-TPD study was implemented to understand the surface acidic properties of the tungstated zirconia catalysts and the metal promoted catalysts. The NH3-TPD spectra are shown in Fig. 2 a and b. The total surface acidity and the peak temperatures are summarized in Table 3 . The ammonia desorption was obtained in a broad range from 100 °C to 800 °C to understand the acid strength of the catalysts. To comprehend and quantify the surface acidity, the peak temperatures below 350 °C were assigned to the weaker acidic sites. The peaks above 350 °C correspond to the stronger acidic sites [25,26]. However, the intensity of stronger acidic strength peaks was very low compared to the lower temperature peaks.The surface acidity was clearly affected by WO3 loading and the calcination temperature (Table 3, entries 1,2 &3). An increase in WO3 loading from 10 wt.% to 15 wt.% induced a hike in the surface acidity from 0.143 mmol/g to 0.201 mmol/g (Table 3, entries 1,2). For 15 wt.% WO3 loading, the surface density was 4.9 W-atoms/nm2 (Table 1, entry 4), which generally corresponds to the monolayer coverage of the W-atoms, thereby increasing the surface acidity [46]. The collapse of the pore structure (BET results, Fig. 1b) at such a higher calcination temperature (900 °C) probably caused the drop in the total acidity to 0.122 mmol/g.The peak corresponding to weaker acidic strength was observed in the range of 175–198 °C for all the catalysts, which indicates that the metal promotion influenced the strong acidic sites upon interaction with the W atoms. The surface acidity of all metal promoted catalysts constrained in a narrow range from 0.199 to 0.209 mmol/g (Table 3, entries 4–9). The yield of BL was also restricted to a narrow range of 10.7 mol% to 14.2 mol% (Table 3, entries 4–9). Even though the enhancement in terms of the acidity of the tungstated catalyst was achieved very slightly by the metal promotion, it has not led to a significant difference in the activity towards increasing the BL yield. Fig. 3 . demonstrates the FTIR spectra of sulfonated carbon catalysts and carbon black at 298 K in the range of 4000 to 400 cm−1, which indicates the functional groups present on the surface of the catalyst and their interaction. The band at 1058 cm−1 correlated to the symmetrical stretching of O = S = O in SO2 bonding and the band at 1162 cm−1 to the asymmetrical stretching of SO2, which were not observed in the carbon black compared to sulfonated catalysts [40,41,48]. These functional groups clearly indicate the incorporation of -SO3H groups onto the carbon black. The reduction in the intensity of the peaks at 1162 cm−1 and 1058 cm−1, which corresponds to asymmetrical and symmetrical stretching of -SO3H, supports the desorption of -SO3H groups after the reaction. The polyaromatic hydrocarbon of C = C stretching was due to the vibration band at 1606 cm−1 [34,41]. The carboxylic acid groups, one of the major functional groups formed by the partial carbonization of the carbon source, were attributed to the presence of a vibration band at 1700 cm−1, which represents the stretching of a C = O of a –COOH group [32]. The band at 2930 cm−1 illustrates aromatic methoxyl groups, which were subsequently suppressed during the sulfonation process. Finally, the band at 3600 cm−1 corresponds to the –OH stretch phenolic functional groups [49]. Thus, FTIR analysis provided insights into the surface functional groups of the carbon catalysts and the desorption of -SO3H groups.To distinguish the nature of the surface acidity of the tungstated zirconia catalyst, Pyridine-FTIR analysis was studied and represented by Fig. 4 . and the peak areas were reported in Table 4 . The band at 1440 cm−1 corresponds to the interaction of pyridine Bronsted molecules with H+ electron-accepting molecules corresponding to Lewis acidity. The band at 1540 cm−1 attributed to the presence of Bronsted acidic molecules. The band at 1490 cm−1 represents the combined acidic strength of Lewis and Bronsted. The intensity of the bands progressed with WO3 loading from 10 wt.% to 15 wt.% (Table, entries 1 & 2). Increasing the WO3 content led to the development of polytungstate species and the formation of Zr-WOx clusters [49,50]. At 15 wt.% of WO3 loading, the monolayer coverage was accomplished by progressing the condensation of monotungstate to polytungstate species by enhancing the surface acidity. The tungsten atom in the polytungstate has the ability to delocalize the adjacent zirconia electrons, thereby generating the proton development to compensate the delocalized electrons, which enhances the growth of the Bronsted acidity [51,52]. At a higher calcination temperature of 900 °C, the band intensity corresponding to Lewis acidity weakened, and other peak intensities reduced considerably. The basic reason attributes to the pore structure collapse at elevated temperatures, accompanied by the Zr atoms sintering [16,53].As described in the NH3-TPD analysis section, the metal promotion slightly altered the surface acidic properties of the catalysts. An incorporation of Pt and Pd metals resulted in close Bronsted to Lewis acidic sites that of neat tungstated zirconia (Table 4, entries 2, 4 & 5). The BL yield obtained for these catalysts also in the close range from 13.5 to 14.42 mol% (Table 4, entries 2, 4 & 5). The Ni and Cu upgradation remarkably increased the intensity of the Lewis acidity compared to other novel metals (Table 4, entries 7 & 8). This change probably surmised to be the interaction between metals (Ni, Cu) and ZrO2 surface, thereby revoking the Zr+4O2- activation, which induces the Lewis acidity and simultaneously making unavailability of electrons for polytungstate to develop the Bronsted acidic sites. The modification with Fe and Co metals also induced in the hike of Lewis acidity compared to the neat and noble (Pt, Pd) metal incorporation (Table 4, entries 6,9). Despite all these structural and physicochemical alterations by metal incorporation of the tungstated catalysts, the yield of BL was not improved significantly. Thus, making this catalyst low selective for the butanolysis of FAL. Fig. 5 and Table 5 . represent the Pyridine-FTIR spectrum and the normalized peak areas corresponding to the acidic sites present on the catalyst, as well as the amount of acidity possessed by the catalyst. The strong Bronsted acidic band was due to SO3H groups and some other functional groups such as phenolic and carboxylic acid groups present on the carbon black represented by the FTIR analysis. The incorporation of -SO3H groups onto carbon black enhanced the Lewis acid groups and combination of Lewis + Bronsted acidic groups (Table 5, entry 2). The reason being that the sulfonation oxidizes the functional groups present on the carbon network, especially carboxyl and methoxyl groups (Bands at 1703 cm−1 & 2930 cm−1 in Fig. 3). The decrease in the spent catalyst's acidic nature was ascribed to the detachment of -SO3H groups that are weakly bonded to polycyclic aromatic carbon network [30,54]. Konwar et al. reviewed the biodiesel production using various carbon-based catalysts, in which the sulfonated carbon-based catalysts exhibited the leaching of -SO3H groups during the reaction [27]. Due to the surface and textural properties of the carbon catalysts observed from the BET results, the pyridine FTIR spectrums were different compared to tungstated zirconia catalysts. Table 5. Compares the presence of Lewis and Bronsted acidic sites of the carbon catalysts. For the spent catalyst (Table 5, entry 3), the Lewis acidic sites were reduced. This was surmised to be the detachment of weakly bonded functional groups caused by the affinity between the reactants and the hydroxyl groups of the carbon network [55]. The total surface acidity values of the catalysts were measured by the acid-base back titration (Table 5, entry 1,2,3) and are in alignment with the reported literature [30,34,53]. The carbon black showed an acidity of 1.12 mmol/g composed of the multiple surface functional groups. The sulfonated catalyst enhanced by SO3H groups, which were building blocks for the strong acidity, displayed 2.357 mmol/g of surface acidity. The spent catalyst after 3 cycles of use and the leaching of the surface functional groups possessed 1.658 mmol/g of acidity contributed by the strong -SO3H groups. The acidity and activity for butanolysis of FAL were in the order of SO3H/C > SO3H/C spent > Carbon black, which complied with the pyridine FTIR results.The butanolysis of FAL was carried out using the tungstated zirconia and sulfonated carbon catalysts. Scheme.1 represents the conversion of FAL to BL through the reaction intermediate 2-BMF over the solid acid catalysts [7,56,57]. The hydroxyl groups of FAL molecule protonated by the catalyst and then attack of 1-butanol to this conjugated FAL molecule to form the reaction intermediate was the initial step of butanolysis of FAL. The reaction intermediate was identified through the GC–MS (GCMS-QP2020 NX) analysis. Further conversion of 2-BMF to BL was a prolonged step, which can be regarded as the rate-determining step of butanolysis of FAL [6,8,58]. The polymerization of FAL molecules in the acidic media is a significant concern to use the lower initial BtOH: FAL ratios and the dehydration of n-butanol to the di-butyl ether was non consuming FAL byproduct of this reaction system. Table 6 shows the activity of the tungstated zirconia catalysts for FAL butanolysis, which indicates the selectivity and yield of 2-BMF & BL. The metal oxide loading and calcination temperature parameters were applied for this catalyst with 10&15 wt.% of WO3, and 800 °C &900 °C calcination temperature. An increase in the WO3 loading enhanced its activity in terms of the FAL conversion from 77% to 95 mol% with 5.5 % to 14 mol% BL yield (Table 6, entries 1,2). As discussed in the NH3-TPD analysis (Fig. 2a, Table 3, entries 1,2), the increase in the acidity of the catalyst enhanced the FAL conversion and the BL yield. Further increase in the calcination temperature for 15 wt.% from 800 to 900 °C resulted in the collapse of the pore structure with widened pores and drop in the acidity caused in decreasing the catalytic activity to 50.5 mol% FAL conversion (Table 6, entry 3). The catalyst activity for this reaction was minimal as only around 14.43 mol% of BL yield was observed after 6 hours of the reaction (Table 6, entries 2). In contrast, the same catalyst showed excellent catalytic activity for other reactions mentioned in the introduction section. The same catalyst was tested for dehydration 1-butanol resulted in 72.5 mol% 1-butanol conversion with 35 mol% of di-butyl ether yield after 1 hour of the reaction (data not shown). The incorporation of metals onto the tungstated zirconia catalyst enhanced its catalytic activity for various reactions [22,59,60]. The promotion with noble and novel metals such as Pt, Pd & Ni, Fe, Cu, and Co also resulted in the similar catalytic activity of fresh catalyst (Table 6, entries: 4–9). The acidity of the catalysts slightly changed and more or less remained in the order of neat 15WZr800 catalyst (Table 3, entry 3). The FAL conversion has reached a maximum for all the catalysts. In contrast, the BL yield was varied in the range of 10–14 mol% with a 25–32% yield range of 2-BMF, indicating that the FAL molecules polymerization and unconverted intermediates were progressed. The reaction temperature was also studied for this catalyst in the range of 130–240 °C with an interval of 20 °C, and a maximum of 28 mol% BL yield was observed at 240 ⁰C after 2 hours of the reaction (SFig. 1). The reaction was also studied with a high initial BtOH to FAL mole ratio, such as 60:1, which resulted in only 11.2 mol% of BL yield with 26 mol% of 2-BMF yield (Table 6, entry10). Due to the lack of sufficient acidity (0.201 mmol/g) of the catalyst, the conversion of 2-BMF to BL was not progressed to obtain higher yields.To study the effect of reaction temperature, the butanolysis reaction was carried in the range of 130–210 °C. Fig. 6 . explains the impact of reaction temperature on FAL butanolysis for sulfonated carbon catalyst on the selectivity and yield of reaction products. All the previous studies reported that the reaction temperature favors the butanolysis reaction by converting the reaction intermediate 2-BMF to BL via furan ring-opening mechanism [6,8,12,15]. The FAL oligomers tend to form FAL polymers in acidic conditions at higher reaction temperatures, which is the primary concern for high-temperature reactions. A reaction with neat FAL in the tetralin solvent at 190 °C is also conducted to understand the formation of FAL polymers in acidic conditions. The GC-FID chromatogram of neat FAL reaction was shown in the Supporting information (SFig. 5). However, Milan et al. reported that the activation energies for forming FAL polymers are identical to that of the primary reaction concluding that high reaction temperatures can be favorable under optimized conditions [61]. At the reaction temperature of 130 °C, though the FAL conversion was near completion, the yield of BL was 49 mol%, accompanied by the unconverted reaction intermediates after 6 hours of the reaction. Further increase in the temperature to 150 °C, there was a slight increase in the BL yield with a reduced yield of others, including FAL polymers. As the reaction temperature progressed, there was a significant change in the BL yield, indicating that higher reaction temperatures are needed for this catalyst system to convert the reaction intermediate 2-BMF to BL. Almost a similar BL yield of around 80 mol% was observed for 190 °C & 210 °C, indicating that the reaction temperature reached its threshold value, thus optimizing the reaction temperature at 190 °C. The selectivity of the BL was steadily increasing with reaction temperature, whereas the 2-BMF selectivity was gradually decreased, and there was no 2-BMF at 210 °C left to convert to BL. The yield of others, including FAL polymers, was on a declining trend with the reaction temperature and kept constant for 190 °C and 210 °C temperatures. Designating that higher reaction temperatures favor the product formation. The total surface acidity of the sulfonated carbon catalyst was way higher (2.357 mmol/g) than the 15WZr800 (0.201 mmol/g) catalyst, which facilitated the transformation of 2-BMF to BL towards higher yields.The butanolysis of FAL was studied at 190 °C with sulfonated carbon catalyst for the reaction time profile with 8.5:1 butanol to FAL mole ratio. Fig. 7 . represents selectivity and yield of BL along with the yield of 2-BMF as a function of time. FAL conversion was proceeded rapidly and almost completely converted after 1 hour while 54 mol% BL yield and 26 mol% of 2-BMF were observed. The temperature has a significant effect on this reaction, which accelerated the conversion of 2-BMF to BL via furan ring-opening (Fig. 6). As the time progressed, the BL yield increased sharply to 82 mol% after 6 hours. The selectivity of the BL also kept on increasing with time and reached a maximum of 98 mol% after 6 hours. Meanwhile, the yield of 2-BMF was decreased over time and completely converted to BL after 6 hours. However, after the complete conversion of 2-BMF, the BL yield has slightly reduced to 80 mol% at 7 hours, which implies the beginning of the BL degradation. Though the complete conversion of FAL was observed after one hour, the formation of BL progressed up on time to achieve the maximum value and settled at 6 hours, thereby optimizing the reaction time to 6 hours.As described in the introduction, most of the studies for butanolysis of FAL were reported at a higher initial molar ratio of butanol to FAL (>30) with higher selectivities. This study intended to reduce the initial mole ratio to a maximum achievable range. The effect of initial FAL concentration was studied in the range of 30:1 to 4:1 at 190 ⁰C for 6 hours using sulfonated carbon catalyst, and results are reported in Fig. 8 . The results depicted that increase in the initial FAL concentration resulted in the decreasing trend of BL yield, and increasing the FAL polymers yield replicates that the formation of FAL polymers was favored upon high FAL concentration. The yield of BL reduced from 84 mol% to 57 mol% from 30:1 to 4:1 initial molar ratio of BtOH to FAL. The complete conversion was obtained even at lower mole ratios because of the high activity of the catalyst, but the respective yields of BL were decreased. Almost no intermediates are left to convert at such high temperatures, as evident from the reaction temperature optimization. Moreover, a decrease in the BL yield was obvious because of the fact that unavailability of FAL conjugates due to FAL polymerization. Higher initial concentrations of FAL led to the polymerization under acidic conditions, thereby decreasing the product formation [11–13]. A maximum of 80 mol% of BL yield was achieved at a much lower initial molar ratio of 8.5:1, whereas for 6:1 initial molar ratio, the BL yield was limited to 68 mol% only. The yield of FAL polymers inclined with increasing initial FAL concentration and relatively increased to 33 mol% for 4:1 initial molar ratio. Therefore, feasible and economic studies are needed to find suitable initial conditions for this reaction to obtain maximum yields of BL. Fig. 9 . exemplifies the variation of BL yield with respect to the amount of catalyst loaded at a reaction temperature of 190 °C for 6 hours. The reaction was first performed without the catalyst, which resulted in no conversion of FAL, indicating that a minimum amount of a catalyst is needed to enhance the FAL conversion. Starting with a 4.5 wt.% of catalyst loading (0.25 g), 94% of FAL conversion was achieved with most of the unconverted reaction intermediates (around 20 mol% of 2-BMF) left to convert to BL. With an increase in the catalyst amount from 0.25 to 0.5 g, the complete conversion of FAL was achieved along with a significant increase in BL yield from 57 to 80 mol%. An increase in the catalyst amount provided more acidity in the reaction, which boosted the FAL conversion and transformed the intermediate 2-BMF to BL completely. Upon further increase in the catalyst loading to 0.75 g, a slight decrease in the BL yield, around 77 mol%, was observed. The yield of FAL polymers increased from 15% to 23 mol% with the catalyst amount. This was bounded to happen in acidic conditions at higher temperatures and probably caused by the presence of the more active sites, which promoted the initial conversion of FAL molecules to polymerize, thereby increasing the yield of FAL polymers [61]. Thus, a minimum of 9 wt.% of catalyst loading was sufficient for the complete conversion of FAL with a maximum of 80 mol% BL yield.The sulfonated carbon catalyst activity was compared with some typical previously reported solid acid catalysts, metal salts such as AlCl3 and dilute H2SO4. The results are summarized in Table 7 . The reaction was performed at 190 ⁰C using a very low concentrated H2SO4 to mild H2SO4 (1 M) to study the effect of acid concentration (SFig. 3). The reaction with AlCl3.6H2O at 123 °C yielded 92 mol% of BL at a very high butanol to FAL mole ratio (Table 7, entry 1). This study depicts that the BL formation not only correlated to Bronsted acidic sites of Al salts but also with the Lewis acidic sites from Al3+ ions. An optimal combination of both Bronstedic and Lewis sites would proceed with a unique selectivity towards the BL. In the present work, the reaction was performed with anhydrous AlCl3 at 190 °C for 2 hours at a lower initial mole ratio of 8.5:1, and 73 mol% of BL yield was observed with 6 mol% of 2-BMF yield (Table 7, entry 8). The reaction was carried out only for 2 hours because of the fact that chloride ion corrodes the reactor at such high temperature and might affect the reaction results [62–64]. Alumina supported SBA-15 catalyst showed 91% of BL yield for a period of 6 hours at 180 ⁰C but at the initial mole ratio of 65 remarking the FAL polymerization to 9% at high temperature even at a high initial mole ratio of butanol to FAL (Table 7, entry2). Waste paper derived magnetic carbanaceous (SMWP) catalyst also showed 91 % BL yield at a reduced initial mole ratio of 41 with consistent activity (Table 7, entry 3). Sulfonated SBA-15 catalyst at a much reduced initial molar ratio of 16 gave 63 % of BL after 4 hours but at a much lower temperature of 100 °C (Table 7, entry 4). The Amberlyst 39 catalyst displayed 63 % BL yield at 110 °C after 6 hours (Table 7, entry 5). The authors have mentioned that unconverted intermediates such as 2-BMF and 4,5,5-tributoxy-2-pentanone at low-temperature reaction caused the 63% yield of BL. Moreover, the ionic resin Amberlyst catalysts are thermally unstable above 120 °C. The reaction with 1 M H2SO4 showed a maximum yield of 85% after 6 hours of reaction (Table 7, entry 7).The carbon black showed the optimal activity for the reaction with 34 mol% of BL yield (Fig. 10 ). The selectivity and the yield of both 2-BMF & BL were identical to each other and settled at 48 mol% selectivity and 34 mol% yield. The carboxylic acid and phenolic groups formed during the incomplete carbonization of sucrose were responsible for the carbon black's optimal activity. The fig. illustrates the reaction results with the carbon black. Upon the results with 0.5 g of catalyst, the loading was increased to 1 g, which resulted in a similar outcome. The product distribution indicated that a maximum of 0.5 g of catalyst loading was enough to convert the FAL to the products.The deactivation study (Fig. 11 ) for the sulfonated carbon catalyst was performed for three recycles. After every reaction, the catalyst was recovered by vacuum filtration and dried for 12 hours at 120 °C for the next use. The heterogenity of the reaction was reported in SFig. 9. which indicated a slight leaching of the -SO3H groups. Moreover, based on the heterogeneity test results, the reaction was mainly catalysed by the acid sites on thesurface of sulfonated carbon catalyst instead of the leached -SO3H groups.. After three recycles, the BL yield was reduced to 49 mol%, indicating that the -SO3H groups bonded to weakly functional groups of the carbon network were desorbed during the reaction. Though the FAL was fully converted after 6 hours for all the recycles, the 2-BMF left unconverted to BL, attributing the need for acidity for the conversion. This problem can be overcome by regenerating the activity of the catalyst by sulfonation. The catalyst after 3 cycles was regenerated by sulfonation at 80 °C under inert conditions, and it replicated the results as that of the fresh catalyst with 78 mol% of BL yield compared to 80 mol% with fresh catalyst.In conclusion, the butanolysis of FAL at a lower initial mole ratio was conducted using two different catalysts. The tungstated zirconia catalyst and the metal promoted catalysts resulted in a maximum of 28 mol% of BL yield, signifying the need for high catalytic activity for butanolysis reaction. The sulfonated carbon catalyst (surface acidity 2.357 mmol/g) resulting in 80 mol% of BL yield showed superior activity caused by the strong Bronstedic –SO3H groups and the aided acidic carbon groups. The partially carbonized sucrose to carbon black showed better catalytic activity than the tungstated zirconia catalyst caused by the presence of phenolic, hydroxyl, and carboxylic acid functional groups. Thus, this work demonstrated that the high BL yields as high as more than 80 mol% with sulfonated carbon catalyst even at low butanol: FAL ratio as low as 8.5. Further decreasing the initial mole ratio resulted in decreasing the BL yield accompanied by FAL polymerization. The deactivation study reveals that the weakly bonded –SO3H groups were detached from the carbon network resulted in the catalytic activity to 49 mol% of BL yield after 3 recycles with 100% FAL conversion and regained the activity to 78 mol% BL yield upon regeneration.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This work was supported in part by Japan Science and Technology Agency Strategic International Collaborative Research Program (JST SICORP) Grant Number JPMJSC18H1, Japan. U.R. Thuppati acknowledges the financial support by JICA IITH-FRIENDSHIP (D1956755) scholarship for suppoeting this studySupplementary data to this article can be found online at https://doi.org/10.1016/j.crcon.2021.03.003.The following are the Supplementary data to this article: Supplementary Data 1

This work presents the formation of butyl levulinate, a potential fuel additive, and an excellent renewable chemical obtained by the butanolysis of furfuryl alcohol (FAL) over a solid acid catalyst. The butanolysis of furfuryl alcohol reaction is a strong function of acidity for which tungstated zirconia (WO3-ZrO2), a robust solid acid catalyst, and a sulfonated carbon catalyst were employed to produce high yields of butyl levulinate targeting a lower initial molar ratio of butanol to FAL. A maximum of 28 mol% yield of butyl levulinate was obtained with tungstated zirconia catalyst. Easily prepared sulfonated carbon catalyst at high reaction temperatures facilitated the complete conversion of reaction intermediate, 2-butoxymethylfuran (2-BMF) through which butyl levulinate was formed, and as high as 80 mol% of butyl levulinate yield was produced at an initial mole ratio of 8.5:1 of butanol to FAL. The better results of sulfonated carbon catalyst could be attributed to the presence of -SO3H, carboxylic acid, and phenolic OH groups on the carbon surface.

The authors are unable or have chosen not to specify which data has been used.Cobalt is widely used in metallurgy, mainly in the manufacturing of alloys which are corrosion and wear resistant while cobalt superalloys are also heat resistant [1]. Moreover, cobalt is also used to make magnets and high-speed tool steels. In non-metallurgical applications, cobalt is used as a catalyst in the petroleum and chemical industries, in batteries, and as a component of drying agents for various paints and inks [1]. Thus, high purity cobalt is required in most of its industrial applications and an effective cobalt extraction process is essential [2]. However, Co(II) and Ni(II) are often present in ores and their separation is challenging since both metals have very similar physicochemical properties due to their adjacent positions in the periodic table [3].Solvent extraction (SX) is the method often chosen by industry for Co(II) and Ni(II) separation due to its high separation efficiency [3]. Acidic organophosphorus extractants such as di-(2-ethylhexyl)phosphoric acid (D2EHPA) [4], 2-ethylhexyl phosphonic acid mono-2-ethylhexyl ester (PC 88A) [5], and bis(2,4,4-trimethylpentyl) phosphinic acid (Cyanex 272) [6] have been applied to the SX of Co(II). However, D2EHPA exhibits poor separation of Co(II) from Ni(II). PC 88A offers a better selectivity over D2EHPA, although both these extractants also extract Ca(II). Cyanex 272, on the other hand, provides good separation of Co(II) from both Ni(II) and Ca(II) [6–8]. Other extractants for Co(II), namely bis(2,4,4-trimethylpentyl)dithiophosphinic acid (Cyanex 301) and bis(2,4,4-trimethylpentyl)monothiophosphinic acid (Cyanex 302), have been reported to offer even better Co(II) and Ni(II) separation than Cyanex 272 [9].However, all the extractants mentioned above are acidic, and thus the extraction process is strongly dependent on the pH of the aqueous phase. Over the past few years, ionic liquids (ILs) have been found to be promising extractants in SX [10].ILs are salts composed of a cation and an anion, which exist as liquids even at low temperatures, including room temperature. ILs have several advantageous properties over common organic solvents, such as negligible vapour pressure, good thermal stability and high intrinsic conductivity [10]. ILs are also known as designer solvents since their properties can be tuned to particular applications (e.g., ILs can be tuned to have low solubility in water, thus limiting their loss to the aqueous phase during the extraction process [11]). Due to these features, ILs have been widely studied as extractants for metal ions. Trialkylmethylammonium chloride (the main component of Aliquat 336) is an example of a quaternary alkylammonium-based IL, which has been studied for the Co(II) and Ni(II) separation.Co(II) and Ni(II) have different coordination preferences in aqueous media. In concentrated electrolyte solutions, Co(II) exhibits a tendency to form tetrahedral complexes which are less pronounced with Ni(II) [3]. Co(II) and Ni(II) separation with Aliquat 336 is based on this difference, i.e., Co(II) is extracted by anion exchange as an anionic tetrahedral chlorocomplex, either as the tetrachlorocobaltate(II) ion (CoCl4 2–) or the hydrogen tetrachlorocobaltate(II) ion (HCoCl4 –) depending on the solution pH [12,13]. Quaternary phosphonium-based ILs, such as trihexyl(tetradecyl)phosphonium chloride ([P66614][Cl]) and trihexyl(tetradecyl)phosphonium bis(2,4,4-trimethylpentyl)phosphinate ([P66614][C272]), are also able to separate Co(II) from Ni(II). The Co(II) extraction mechanism using [P66614][Cl] or [P66614][C272] in the presence of a high concentration of HCl or NaCl is the same as that involving Aliquat 336 (i.e., ion-exchange extraction) [14–17]. However, when using [P66614][C272] in the absence of HCl, it is suggested that Co(II) forms a neutral complex with the IL's phosphinate anion and its counter anion is extracted by the IL's cation to preserve solution electroneutrality [14]. It has also been reported that [P66614][Cl] can achieve high separation of Co(II) from Ni(II) in sulfate media via an anion-exchange mechanism as described by Onghena et al. [18]. In this case it is still the tetrachlorocobaltate(II) anion that is extracted, although the extraction process is more complex since it only occurs in the presence of sulfuric acid in the aqueous phase. It is suggested that the HSO4 − ion firstly replaces the chloride ion in the IL which then complexes with Co(II) in the aqueous phase to form the tetrachlorocobaltate(II) anion. This anion is then transferred to the organic phase by anion-exchange with the HSO4 − ion of the IL in its HSO4 form. However, in most cases, the extraction of Co(II) from chloride solutions involving anion-exchange requires the use of hydrochloric acid and/or lithium chloride solutions with concentrations as high as 7 M [19,20] which can present some difficulties for large scale industrial applications.Larsson and Binnemans [21] described the conversion of [P66614][Cl] into a form which contains a lipophilic and metal complexing anion, such as thiocyanate, rather than the more hydrophilic chloride anion (i.e., [P66614][SCN]). This IL can then be used to extract a target cation from the aqueous phase by direct complexation rather than by the formation of a complex anion in the aqueous phase first as in the conventional extraction by anion-exchange. Hence, the anion of such ILs can strongly complex with the target metal cation from the aqueous phase to form an anionic complex, while the cation of the IL forms two ion-pairs: one with the newly formed and extracted into the organic phase anionic metal complex, and the other with the counter anion of the target metal ion (which is only extracted into the organic phase in order to preserve electroneutrality in both phases). Larsson and Binnemans [21] have termed this process as ‘split-anion’ extraction. We are of the opinion that the term ‘bifurcated extraction’ better depicts this process, and it will thus be used throughout this manuscript.The bifurcated extraction mechanism using [P66614][SCN] has not only been used for the separation of Co(II) from Ni(II) in sulfate media [18], but also for the separation of transition metal ions (including Co(II)) from rare earth ions in aqueous nitrate or chloride solutions [22]. This extraction mechanism has also been explored with quaternary ammonium-based ILs with different anions for the separation of cobalt from samarium [23]. However, the stoichiometry proposed in the above-mentioned cases has not been determined experimentally.Since bifurcated extraction does not require a high chloride concentration in the aqueous phase and is independent of the aqueous solution pH, it has the potential to provide an efficient way for the separation of Co(II) and Ni(II) in industry. Thus, this paper describes a detailed study of the extraction process associated with the use of [P66614][C272] and [P66614][SCN], both dissolved in toluene, for extracting and separating Co(II) without the need for high concentrations of chloride in the aqueous phase or adjustment of the pH.All the chemicals used in this study were AR grade unless stated otherwise. [P66614][Cl] (commercially known as Cyphos® IL 101, >95.0%, Aldrich) and [P66614][C272], also known as Cyphos® IL 104 (>95.0%, Strem Chemicals), were used in toluene (99.5%, Ajax) as the diluent. Solutions for the SX studies were prepared from Co(II) stock solutions in deionised water (≥18.2 MΩ cm, Synergy 185, Millipore) using, CoCl2·6H2O (99.7%, J.T. Baker), Co(SCN)2 (99.9%, Aldrich), Co(NO3)2·6H2O (BDH), and CoSO4·7H2O (BDH). KSCN (VWR Chemicals) was used to convert [P66614][Cl] to its SCN− form. Deionised water, 0.5 M HCl (32 wt%, Ajax), 0.5 M HNO3 (70 wt%, Ajax) and ethylenediaminetetraacetic acid disodium salt (EDTA) (Chem-Supply) were used for the preparation of aqueous solutions for back-extraction. The pH of the EDTA solution was adjusted to 7–8 by the addition of NaOH pellets (Chem-Supply). The pH was measured using a smart Chem-Lab Multi-Parameter Laboratory Analyser (TPS).For the study of the extraction of other metal ions, the following chemicals were used to prepare stock solutions containing 1000 mg L−1 of the associated cations: NaCl, NaNO3, CaCl2·2H2O, Ca(NO3)2·4H2O (all from Chem-Supply), Ni(NO3)2·6H2O, CdCl2·2.5H2O, CuCl2·2H2O, Cu(NO3)2·2.5H2O (all Ajax), Cd(NO3)2·4H2O, MgCl2·6H2O, Mg(NO3)2·6H2O (all BDH), Zn(NO3)2·6H2O (98%), NiCl2·6H2O (both Sigma-Aldrich), ZnCl2 (UniLab). The above-mentioned salts were dissolved in deionised water.Acidic impurities in [P66614][Cl] were determined by potentiometric titration in ethanol solution using 0.01 M NaOH and monitoring the pH with a Chem-Lab Multi-Parameter Laboratory Analyser (TPS).Co(II) concentrations in aqueous solutions were determined by atomic absorption spectrometry (AAS) (Z-2000 Series Polarized Zeeman AAS, Hitachi) using the following conditions: acetylene flow − 1.8 L min−1, acetylene pressure − 160 kPa and air flow − 15.0 L min−1, hollow cathode lamp (Hitachi) current and wavelength – 15 mA and 240.7 nm, respectively.Other metal ion concentrations were determined by inductively coupled plasma optical emission spectrometry (ICP-OES, Optima 4300 DV, Perkin Elmer) using the following conditions: RF power − 1300 W, plasma flow − 15.0L min−1, auxiliary flow − 0.2L min−1, nebuliser gas flow − 0.7L min−1.UV–visible spectroscopic studies of the organic phases (Libra S12, Biochrom) were carried out using toluene as the reference.Fourier transform infrared measurements (FTIR, Tensor 27 IR, Bruker) were conducted on KBr crystal disks to examine the [P66614][SCN] organic phase before and after extracting Co(II) from a 20 g L−1 Co(II) solution as well as to assess the presence of water in the organic phases containing [P66614][C272] or [P66614][SCN] loaded with Co(II).Electrospray ionisation mass spectrometry (ESI-MS) (Agilent 6520 Quadrupole Time of Flight (Q-ToF) mass spectrometer, USA, coupled to an Agilent 1100 autosampler) was used in the negative mode under the following conditions: drying gas flow rate, 7L min−1; nebuliser pressure, 40 psi; drying gas temperature, 300 ˚C; capillary voltage, 4000 V; skimmer voltage, 65 V; scan range acquired, 50–500 m/z. Each analysis involved the injection of dissolved in acetonitrile (HPLC grade, Sigma, USA) 0.005 M [P66614][C272] samples (1 μL), collected before and after extraction of Co(NO3)2, into the carrier solvent stream of 70:30 (v/v) acetonitrile (HPLC grade, Sigma, USA): 0.1% formic acid (HPLC grade, Sigma, USA), flowing at 0.3 mL min−1.The procedure used to convert [P66614][Cl] to its thiocyanate form was similar to that described by Rout and Binnemans [22]. This was carried out by shaking 100 mL of 0.1 M [P66614][Cl] dissolved in toluene with 100 mL of 3 M KSCN solution in a separation funnel for 5 min. After separation, an orange-coloured organic phase was obtained, which was washed twice with 100 mL of 3 M KSCN and four times with 100 mL of deionised water. After each washing with deionised water, the orange colour diminished to yield a faint yellow organic phase.SX studies using both [P66614][C272] and [P66614][SCN] were carried out by shaking 20 mL of 0.1 M [P66614][C272] or [P66614][SCN] in toluene with 20 mL of 100 mg L−1 cobalt(II) chloride, thiocyanate, nitrate or sulfate in 125 mL glass jars for 1 h using an orbital shaker set to 200 rpm (Platform Mixer OM06, Ratek) or by mixing with a magnetic stirrer set to 750 rpm (magnetic multistirrer, VELP Scientifica) for 2 h. After phase separation, 0.5 mL of the aqueous phase was removed, diluted 20 times and the Co(II) concentration was determined by AAS.The organic phase (15 mL) after the Co(II) extraction experiments was back-extracted using 15 mL of deionised water, 0.1 M EDTA, 0.2 M EDTA, 0.5 M HNO3 (for [P66614][C272]) or 0.5 M HCl (for [P66614][SCN]) with contact times of 1 h for [P66614][C272] and 3 h for [P66614][SCN]. Solutions were stirred using a magnetic stirrer and, after phase separation, 0.5 mL of the aqueous phase was withdrawn for Co(II) determination.The percentage of extraction (%E) and back-extraction (%BE) were calculated using Eqs. (3) and (4), respectively: (3) % E = M IF - M FF M IF × 100 (4) % B E = M FR M IF - M FF × 100 where M IF is the initial concentration of metal in the aqueous (feed) solution, M FF is the concentration of metal in the aqueous solution after extraction, and M FR is the concentration of metal in the back-extraction solution.In order to investigate the stoichiometry for the extraction of Co(II) by [P66614][C272] and [P66614][SCN], a 20-mL aliquot of 0.1 M of extractant in toluene was mixed using a magnetic stirrer with 20 mL of increasingly high concentrations (up to 20,000 mg L−1) of a Co(II) solution (nitrate for [P66614][C272] and chloride for [P66614][SCN]) for 3 h for each concentration until a plateau was reached to signify that the organic phase had been fully loaded with Co(II). After phase separation, 15 mL of the organic phase were collected and back-extracted with an equal volume of 0.5 M HNO3 for [P66614][C272] and 0.5 M EDTA for [P66614][SCN]. Complete back-extraction was achieved in 5 h for [P66614][C272] and 3–4 days for [P66614][SCN]. The back-extraction solutions were diluted with deionised water and the Co(II) concentration was determined by AAS.For [P66614][C272] only, an extraction experiment with a lower concentration of extractant (0.005 M) was also conducted in order to be able to analyse the Co(II) concentration directly in the aqueous feed phase. Aqueous feed solutions with increasing concentrations of Co(II) of up to 600 mg L−1 (with nitrate as the counter ion) were used, following the same procedure as described above.Studies were carried out using the same extraction procedure as described in Section 2.4 with aqueous feed solutions containing a mixture of Co(II), Ni(II), Cd(II), Cu(II), Zn(II), Na(I), Mg(II) and Ca(II) (1.7 mM each).As mentioned in Section 2.3, an orange colour appeared in the toluene organic phase during the conversion of [P66614][Cl] to its thiocyanate form (i.e., [P66614][SCN]). The UV–visible spectrum of the organic phase is shown in Fig. 1 along with that of the original [P66614][Cl].There is an absorption peak at 480 nm for [P66614][SCN] which is absent in the spectrum of the original [P66614][Cl]. This peak is characteristic for the [FeSCN]2+ species [24] and demonstrates that [P66614][Cl], as received from the supplier, contained a small amount of Fe3+ as an impurity. This was confirmed by the fact that the spectrum of [P66614][Cl] showed a peak at 360 nm, characteristic of the FeCl4 − species [25].Another important observation made when initially using [P66614][SCN] to extract Co(II) from neutral solutions was the fact that the pH of the aqueous solution decreased unexpectedly by approximately 3 pH units after extraction (from pH 5.1 ± 0.2 to 2.2 ± 0.1, n = 3). Bradaric et al. [26] have described the synthesis of trihexyl(tetradecyl)phosphonium chloride using trihexylphosphine and 1-chlorotetradecane as starting materials yielding 98 wt% purity, of which 93.9% corresponded to the IL, and the remaining were impurities (e.g., 4.4% trihexylphosphonium hydrochloride, 0.3% HCl, <0.3% tetradecane isomers, <0.7% secondary alkylphosphines). Potentiometric titration of commercial [P66614][Cl] (Section 2.2) with a standard solution of NaOH confirmed that the commercial product, as supplied, contained 58.4 mmol L−1 of a monobasic acid (presumably HCl) which after dilution with toluene was sufficient to decrease the aqueous solution pH by 3 pH units as observed in the preliminary extraction experiments using unwashed IL. It is interesting to note that the hydrogen ion associated with trihexylphosphonium hydrochloride was not removed during the process of its conversion to the thiocyanate form but only when the thiocyanate form of the IL was used to extract Co(II). This is possibly due to the complexation of trihexylphosphine with CoCl2.Extraction experiments were initially conducted with 0.1 M [P66614][C272] or [P66614][SCN] in toluene as the organic phase, and cobalt(II) nitrate or cobalt(II) chloride in the aqueous phase, respectively (both 100 mg L−1 Co(II), Fig. S1, Supplementary Material). After the extraction, different aqueous phases were investigated for their ability to back-extract Co(II), namely deionised water and solutions of EDTA, and HNO3 for [P66614][C272], and deionised water and solutions of EDTA and HCl for [P66614][SCN]. For [P66614][C272], ≥ 90% back-extraction of Co(II) was achieved within 10 min using 0.2 M EDTA and 0.5 M HNO3 solutions, but only 20% back-extraction was achieved in 40 min with deionised water (Fig. S2, Supplementary Material). For [P66614][SCN], only the EDTA solution was able to back-extract Co(II) due to the high stability of the corresponding Co(II)-EDTA complex (log K Co(II)-EDTA = 16.26 [27]). Complete back-extraction was achieved in 1 h for 0.2 M and 2 h for 0.1 M EDTA solutions (Fig. S3, Supplementary Material). Back-extraction of Co(II) in the case of [P66614][SCN] was significantly slower than that for [P66614][C272] because of the high stability of the cobalt(II)-thiocyanate complex.The intense blue colour of the organic phase after Co(II) extraction for each of the extractants suggested tetrahedral coordination of the ligands around the Co(II) ion. This was confirmed on examination of the UV–visible spectrum of the organic phase as shown in Fig. 2 for each extractant.The spectrum of [P66614][C272] organic phase after extraction (Fig. 2A) shows three absorption peaks consistent with a Co(II) complex with asymmetric tetrahedral coordination [27,28]. Xun and Golding [29] obtained a similar spectrum for the Co(II)-Cyanex 272 complex, however, in their case the extractant existed in a dimeric form and so a direct comparison with [P66614][C272] should be made with caution, even though both extractants have the same anion.The spectrum of the [P66614][SCN] organic phase after extraction of Co(II) (Fig. 2B) shows a single absorbance peak at 625 nm. According to Bjerrum [30], this peak is characteristic of a tetrahedral tetrathiocyanatocobaltate(II) complex anion.In order to elucidate the Co(II) extraction stoichiometry in the cases of [P66614][C272] and [P66614][SCN], studies were carried out to fully load the organic phase with Co(II) as described in Section 2.5 to determine the mole ratio of IL to Co(II). The results obtained for [P66614][C272] (Fig. 3 ) indicated that on fully loading of the organic phase with Co(II), the mole ratio of [P66614][C272] to Co(II) was equal to 2.8:1 thus suggesting that 3 molecules of [P66614][C272] were required to extract one Co(II) ion. However, since the extracted Co(II) species had a tetrahedral configuration, it was virtually impossible to form a complex composed of one cobalt ion and three phosphinate anions. This experiment was repeated with a lower concentration of IL in the organic phase (where the Co(II) concentration after extraction was measured in the feed solution, with no need to do back-extraction) and the same ratio was obtained (Fig. 3). Moreover, it should be noted that the pH before and after solvent extraction was not significantly different (e.g., 6.0 vs 6.3, respectively). It has been reported for the extraction of Co(II) using Cyanex 272 by SX that oligomeric species with 2 or even 3 cobalt centres are formed at >42% Co(II) loading of the organic phase [31,32]. Since the ratio [P66614][C272]/Co(II) determined here experimentally was 2.8 and the organic phase was fully saturated, it was suggested that an oligomeric species, consisting of 3 cobalt centres and 8 phosphinate anions, similar to that proposed by Best et al. for Cyanex 272 [32], was present. Due to the formation of such oligomeric species, the viscosity of the organic phase increased and for that reason longer equilibration times were used to guarantee that equilibrium was reached.The extraction mechanism is thus described by Eq. (1) (Table 1 ) which is clearly an example of bifurcated extraction in which the nitrate ion (presence in the organic phase confirmed by ESI-MS in negative mode, Fig. S4, Supplementary Material) only serves the purpose of preserving electroneutrality and the complexing anion is that originally associated with the IL.It should be noted that Rybka and Regel-Rosocka [14] proposed an alternative stoichiometry (Eq. (2), Table 1) suggesting a 2:1 ratio of [P66614][C272] to Co(II). However, this was not based on the experimentally determined mole ratio of IL to Co(II) and required each phosphinate anion to act as a bidentate ligand in order to explain the tetrahedral structure of the complex. As discussed by Carson et al. [31] the phosphinate ligands are insufficiently flexible to allow for O-Co-O angles to be close to the ideal tetrahedral angle when acting in their bidentate coordination mode. The asymmetric tetrahedral structure, as evidenced by the UV–Visible spectrum of the Co(II)/[P66614][C272] complex, shown in Fig. 2A, is consistent with the proposed oligomeric complex.Co(II) back-extraction in the present work with 0.5 M HNO3 solution can be described by Eq. (3) (Table 1). In this back-extraction stoichiometry it is suggested that the H+ ion protonates the bis(2,4,4-trimethylpentyl)phosphinate anion thus forming Cyanex 272 (i.e., bis(2,4,4-trimethylpentyl)phosphinic acid) [33]. This means that the IL does not fully return to its original form. However, washing the back-extracted organic phase with a low concentration of sodium hydroxide solution is expected to allow the complete regeneration of [P66614][C272], and consequently its reusability [33]. On the other hand, back-extraction with EDTA is expected to return the IL to its original form as described by Eq. (4) (Table 1).The results of the Co(II) extraction stoichiometry study in the case of [P66614][SCN] are shown in Fig. 4 . It should be noted that longer equilibration times were implemented to guarantee that equilibrium was reached even in the presence of very high Co(II) concentrations (Fig. S5, Supplementary Material). Moreover, no significant pH change was observed before and after the solvent extraction (e.g., pH 5.3 vs 5.0, respectively). The results obtained suggest a [P66614][SCN] to Co(II) mole ratio of 2.6:1.An extraction stoichiometry of 2.6:1 is in contrast to the 4:1 stoichiometry suggested by Rout and Binnemans [22] as discussed earlier, however, their proposal was not based on an exhaustive study of the stoichiometry as presented above. An extraction stoichiometry that is in agreement with a [P66614][SCN] to Co(II) mole ratio of 2.6:1 can be described by Eq. (5) (Table 2 ) where the mole ratio is 2.5:1.In order to explain the tetrahedral coordination associated with the Co(II) ion and the 2.5:1 stoichiometry, we propose a structure in which two Co(II) ions are bridged by a thiocyanate ion as shown in Fig. 5 .Support for this structure arises from the fact that the thiocyanate anion has a linear structure and is ambidentate since both the S and N atoms can act as electron pair donors in the formation of transition metal complexes. Transition metal complexes are known, where the thiocyanate anion acts as a bridging group, such as that suggested in Fig. 5, which can be confirmed by FTIR [34,35]. Hence, the organic phase containing [P66614][SCN] was analysed by FTIR before and after saturation with Co(II), and the results are presented in Fig. 6 .The spectral band at 2052 cm−1 observed for the organic phase before extraction corresponds to the thiocyanate anion of the IL (CN stretching frequency), which shifts to 2071 cm−1 with a shoulder at 2088 cm−1 after loading the organic phase with Co(II). It has been reported that, pronounced υC–N bands within the 2020–2096 cm−1 range indicate the presence of M−SCN−M coordination of the thiocyanate group (M standing for a metal ion). Moreover, the shoulder is indicative of the presence of SCN− ions in bridging and terminal positions [35]. Hence, the FTIR spectra, shown in Fig. 6, provide the evidence required to support the structure presented in Fig. 5.Equation 7 (Table 2) represents the back-extraction of Co(II) with EDTA and, as for the case with [P66614][C272], EDTA returns the IL back to its original form.The presence of water in the organic phases containing [P66614][C272] or [P66614][SCN] loaded with Co(II) was also assessed by FTIR (Fig. S6, Supplementary Material). Within the range 3500–3300 cm−1, no broad peak was observed which indicated the absence of any O–H stretch vibration. Hence, it can be concluded that water was not present in the organic phases, thus having no influence on the extraction process.The results from the Co(II) SX studies using different aqueous phase counter anions in the cases of both [P66614][C272] and [P66614][SCN] are shown in Fig. 7 . The organic phase acquired an intense blue colour after extraction of each of the four counter anions and the extent of extraction was similar for both extractants. It can be seen that 100% of the Co(II) was extracted with thiocyanate and nitrate as the counter anions while lower degree of extraction was obtained when the counter anions were chloride and sulfate. The order of extraction clearly follows that predicted from the Hofmeister series [36] for anions in which the less hydrated, and hence more lipophilic anions (i.e., thiocyanate and nitrate), are more readily extracted than the more hydrated and less lipophilic anions (i.e., chloride and sulfate).The extraction of other metal ions using [P66614][C272] and [P66614][SCN] was studied for Ca(II), Mg(II), Na(I), Cd(II), Ni(II), Cu(II), or Zn(II) ions using their chloride or nitrate solutions and the results are presented in Figs. 8 and 9 .It can be seen in Fig. 8 that, except for Na(I), [P66614][C272] extracts all the other cations studied from their aqueous solutions, which is not surprising since the phosphinate anion of [P66614][C272] is a strong complexing agent, particularly for transition metal cations. It should be noted that the extraction percentage from nitrate solutions, with the exception of Cu(II), is higher than that from chloride solutions due to the higher lipophilicity of nitrate.In the case of [P66614][SCN], Fig. 9 shows that Mg(II), Ca(II) and Na(I) are not extracted because these cations do not form stable complexes with thiocyanate unlike Cd(II), Cu(II), Co(II), and Zn(II) which form stable complexes and are successfully extracted. The most important result in this experiment is that Ni(II) is not extracted from both its chloride or nitrate solutions and this leads to the conclusion that [P66614][SCN] presents itself as a promising extractant for the separation of Co(II) and Ni(II).The results of the present study have demonstrated that both [P66614][C272] and [P66614][SCN] in toluene as diluent can act as bifurcated extractants for Co(II) from nitrate or chloride solutions in the absence of strong acids or high concentrations of salts. This further supports the argument that ILs can extract the target species by ion-exchange or ion-pair extraction depending on the system being studied.The UV–visible spectral studies have shown that the Co(II) coordination sphere geometry in the organic phase for both extractants is tetrahedral which is also supported by the intense blue colour of the organic phase. A study of the Co(II) extraction stoichiometry has determined that the [P66614][C272] to Co(II) stoichiometric mole ratio is 2.8:1 (with saturated organic phase), which along with the fact that the structure is tetrahedral suggests eight phosphinate anions bridging three cobalt centres. However, it should be noted that if the Co(II) loading in the organic phase is low, a different stoichiometric mole ratio would be expected (i.e., 4:1) [31]. In the case of [P66614][SCN], the extraction stoichiometry, also confirmed by FTIR analysis, has been found to consist of two tetrahedral Co(II)-thiocyanate complex anions bridged by a thiocyanate ion. In the case of [P66614][SCN], an EDTA solution has been found to be able to completely back-extract Co(II), while regenerating the organic phase to its original form. In the case of [P66614][C272], Co(II) could be completely back-extracted with a solution of HNO3, although the IL likely did not return to its original form due to protonation and formation of Cyanex 272. Therefore, it can be expected that washing the back-extracted organic phase with a low concentration of sodium hydroxide solution or using EDTA instead of HNO3 as receiving phase could potentially regenerate the organic phase to its original form.It is also shown that the extraction of Co(II) is enhanced if its counter anion in the aqueous phase is more lipophilic (e.g., nitrate, thiocyanate). As expected for [P66614][C272], limited discrimination has been found in its ability to extract cations other than Co(II) due to the strong complexing ability of the phosphinate anion and the fact that neutral aqueous solutions have been used. On the other hand, [P66614][SCN] has not extracted Ni(II) from nitrate or chloride solutions thus proving to be a useful extractant for the separation of not only Co(II) and Ni(II) without the use of acidic solutions or high concentrations of chloride, but also for the separation of Co(II) from Ca(II), Mg(II) and Na(I). Syane A. Satyawirawan: Validation, Formal analysis, Investigation, Writing – original draft. Robert W. Cattrall: Conceptualization, Writing – review & editing, Supervision. Spas D. Kolev: Conceptualization, Resources, Writing – review & editing, Supervision. M. Inês G.S. Almeida: Conceptualization, Investigation, Writing – review & editing, Supervision, Project administration.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.Supplementary data to this article can be found online at https://doi.org/10.1016/j.molliq.2023.121764.The following are the Supplementary data to this article: Supplementary data 1

The solvent extraction of Co(II) from its aqueous solutions free of acids or chloride salts at high concentration by the ionic liquids trihexyl(tetradecyl)phosphonium bis(2,4,4-trimethylpentyl)phosphinate ([P66614][C272]) and trihexyl(tetradecyl)phosphonium thiocyanate ([P66614][SCN]) was characterised at saturation of the organic phase with Co(II) using toluene as diluent. The ionic liquids’ cation participated in the formation of two ion-pairs in the organic phase, i.e., one with the complex anion containing cobalt, and the other with the cobalt counter anion, thus preserving electroneutrality in both the organic and aqueous phases. This was considered to be an example of bifurcated extraction. UV–visible spectral studies of the intensively blue organic phase after Co(II) extraction demonstrated that the Co(II) coordination sphere geometry in the case of both extractants was tetrahedral. The Co(II) extraction stoichiometry was experimentally determined for the first time by saturating the organic phase with Co(II). The [P66614][C272] to Co(II) stoichiometric mole ratio in the extracted adduct was found to be 2.8:1, suggesting that eight phosphinate anions acted as bridging ligands around three cobalt centres. For the [P66614][SCN], the stoichiometric mole ratio was found to be 2.5:1 with the Co(II)-thiocyanate adduct consisting of two Co(II)-thiocyanate tetrahedra linked by a bridging thiocyanate ion. The Co(II) extraction was enhanced when the aqueous phase contained an anion with high lipophilicity (e.g., nitrate, thiocyanate). [P66614][SCN] did not extract Ni(II) from nitrate or chloride solutions, thus demonstrating its potential for the separation of Co(II) from Ni(II) without requiring the addition of high concentrations of acids or chloride salts to the aqueous phase.

Industrial waste management is beneficial for the energy security of Poland and development of local economy. The by-products of food industry are abundant sources of biomass which can be thermally converted into fuel, chemicals, or used for environmental applications [1–3]. One of the methods to utilize the by-products of food industry is pyrolysis. Fast pyrolysis is an economical method to convert biomass waste into oil, which is characterized by a high energy density. However, oil obtained by fast pyrolysis has undesirable properties such as low chemical stability and high acidity [4]. The main disadvantage of oils obtained by fast pyrolysis is the high oxygen content; therefore, deoxygenation and aromatization of oils have been investigated in many studies [5–8]. Zeolites have been effectively used to decrease the oxygen content in oils. Zeolite Socony Mobil #5, abbreviated ZSM-5, and Zeolite Y are commonly used to deoxygenate the biomass pyrolysis vapors [4,5]. Silica–alumina ratio (SAR) significantly affects the acidity and reactivity of a catalyst during biomass pyrolysis [6]. The catalyst acidity increases with decreasing SAR values; zeolites with low SAR values exhibit better cracking capabilities [7,8]. Moreover, zeolites with large pore sizes allow easier access to more long-chain compounds at the acidic sites for decarbonization and dehydration reactions [9,10]. Therefore, the catalytic performance can be improved by the addition of hydrogen donors during pyrolysis [11,12]. Zeolites with low SARs afford relatively lighter oils [13]. One of the major technical barriers inhibiting the commercialization of the catalytic pyrolysis technology is rapid catalyst deactivation [14,15]. Despite promising advances in catalytic fast pyrolysis, the issue of coke deposition on the catalyst surface persists [16,17]. Catalyst deactivation involves active site poisoning, pore blockage, mineral deposition, and is inversely proportional to the SAR [18,19]. A catalyst with the high aluminum content (SAR < 30) promotes the aromatization reaction, which corresponds to a high amount of acid sites. Aromatic compounds decrease with increasing SAR ratio. When increasing the SAR ratio of the catalyst, its deactivation was considerably as a result of the lower rate of coke precursor condensation reactions [20]. As a result, the significantly prolonged catalytic lifetime increases with an increase in the SAR value [21]. Heracleous et al. [22] reported catalyst regeneration under air oxidation at 500 °C, without adversely affecting the catalyst structure and strength of the acidic sites; catalyst activity was maintained even after seven catalyst regeneration cycles [23]. Yung et al. [24] reported that 89% of coke could be removed from the catalyst during 120 min of regeneration at 550 °C. Interestingly, the coke deposition rate was higher in the fluidized bed method than in the fixed bed method [25]. Another significant aspect that determines coke deposition on the catalyst is the catalyst-to-biomass (C/B) ratio [15,26–28]. Excess catalyst increases the reaction probability of the pyrolysis products with the acidic sites on the catalysts [15]. An increase in the C/B ratio from 0.5 to 10 resulted in an increase in the hydrocarbon vapor content in the range of 6–15% liquids [26]. Additionally, the catalytic activity could be lost when the C/B ratio was 1/10. With an increase in the C/B ratio from 0 to 15, the yields of the monocyclic and polycyclic aromatics increased, whereas those of the oxygenates progressively decreased [27].Compared to fast pyrolysis, intermediate pyrolysis is a relatively new method of thermal conversion, and the first pilot-scale reactor for intermediate pyrolysis was developed in 2009 [29]. In addition to char and gas, the products of intermediate pyrolysis include two-phase liquids (aqueous phase + bio-oil). Fast pyrolysis is often selected owing to the maximum liquid yield obtained; however, it is not useful for industrial residue or agricultural biomass [29]. In terms of agricultural products like straw, switch grass, husks, or miscanthus - fast pyrolysis produces a low yield of bio-oil, from 35% for wheat straw to 58 wt% for miscanthus [30]. Additionally, high water content in agriculture biomass is disadvantages in fast pyrolysis. The formation of high-molecular-weight tars and dry and brittle char is a critical issue, rendering intermediate pyrolysis more attractive than fast pyrolysis. Intermediate pyrolysis is characterized by longer vapor and feedstocks residence times than those in fast pyrolysis, as shown at Table 1.Although the yield of bio-oil obtained by intermediate pyrolysis is lower than that obtained by fast pyrolysis, the bio-oil is more stable and contains less oxygen [33]. Research on intermediate pyrolysis has mainly focused on the effects of the type of feedstock and process temperature on the yield and product quality [34–39]. High pyrolysis temperatures (500 °C) favor fragmentation reactions and result in a better quality oil with a low oxygen mass fraction [34]. Oils produced during the intermediate pyrolysis of waste sludge have satisfactory characteristics for use as diesel engine fuel [35]. Interestingly, even an aqueous phase with a low heating value and high water content can be effectively used. The anaerobic digestion of the water phase with the addition of char allows fuel (methane) production [36]. According to Primaz et al. [37], the char yield decreased with an increase in temperature from 400° to 600°C. Higher pyrolysis temperatures resulted in secondary char cracking and increased the gas product yield. Pyrolysis gas can be recycled to the reactor as a carrier gas which consists of nitrogen > carbon dioxide > carbon monoxide > methane > hydrogen [38]. Recent studies have suggested the risk of clogging pipes by tars formed during the intermediate pyrolysis of biomass [40]. Therefore, use of a catalyst is advisable to minimize tar formation, but only a few studies have investigated catalytic intermediate pyrolysis [34,41–44]. The commercial Ru/C catalyst is better suited for oil deoxidation than NiCu/Al2O3. In addition, more aqueous phase was formed when Ru/C was used instead of NiCu /Al2O3 [34]. Mohammed et al. [41] determined the effects of ZSM-5 and Zeolite A (zinc-exchanged) on the intermediate pyrolysis of Napier grass, where the yield of biochar did not change after introducing the ZSM-5 catalyst into the reactor. Furthermore, a low proportion of ZSM-5 catalyst (up to 1.0% by weight of feedstock) had no effect on the oil yield [41]. Catalytic biomass pyrolysis with and without a steam reformer was the aim of the study reported by Mahmood et al. [42]. The addition of steam increased the calorific value of the gas from 3 to 10.8 MJ/m3 at 500 °C, mainly due to an increase in the hydrogen content. Intermediate biomass pyrolysis with thermocatalytic reforming promoted the formation of phenol in oil (30.88% mass) and hydrogen in gas (19.4 vol%) [43]. Recent thermogravimetric analysis–Fourier-transform infrared spectroscopy (TGA–FTIR) studies have shown that zinc-containing nanopowders affect the pyrolysis yield. The bio-oil yield increased in the presence of ZnWO4, ZnAl2O4, and Mn–Zn ferrite, while the highest non-condensable gas yield was obtained with the addition of 2% Ag/ZnO catalyst [44].Only a few reports on catalytic intermediate pyrolysis are currently available, which is the main motivation to investigate this subject area. It is well-known that the application of catalysts with low SAR values of < 30 in fast pyrolysis is advantageous, but this effect is not as prominent in intermediate pyrolysis. Therefore, the effect of catalyst acidity (Zeolite Y: SAR = 26 and ZSM-5: SAR = 352) on biomass pyrolysis was investigated. In this study, the physical and chemical properties of the biomass and catalysts before and after pyrolysis were examined. Catalytic pyrolysis with 2-, 4-, 6-, 8-, and 10-fold usage of the same catalyst (ZSM-5 or Zeolite Y) was performed, and the coke contents of the catalysts, trace elements in the catalysts, and compositions of the pyrolysis products were analyzed. The novelty of this work includes the determining the effect of coked catalysts after 2-, 4-, 6-, 8-, and 10-fold usage in biomass pyrolysis. The effect of the multiple usage of the catalysts on impurity deposition was also investigated. Moreover, the catalysts that were reused 10 times during pyrolysis were regenerated, and impurity removal was examined. To the best of our knowledge, such studies have not been previously reported in the literature. Additionally, it was studied the amount of coke after the regeneration of ZSM-5 and Zeolite Y under intermediate pyrolysis conditions. This is significantly important to define the kind of coke e.g., hard (Cn) or light coke (-CnH2n).Post-extraction rapeseed meal (abbreviated as RM), a by-product of oil production, was obtained from an industrial oil-pressing plant located in Poland. In 2020, rapeseed harvest in Poland was 2.7 million tons [45], from which approximately 1.6 million tons of meal was produced. In the 2020/21 season, global rapeseed harvest forecasts will amount to about 63.0 million tonnes. In the group of key producers, crops increased in Canada, India, China and Australia [45]. RM is traditionally used for feeding animals because of its high protein content. However, many recent studies have reported new methods of waste management, which can allow the expansion of its application potential [23,46,47]. Xia et al. [23] demonstrated that porous carbon derived from RM could be successfully used for various energy storage applications such as in lithium-ion batteries and supercapacitors. Zhang et al. [46] proposed the synthesis of a catalyst based on carbon derived from RM with nitrogen and sulfur doping. The new catalyst was an excellent replacement for the commercial Pt/C catalyst used for energy conversion in fuel cells. Poskrobko and Król [47] reported the significant potential of RM for use in the co-gasification of wood biomass and RM. The addition of RM to the wood biomass contributed to an increase in the calorific value of the synthesized gas.The catalysts used in this study were the hydrogenated forms of ZSM-5 and Zeolite Y (ZY) that were purchased from Acros Organics and Alfa Aesar, respectively. The catalysts were in the form of gray (ZSM-5) and white (ZY) powders.A Truspec CHNS 628 Leco (USA) analyzer was employed for the ultimate analysis (carbon (C), hydrogen (H), nitrogen (N) and sulfur (S)) of RM, catalysts, char, bio-oil, and aqueous phase. TGA of the catalysts under air and nitrogen atmospheres (constant flow rate of 50 mL/min, approximately 4 mg sample, and heating rate of 10 °C/min) was performed using a Mettler Toledo TGA/SDTA 851 system (Switzerland). The elemental contents of the catalysts were evaluated using the X-ray fluorescence (XRF) method (ZSX Primus II Rigaku spectrometer, USA). The morphology and structures of the chars were determined using scanning electron microscopy (SEM; Inspect S50 apparatus, FEI, the Netherlands). SEM images were collected using a secondary electron detector in the high-vacuum mode, and the applied acceleration voltage was 3 keV. The specific surface areas and average pore diameters of the fresh, used, and regenerated catalysts were determined using the Brunauer–Emmett–Teller (BET) method. The total pore volume was determined using the Barrett–Joyner–Halenda (BJH) method. The analysis was done by apparatus of Micromeritics, USA. FTIR spectroscopy was employed to identify the functional groups of the organic compounds in the studied samples using a Bruker Alpha II system (Bruker Optics Inc., USA). The infrared absorption frequency was in the 400–4000 cm−1 range. The bio-oil composition was determined by gas chromatography-mass spectrometry (GC-MS; Agilent GC 7890 B equipped with an MS 5977 A mass spectrometer and a flame ionizer detector, Agilent Technologies, USA) technique. The gas phase was analyzed by GC (Agilent Technology 7890 A, Agilent Technologies, USA).A simplified diagram of the experimental set-up employed to investigate the pyrolysis process is shown in Fig. 1. Preparation of RM for the experiments included drying under ambient conditions and sieving of the particles (300–750 µm). Each experiment began by weighing 1.5 g of the catalyst and placing it in a reactor. Next, the reactor was electrically heated to 500 °C and maintained at a constant temperature for 1 h. The RM sample (1.5 g) was placed on a boat into the zone of the water cooler, and the reactor was purged with nitrogen at a flow rate of 100 mL/min for 5 min. The main process of intermediate pyrolysis started with the insertion of the boat with the RM into the heated furnace. The residence time of the sample was 7 min. During this time, the temperatures of the sample and reactor were monitored using K-type thermocouples. Fig. 2 shows the temperature profiles of the sample and reactor. One RM sample was investigated for 30 min, including 5 min of nitrogen purge, 7 min of pyrolysis, and 18 min of sample cooling. After starting the pyrolysis, that is, from the fifth minute of investigation, the temperature of the sample increased rapidly and reached 412 °C after 2 min of pyrolysis. Then, the temperature of the sample increased slowly. After the pyrolysis was completed, the sample boat was moved back to the water-cooler zone. The pyrolysis vapors flowed to the ice tank, where the aqueous and bio-oil phases were condensed. A gravimetric settler was used to separate the aqueous phases and bio-oil. The non-condensed dried gases were collected in a Tedlar bag for GC analysis.Five series of measurements were performed for the investigated catalysts used for the pyrolysis of 2, 4, 6, 8, and 10 RM samples; here, 10 RM sample indicates that the catalyst was held in the reactor for 10 pyrolysis cycles of the new biomass sample. After each series of measurements, the catalyst was removed from the pyrolysis reactor, and the composition of the pyrolysis gas was determined for each series of measurements. Liquid phases (bio-oil and aqueous phase) were not collected for each sample series (2, 4, 6, 8, and 10) because of the need to clean the system and difficulty in collecting an appropriate amount for analysis. After 10-fold usage of ZSM-5 and Zeolite Y catalysts, they were regenerated in muffle furnace. Regeneration took place in an air atmosphere at a temperature of 500 °C for 5 h. Table 2 shows proximate, ultimate and compositional analysis of rapeseed meal (RM). The volatile matter (74% mass) is comparable to that of wood, and the moisture content (5.66%) is low. The low moisture content is an advantage for RM as it prevents biological degradation during storage. However, the ash content is relatively higher compared to those in other biomass wastes [9,43]. Based on the ultimate analysis, the effective hydrogen to carbon molar ratio (H/C eff ) was calculated using Eq. (1). (1) H / C eff = ( H − 2 O ) / C The H/C eff ratio of the biomass commonly ranges from 0 to 0.35 [11,27,28]. The high amount of oxygen with moderate carbon and hydrogen contents in the biomass results in an H/C eff ratio of 0.22 for the RM. Biomass with a low H/C eff ratio tends to produce aromatics during pyrolysis [27]. In RM, the cellulose, hemicellulose, and lignin contents are comparable (24.23%). The RM composition is dominated by other biopolymers such as protein, fat, and polysaccharides [48].Additionally, RM was ashed at 550 °C to identify the components present in the ash. The ash-forming elements in the biomass may play important roles during pyrolysis. Elements from the alkali metal group, mainly potassium, can promote primary dehydration reactions [49]. Furthermore, a significant amount of ash may accumulate on the catalyst surface during subsequent pyrolysis cycles. The chemical analysis data for ash is presented in Table 3, indicating that ash is phosphorus-rich and contains relatively large amounts of potassium, calcium, and magnesium. The presence of potassium and calcium increase the intrinsic catalytic reactivity of RM. Fig. 3 shows the FTIR spectrum of RM, where the stretching and bending vibrations are marked. The wavenumber ranges for different functional groups are listed in Table 4, based on the research equipment database. The exception is the class of compounds corresponding to the wavenumber below 700 cm−1, which was identified based on the literature data [50,51], because of a lack of data in the system database. N–H, C–N, and CO bonds of the proteins present in RM (protein content is 35% of dry mass [48]) were identified in the FTIR spectrum.Two catalysts, ZSM-5 and ZY, were used for pyrolysis. Hereafter, the following representations are used to describe the catalyst samples: ZSM-5_0 and ZY_0 indicate fresh catalysts before pyrolysis, ZSM-5_10 and ZY_10 represent the catalysts used 10 times for pyrolysis of the RM samples, and ZSM-5_reg and ZY_reg are the catalysts after regeneration at 500 °C for 5 h. Fig. 4 shows the SEM images of the catalysts, where significant differences are observed in the shapes and structures of ZSM-5 and ZY. Particles of the ZSM-5 catalyst are larger than the ZY particles, and are crystals with round shapes and different sizes, which are conglomerated. No noticeable changes in the main crystal structure and morphology are observed after the use and regeneration of ZSM-5, but the particles appear to be less dense. The ZY catalyst has smaller particles with a regular cubic shape compared to ZSM-5. The particles have rough edges, and slight agglomeration of particles is observed. The regeneration of ZY does not affect the morphology and structure of this zeolite.FTIR was used to examine the structures of the ZSM-5 and ZY samples. The structure of the zeolite can be considered as a set of interconnected TO4 (T = silica or alumina) units of SiO4 and AlO4 tetrahedra. Fig. 5 shows the FTIR spectra of the catalysts, where the peaks are observed only in the mid-infrared region (1250–400 cm–1). The adsorption bands are observed at 1224, 1070, 797, 546, and 437 cm−1 for the ZSM-5 catalyst, and 1205, 1053, 832, 607, 527, and 455 cm−1 for the ZY catalyst. The adsorption bands at 1224 cm−1 (external asymmetric stretching), 1070 and 1053 cm−1 (internal asymmetric stretching), 797 cm−1 (external asymmetric stretching), and 455 cm−1 (T–O band) correspond to highly siliceous materials [52]. The representative signals, but with lower intensities at 546 and 547 cm−1, are attributed to the bending vibrations of the five-membered rings of O–T–O in the zeolite topological structure. Wavenumber ranges are provided in accordance with the literature reports [53].The fresh (0), used (10), and regenerated (reg) catalysts were characterized using the nitrogen physisorption method to determine the effects of deactivation and regeneration on the physicochemical properties of the catalyst ( Table 5). BET analysis was performed to determine the specific surface area (SBET). A comparison of ZSM-5 and ZY showed that the SBET value of ZY was twice that of ZSM-5. The 10-fold usage of ZY during pyrolysis resulted in a significant decrease in the surface area (up to 34%), whereas the change was small (only 7%) for ZSM-5. After heating at 500 °C, ZY was regenerated, attaining its original or higher surface area. For ZY_reg, the surface of the micropores increased, which caused an increase in the BET surface. Regeneration can change the physical structure of the catalyst [54]. As it was reported by Brito et al. [55] an increase in BET surface area can be observed during the first catalyst regeneration, and subsequent regeneration can lead to SBET drop. The increase in BET surface area is attributed to the dehydrogenation of the coke resulting in the formation of additional mesoporosity at the regeneration temperature [24]. After regeneration of ZSM-5, the surface area value did not return to that of the fresh catalyst, indicating that a temperature of 500 °C was very low to clean the catalyst; this observation was consistent with that described in a prior literature report [24]. The total pore volume (Vtot) for ZSM-5 was relatively lower and decreased after usage and regeneration, whereas the volume was 4-fold higher for ZY and did not change after pyrolysis. In terms of the porosities of these zeolites, the diameter of micropores (<2 nm) and mesopores (2–100 nm) were almost equal in ZY, and thus the surface area of the micropores was approximately two times the area of the mesopores. In ZSM, both the volume and surface area of the mesopores were higher than those of the micropores.To determine the rate of coke deposition on the catalysts, ultimate analysis (carbon, hydrogen and nitrogen) was performed for the ZSM-5 and ZY samples. As shown in Fig. 6a, the carbon content in ZSM-5 increases from 2% to 4.16% (two-fold) after 10 uses, while it increases from 1.21% to 20.63% (17-fold) in ZY (Fig. 6b). Carbon deposited faster on ZY than on ZSM-5, and the carbon content in ZY_4 after fourth use was the same as that in ZSM-5_10 after 10 uses. The coke deposited on the analyzed catalysts is classified as hard Cn, because the increase of hydrogen content was negligible compared to carbon. As reported in the literature [56], removal of hard coke is actually more difficult than light coke (-CnH2n). Moreover, elemental analysis revealed that the ZY_10 catalyst contained more hydrogen and nitrogen than ZSM-5_10. Carbon deposition reduces the porosity of the catalyst and hinders the pyrolysis active site reaction [15]. As reported by Heracleous et al. [22], coke deposition depends on the duration of pyrolysis. The ZSM-5 catalyst (SAR = 45) contained 4.8% carbon and 6.9% carbon after 20 and 40 min, respectively.TGA data of the investigated catalysts are presented in Fig. 7a-f. The mass losses of the catalysts upon heating to 900 °C under nitrogen and air atmospheres are shown in Figs. 7a-b and 7c-d, respectively. Under nitrogen atmosphere, the mass losses for ZSM-5_0 and ZY_0 are 4.2% and 12.8%, respectively. In contrast, the catalyst samples after pyrolysis under nitrogen atmosphere are characterized by lower mass losses of 3.7% and 6.7% for ZSM-5_10 and ZY_10, respectively. As shown in Fig. 7c-d, ZSM-5_10 and ZY_10 display two distinct mass loss regions. The first mass loss is observed at 25–300 °C, which is attributed to the desorption of water and volatile species (i.e., reactants, products, and reaction intermediates) adsorbed on the catalyst surface [57]. The second mass loss is observed at 300–900 °C, which corresponds to the decomposition of coke species deposited on the catalyst surface. The mass losses for the ZSM-5_10 and ZY_10 catalysts in air atmosphere at 300–900 °C show good agreement with the results of the coke content analysis shown in Fig. 6a-b. Catalyst regeneration was performed in a muffle furnace in air. The mass loss trends for the ZSM-5_reg and ZY_reg catalysts are shown in Figs. 7e and 7f, respectively. Analyzing the results of the thermogravimetric investigations, the increase in the specific surface area is due to the large weight loss (12%) of the fresh catalyst when heated to a pyrolysis temperature of 500 °C. The loss in weight of the catalyst leads to an increase in SBET. However, as mentioned earlier, it was done only after the first regeneration. It was proved that it is more difficult to regenerate the catalyst used for intermediate pyrolysis than for fast pyrolysis [58].Elemental analysis of the regenerated catalysts was performed, which afforded the following results: 1.36% C, 1.15% H, and 0.33 N in the ZSM-5_reg catalyst, and 1.09% C, 0.54% H, and 0.26% N in ZY_reg. However, more carbon and hydrogen remained in ZSM-5_reg than in ZY_reg, but less coke was deposited on ZSM-5_10. This result indicated that the complete removal of coke from the catalyst during regeneration was very difficult. As a result of biomass pyrolysis, other elements were also deposited on the catalysts [16]. Table 6 summarizes the elements detected in fresh, used, and regenerated catalysts. Increases in most of the trace elements (except Na, Zr, and Mg) in ZY_10 are observed in comparison to the contents in the ZY_0 sample. Interestingly, the most abundant elements in RM ash, such as phosphorus and potassium, were not deposited on the ZSM-5_10 and ZY_10 catalysts. The regeneration resulted in the removal of coke, but also a significant loss in sulphur deposited on the catalysts. The sodium content of the catalysts increased after regeneration.After the pyrolysis of each 1.5 g sample of RM, the boat retained 0.446 g of char (29.7% by weight). The type of catalyst used did not affect the char composition. Elemental analysis data of char are listed in Table 7. As expected, char has a significantly higher content of carbon and lower percentages of hydrogen and oxygen compared to those of the feedstock. The second most abundant element in the char is potassium. Potassium is a desirable element for pyrolysis because it increases char production [59]. Moreover, potassium has a better impact on the pyrolysis product yields than phosphorus [60]. Fig. 8 shows the morphologies of raw biomass and char obtained after pyrolysis using ZSM-5 (as an example). The data for the char obtained using ZY is not presented because the catalyst types does not significantly affect the char properties and morphology. As shown in Fig. 8, raw post-extraction RM mainly has a fibrous texture with a few spherical cavities, and the surface is coherent. Analysis of char shows the effect of pyrolysis on the material structure. The char has many openings and a tubular-shaped structure, which can be compared to a honeycomb structure, where the diameters of the honeycomb pores are <10 µm. The obtained honeycomb structure of char can play an important role in oxygen reduction during catalytic processes [61]. Notably, some char particles do not have open pores, suggesting that a pyrolysis temperature of 500 °C is not sufficiently high to remove all volatile matter.One of the most important characteristics of char is its functional groups, corresponding to the wavenumber range of 1700–400 cm−1, as shown in Fig. 9. The FTIR spectra of the chars obtained by pyrolysis using ZSM-5 and ZY were similar because the catalyst was only in contact with the released vapors of the heated RM samples. Two functional groups in the broadest wavenumber ranges, corresponding to the aromatics (1700–1500 cm−1) and alkanes (1200–980 cm−1), were notable.The aqueous phase is produced because of the presence of residual moisture within the RM and during dehydration reactions in the pyrolysis process. The total weight of the liquid phase (aqueous phase + bio-oil) after pyrolysis of 10 RM samples (1.5 g per sample) did not depend on the type of the catalyst used and was 8.0 g +/- 0.5 g. However, the weight percentage of bio-oil in the liquid phase was higher when ZSM-5 was used instead of ZY, i.e., 21% and 19%, respectively. Unfortunately, during the pyrolysis variants of eight or less samples, it was more difficult to obtain a repeatable mass of the liquid. Hereafter, the combined aqueous phase and bio-oil samples from all variants (2, 4, 6, 8 and 10 RM samples) were analyzed. As shown in Table 8, the aqueous phase formed using ZY has 2.51% more carbon than the aqueous phase formed using ZSM-5, which contains 2.5% more oxygen. In contrast, more carbon and less hydrogen percentages are observed in the bio-oil collected after pyrolysis using ZSM-5 instead of ZY.Aqueous phases and bio-oils were also characterized using FTIR, and the data are shown in Fig. 10. The broad peak at 3400–3200 cm−1 implies that the aqueous phase and bio-oil samples contain compounds with O-H groups, such as water, alcohols, and phenols. The sharp bands at approximately 2956, 2925, and 2853 cm−1 are attributed to the C-Hx vibrations, indicating the presence of hydrocarbon molecules [62]. The vibrations observed between 1780 cm−1 and 1680 cm−1 are attributed to the presence of the CO bond, which indicates the presence of aldehydes, ketones, and carboxylic acids. A few low-intensity peaks observed at 1200–970 cm−1 correspond to the presence of alcohols and phenols.To better understand the effects of the two different catalysts on the RM intermediate pyrolysis, GC-MS analysis of the bio-oils was performed. Table 9 shows the detected and identified compounds in the bio-oil and their peak area percentages. The bio-oil contains a complex mixture of low- to intermediate-molecular-weight hydrocarbon chains with C4 to C24 units, which is consistent with the results reported by Mahmood et al. [42]. The bio-oils obtained by pyrolysis were categorized into alkanes, alkenes, alkynes, alcohols, acids, ketones, single-ring and polycyclic aromatic hydrocarbons, phenols, N-containing aromatic compounds, nitriles, amines, and other oxygenated compounds. Aromatic compounds, ketones, acids, and oxygenates were the dominant components in the bio-oils obtained by catalytic pyrolysis. Phenols, cyclopentanes, and furans are classified as flammable organics [43]. The SAR value of the catalyst (26 for ZY_0 and 352 for ZSM-5_0) in intermediate pyrolysis slightly affected the contents of carboxylic acids, ketones, polycyclic aromatic hydrocarbons, phenols, and N-containing aromatic compounds in the bio-oils. Interestingly, the total content of the oxygenates in the bio-oil (alcohols + ketones + phenols + other oxygenates) was lower with the use of ZSM-5 (26.51%) than that of ZY (29.57%). This result is in contrast to that of fast pyrolysis [27]. This may be because of different vapor residence times in the reactor for the aforementioned types of pyrolysis. In the present study on intermediate pyrolysis, the residence time was 20 s, while in fast pyrolysis this time was approximately 1.5 s [8].The concentrations of the detected pyro-gas components, converted to nitrogen-free composition, are shown in Fig. 11a-j. The main component of pyro-gas was CO2, whose concentration increased with the reuse of catalysts. The highest concentrations of carbon monoxide from the catalytic pyrolysis process were obtained with the first use of catalysts (both ZSM-5 and Zeolite Y). After 4-, 6-, 8-, and 10-fold uses of the catalysts (ZSM-5 or Zeolite Y) a reduction in the concentration of carbon monoxide in the pyro-gas was observed, which may be attributed to the weakening of the decarbonylation and decarboxylation reactions [41]. The hydrocarbons consisted of alkanes, alkenes, and alkynes. The pyro-gas obtained using the ZSM-5_0 and ZY_0 catalysts was characterized by the lowest carbon dioxide and highest hydrogen concentrations. The use of ZY_2 afforded higher concentrations of hydrogen, ethene, and ethyne compared to those obtained using ZSM-5_0. In particular, for the ZY catalyst, increases in the percentages of hydrogen and ethyne in the pyro-gas were accompanied by a low methane content. This phenomenon may be represented by reaction (2). (2) 2 C H 4 → 3 H 2 + C 2 H 2 The concentration of methane in obtained pyro-gas in the presence of ZSM-5 catalyst was lower and constant during the reused of the catalyst. This means that the catalyst promotes the conversion of methane precursors to aromatic hydrocarbons [41,63]. In the case of Zeolite Y, the concentration of methane increased significantly from its eighth use. The concentrations of methane, acetylene, propane, and butane were higher for each increase in the reusage cycle of the ZY catalyst compared to those observed with ZSM-5.For the catalytic intermediate pyrolysis of rapeseed meal, mass and carbon balances of the products were also calculated and shown in Fig. 12. Due to the inability to analyze samples of oils and aqueous phases after 2- and 4- fold usage of catalysts, the mass and carbon balances were based on average values from pyrolysis of 10 samples. The catalyst samples were not weighed after pyrolysis and they were not included. When analyzing the mass balance (Fig. 12a-b)), the dominant product is the aqueous phase: 42% for ZSM-5% and 43% for Zeolite Y. This result was confirmed by other literature studies [42,43]. The char accounted for nearly 30% of the mass of products, and it did not depend on the type of catalyst used. This conclusion was expected because the catalyst was not in direct contact with the char. Compared to the intermediate pyrolysis of wood (see Table 1), lower yield char and pyro-gas weight percentages were obtained. On the other hand, from the carbon balance shown in Fig. 12 c-d resulting that char contains 41.4% carbon. A similar percentage of carbon from this balance was reported by Funke et al. [33] for four types of feedstocks. The second pyrolysis product in terms of carbon content was pyro-gas. Interestingly, the carbon percentage in pyro-gas formed in the presence of the ZSM-5 catalyst was 5% higher than in the case of Zeolite Y. It is worth noting that the Zeolite Y catalyst was responsible for 5% by mass of carbon in the balance.Catalytic intermediate pyrolysis of RM with 2-, 4-, 6-, 8-, and 10-fold usage of the catalyst (ZSM-5 or Zeolite Y) was performed. The repeated use of less acidic ZSM-5 catalyst (SAR = 352) was better than that of ZY (SAR = 26) for bio-oil production with a low oxygen content. The type of catalyst did not significantly affect the yield and char composition because the catalyst was used downstream to the pyrolysed sample. Comparable masses of the liquid phases were obtained for both ZSM-5_10 and ZY_10 catalysts. The main component of the liquid phase was the aqueous phase (approximately 81% for ZY and 79% for ZSM-5). The carbon and hydrogen contents in the bio-oils obtained using both catalysts were similar. GC-MS analysis allowed the identification of the main components in bio-oils, which included carboxylic acids, ketones, phenols, and compounds containing nitrogen and oxygen. In the bio-oils obtained using ZSM-5, a lower total oxygenate content (alcohols + ketones + phenols + other oxygenates) was observed compared to the bio-oils obtained using ZY. The elemental analysis data confirmed the higher oxygen content (greater by 0.7%) of the bio-oil obtained using the ZY catalyst compared to that obtained using the ZSM-5 catalyst.Notably, repeated use of the same catalyst significantly affected the composition of the produced pyro-gas, particularly methane and hydrogen contents. With an increase in the catalyst usage, the contents of hydrogen and carbon monoxide decreased, while carbon dioxide concentration in the pyro-gas increased. The 10-fold usage of the catalyst resulted in significantly higher carbon deposition on ZY than on ZSM-5. In contrast, after the regeneration of the catalysts (500 °C and 5 h), more coke remained on ZSM-5 than on ZY. Wojciech Jerzak: Conceptualization, Methodology, Writing – original draft, Data Curation, Investigation. Aneta Magdziarz: Project administration, Funding acquisition, Writing – review & editing, Investigation. Ningbo Gao: Writing – review & editing, Investigation. Izabela Kalemba-Rec: Writing – review & editing, Investigation.The authors declare no known competing financial interests or personal relationships that can influence the work reported in this paper.The research project was supported by the program "Excellence initiative – research university" of the AGH University of Science and Technology, Poland (Grant AGH No. 501.696.7996), the Key Program for China EU International Cooperation in Science and Technology Innovation, China (Grant No. 2018YFE0117300), and Shaanxi Provincial Natural Science Foundation Research Program-Shaanxi Coal Joint Funding, China (2019JLZ-12).

This study describes an experimental investigation comparing the effects of two catalysts (ZSM-5 and Zeolite Y) on intermediate pyrolysis. The catalysts are characterized by different acidities, expressed by silica-to-alumina ratios of 352 (ZSM-5) and 26 (Zeolite Y). The post-extraction rapeseed meal was pyrolysed in a fixed-bed furnace at a temperature of 500 °C, and the pyrolysis products (char, liquid phase, and pyro-gas) were characterized in detail. The catalysts were assessed based on their reusage capability. Five-fold more carbon was deposited on Zeolite Y than on ZSM-5 after 10-fold use during the pyrolysis of rapeseed meal. Moreover, the ultimate analysis of the catalysts showed increases in the hydrogen and nitrogen contents, which were significantly higher for Zeolite Y than those for ZSM-5. The catalysts showed different effects on the properties of the products. Better-quality pyro-gas was obtained with Zeolite Y, but reusage of this catalyst resulted in decreases in the hydrogen and carbon oxide concentrations. Compared to Zeolite Y, ZSM-5 afforded bio-oil with a lower oxygen content. Phenols and ketones were dominant compounds in both bio-oils. Regeneration of Zeolite Y caused to increase of specific surface area.

No data was used for the research described in the article.Dibenzyl phosphateDimethylglyoximePolyethylene glycolNickel-metal hydride batteryNiMHB anodeIonic liquidRare earth elementLight rare earth elementHeavy rare earth elementRare earth element oxideLanthanidesLayered double hydroxideAqueous biphasic systemExtraction efficiencyMetal−organic frameworkNanoporous grapheneA pluronic triblock copolymerPr and NdSodium rare earth double sulfate (NaRE(SO4)2)Polyethylene oxide polymer with an average molar mass of 1500 g mol−1 Tributyl phosphateTri-n-octyl phosphine oxideDi-(2-ethylhexyl) phosphoric acid1-methylimidazolium hydrogen sulfateTrialkylphosphine oxideBis (2,4,4-trimethylpentyl) phosphinic acidBis(2,4,4-trimethylpentyl) dithiophosphinic acidBis [2, 4, 4-trimethyl pentyl] mono thiophosphinic acidTrioctyldecylamine chloride2-ethylhexylphosphonic acid mono-2-ethylhexyl esterTricaprylmethylammonium chloride (dihexyl diglycolamate) Trihexyl(tetradecyl)phosphonium chloride (nitrate) Trihexyl(tetradecyl)phosphonium bis(2,4,4-trimethylpentyl) phosphinate1-ethyl-3-methylimidazolium chloride[Trihexyl(tetradecyl) phosphonium]2[2,2' -(1,2-phenylenebis(oxy))dioctanoate]Trihexyl(tetradecyl)phosphonium thiocyanate (bis(trifluoromethanesulfonyl)imide) Tetrabutylammonium nitrateEthylenediaminetetraacetic acidDiethylenetriaminepentaacetic acidEthylenedinitrilo tetra acetic acid disodium salt dihydrateSec-octyl phenoxy acetic acidBenzyltributylammonium myristic acetateBenzyltributylammonium dodecanedioic acetate2-thenoyltrifluoroacetone1-butyl-3- methylimidazolium bis(trifluoromethanesulfonyl)imide (bis(perfluoroethanesulfonyl)imide) (di(2-ethylhexyl)-oxamate) Tetraoctylammonium dioctyl-diglycolamate (di(2-ethylhexyl)-oxamate) (oleate) Methyltrioctylammonium naphthenic acid (Peanut oil) (Rapeseed oil) (Sunflower seed oil) (Flaxseed oil) Trioctyl(2-ethoxy-2-oxoethyl)ammonium dihexyl diglycolamateN,N,N,N-tetrabutyl-3-oxapentane-diamideN,N,N,N-tetraoctyl-3-oxapentane-diamide1,4-diisopropylbenzene1-methyl-1-butylpyrrolidinium bis(trifluoromethanesulfonyl)imideTributyl-methyl ammonium nitrate (bis(trifluoromethylsulfonyl)imide) (2,4-dimethylheptyl) (2,4,4′-trimethylpentyl) phosphinicRare earth elements (REEs) are a group of metals comprising 15 lanthanides, as well as yttrium and scandium [1]. These elements are naturally present in various minerals, including silicates, carbonates, and phosphates. Due to their similar ionic radii, REEs exhibit interchangeable properties within a given mineralogy, resulting in difficulties in their separation during mineral processing [2]. The increase of the atomic number in REEs adds electrons to their inner incomplete orbital (4f) rather than the outer orbital, resulting a decrease in the ionic radius of the trivalent lanthanides from La3+ to Lu3+, a phenomenon known as “lanthanide contraction” [3]. As a result of lanthanide contraction, it is possible to separate different rare earth elements individually [4].REEs are inputs to a range of critical applications from the electronics and energy industry to defense and military technologies, accounting for more than 10% of the US$75 trillion worth of the global economy [4,5]. Due to the strategic role of these elements, REE-bearing minerals were categorized as critical minerals in 2018 in the United States [6]. Fig. 1 illustrates the utilization of REEs in various applications in the United States (2018), the European Union (2019), and globally (2019). The major industrial applications of individual REEs are also featured in Table 1 . REEs are primarily utilized in the catalysts, glass, and ceramics industries in the United States and the European Union. However, affected by China's REEs industry, the production of magnets is the main end-use application of these elements globally.Although the term “rare” is used for rare earth elements, they are not particularly rare in terms of average crustal abundance [4,9]. The scarcity of REEs is relevant to their low concentration in most of their deposits [4]. Deposits containing a sufficient concentration of REEs to support mining operations are not frequently found [4,9]. This can make the mining and extraction of REEs more challenging and costly. Carbonate igneous rocks, known as carbonatites, are the most significant REEs deposits in the world [4]. Alkaline igneous rocks, placer deposits with monazite-xenotime mineralization, and ion-adsorption clay deposits are other principal economic REEs sources [9]. The major secondary resources of REEs are permanent magnets, batteries, fluorescent lamps, catalysts, phosphogypsum, bauxite residue (red mud), mine tailings, metallurgical slags, and fly ash [5]. According to the United States Geological Survey (USGS) 2021 report, world economic reserves of REEs in 2020 totaled 120,000 kt REEs oxides (except for yttrium), and more than 500 kt for Y2O3. Based on the report, as shown in Fig. 2 , the world mine production of REEs was estimated to be 240 kt of REEs oxides (REOs) in 2020. China is the leading country in both reserves and mine production, with 44,000 kt and 140 kt, respectively [10].REEs as strategic high-tech metals currently play a critical role in the global technological advancement. As the world is moving towards a cleaner and greener future, meeting the growing demand for REEs is becoming increasingly challenging due to the limited availability of economically viable resources and the monopoly of a few countries on their production and supply. Primary and secondary resources are the two main options for securing demand for REEs. Factors such as the relative distribution of light (LREEs) and heavy (HREEs) REEs within a deposit, the complexity of processing, and the environmental impacts associated with REEs mining and processing greatly affect the economic viability of REEs prospecting. The co-occurrence of several LREEs (e.g., Nd, La, and Ce) in a single deposit is often accompanied by simultaneous extraction of the others, which makes the overall process more cost-effective. However, HREEs are typically found in lower concentrations within deposits, which raises concerns about potential shortages in the near future due to limited production volume [2].Electronic waste (e-waste) is considered one of the main secondary sources of REEs that can potentially secure a significant part of their demand. Fluorescent materials, battery alloys, and permanent magnets are the major sources of REEs. The total amount of e-waste generated worldwide in 2019 was estimated to be 53.6 Mt, and it is projected to reach 74.7 Mt by 2030 [11]. Only 17.4% of the generated e-waste in 2019 was recovered, which means about $57 billion worth of recoverable materials were mostly dumped or incinerated [11]. Currently, less than 1% of the REEs in end-of-life products are recycled globally [12], indicating a significant potential for utilizing e-waste as a source of REEs. Studies have recently focused on replacing REEs in critical technologies with alternative materials. Hitachi, Tesla, Renault, and BMW are among the leading companies actively engaged in the minimization or substitution of REEs in the compartments of their products [2].A variety of physical, chemical, and thermal processes have been developed to extract REEs from primary or secondary resources. Physical processes are mainly utilized to concentrate REEs from gangue materials in raw ore (e.g., through froth flotation), and to liberate REEs-containing materials from other components in secondary resources. Despite being simple and environmentally friendly, they often require further chemical or thermal treatment to remove trace impurities. Pyrometallurgical recovery of REEs, which involves the separation of REEs in oxide form (slag) via a high-temperature process, is mainly used on a large scale for commercial operations. The resulting REEs slag is then subjected to hydrometallurgical processing to remove various impurities. Alternatively, hydrometallurgy can be used as the sole recovery technique for the recovery and refining of REEs from different sources. The process usually begins with leaching the raw material using acidic or alkaline solutions. Depending on the composition and chemical structure of the raw material, a roasting step may be incorporated into the process in order to remove volatile components and enhance the water-solubility of the solid phase. The resulting aqueous solution (pregnant leach solution, PLS) is then used in various hydrometallurgical processes, such as precipitation, solvent extraction, or adsorption, for group or individual separation of REEs. Due to the chemical similarities of REEs, efficient separation and purification of individual REEs require the development of new technologies to reduce the cost and environmental impact associated with the process.Due to the hydrogen-absorbing properties of nickel-lanthanide alloys, REEs have been utilized in energy storage since the early ’90s, leading to their extensive use in nickel-metal hydride batteries (NiMHBs). NiMHBs were commercialized in 1991 and have since found applications in electric vehicles and rechargeable products [5]. The components of NiMHBs include an anode composed of hydrogen-absorbing alloys (MH), a cathode made of nickel hydroxide, a separator between the electrodes, and a potassium hydroxide electrolyte. The anode is an AB5-type alloy (A is a mixture of La, Ce, Nd, and Pr; and B is Ni, Co, Mn, and Al) containing about 33 wt% REEs, mainly La with about 21 wt%, and Ce, Nd, and Pr with 6.5 wt%, 3 wt%, and 2 wt%, respectively. Fig. 3 depicts a schematic representation of a NiMHB cell, along with the average chemical composition of its anode and cathode.This review paper aims to investigate various methods proposed for recovering REEs from spent Ni-MH batteries and study the available techniques for individual separation of the elements. The current state-of-the-science and an in-depth understanding of hydrometallurgical, pyrometallurgical, and combined methods for recovering REEs from the spent batteries will be introduced and discussed. To complete the recovery cycle of REEs from NiMHBs, the current and emerging techniques for individual separation of these elements are investigated, and their challenges and potential future directions are highlighted. The study is a comprehensive examination of the available techniques, which combines both narrative and systematic reviews to assess the potential of each method for sustainable recycling of NiMHBs.The preliminary processing of NiMHBs usually includes discharging (to avoid short circuits), opening of casings, liberation of seals and separators, shredding, and separation of different fractions (fluff, metals, and black mass). The black mass fraction, which mainly contains anode and cathode, is processed for REEs recovery. Fig. 4 is a process flow diagram illustrating the various stages involved in the recovery and individual separation of REEs from NiMHBs. The recovery of REEs from spent NiMHBs can be achieved through either hydrometallurgical or pyrometallurgical processes. However, the purification and individual separation of the elements is typically performed via hydrometallurgical methods. The methods are further explained in the following sections.Electroslag refining [13,14], liquid metal extraction [13,15–19], glass slag process [20–26], direct melting [27–30], and gas-phase extraction [28,31–37] are the most common pyrometallurgical techniques developed for the recovery of REEs from different sources [14]. Among the mentioned processes, molten slag extraction is the most favorable technique for the recovery of REEs from spent NiMHBs, due to the efficient recovery of both REEs and Ni. The method is suitable for group separation of REEs oxides in the form of a slag phase, which requires further hydrometallurgical processing for purification and individual separation of the elements. During the smelting process, REEs present in the waste batteries are oxidized and isolated in the slag phase, while Ni, Co, and Fe are transferred to the metallic phase. The properties of the slag, such as viscosity, melting point, and vapor pressure, are directly related to its composition. Effective separation of the REEs in the slag phase can be achieved through the adjustment of the slag composition by the addition of fluxing agents. Fluxing agents are substances introduced to the smelting system to improve the fluidity and to effectively isolate unwanted impurities in the form of a slag [38]. Limestone, silica, dolomite, lime, borax, and fluorite are the most common fluxes used in this technique [38]. Both the metal and slag phases are valuable products in the molten slag extraction process. However, in the context of recovering REEs from batteries, selecting the appropriate slag system is a crucial step in the process. The slag system selected must meet the criteria outlined in Table 2 .As previously stated, the batteries undergo pre-processing, and the separated black mass is utilized as the raw material for the molten slag extraction process. The black mass is introduced into the smelting furnace in conjunction with fluxing agents and reductant materials, either directly or after a pre-oxidation step. As a result, smelting can be divided into two types of direct smelting and oxidized smelting.During direct smelting of NiMHB, metallic REEs act as reducing agents due to their high affinity with oxygen. Since the Gibbs free energy of the REEs oxide is lower than the other oxides in the system, the following reactions are expected to take place during the smelting process [39]: (1) 3 S i O 2 ( s ) + 4 R E ( l ) → 3 S i ( l ) + 2 R E 2 O 3 ( s l a g ) (2) 3 M g O ( s ) + 2 R E ( l ) → 3 M g ( g ) + R E 2 O 3 ( s l a g ) (3) C a O ( s ) + R E ( l ) → C a ( g ) + R E 2 O 3 ( s l a g ) ( T > 1700 o C ) (4) 3 N i O ( s ) C a t h o d e + 2 R E ( l ) → 3 N i + R E 2 O 3 By reduction of SiO2 according to (Eq. (1)), Si either dissolves in the Ni–Co alloy or forms intermetallic compounds. MgO is reduced to metallic Mg (Eq. (2)); however, it will eventually evaporate at ∼1090 °C, implying that magnesia is not a suitable refractory for this process. The same type of reaction is valid for Al2O3; however, metallic Al may be re-oxidized due to its high reactivity with oxygen, or it may form intermetallic compounds. Müller and Friedrich [39] investigated the effectiveness of CaO–SiO2 and CaO–CaF2 fluxing systems to recover REEs from NiMHBs. According to the study, the CaO–SiO2 system shows poor metal-slag separation due to high slag viscosity, leading to Si enrichment in the metal phase. In contrast, the CaO–CaF2 system exhibits excellent metal-slag separation, highly efficient REEs separation, and a near-complete transformation of Ni and Co to the metal phase. Maroufi and colleagues [40,41] suggested an alternative approach in which REEs are oxidized via iron oxide (hematite) during smelting. The process involves heating a 1:1 mixture of MHA-hematite in a graphite crucible, where the CO2 released from the reaction of hematite with graphite oxidizes the REEs, resulting in the formation of a REEs-rich slag phase.By heating the battery's black mass in air atmosphere, nickel hydroxide is first decomposed into NiO and water at a temperature in the range of 250–300 °C [42], and by increasing the temperatures, all the constituent elements eventually transform into their oxide forms. Maroufi and colleagues employed pure iron as an alloying solvent and reducing agent to separate Ni and Co from REEs in MHA. A metallic alloy and a REEs-rich slag were produced by heating the mixture of oxidized MHA and pure iron at 1550 °C in argon atmosphere. The presence of carbon in this process is essential as it contributes to the metallization of Ni, Co, and Fe and improves phase separation by reduction of the metallic phase viscosity [40,43].An alternative method for selective reduction of oxidized Ni, Co, and Fe is to introduce a hydrogen gas atmosphere, eliminating the need for adding reducing agents to the charge. Deng and colleagues studied the recovery of REEs from NiMHBs employing the thermal isolation method. A pre-oxidized mixture of anode-cathode was reduced by hydrogen at 800 °C. The product was mixed with Al2O3 and SiO2 as fluxing agents, and the mixture was smelted at 1550 °C in argon atmosphere, resulting in a slag phase rich in REOs, Al oxide, and Mn oxide, and a Ni-based alloy. The slag was water-quenched and then crushed to facilitate the acid leaching process [44]. The SiO2–CaO-based slag is another system that can efficiently absorb REEs oxides from pre-oxidized NiMHB. Despite Tang and colleagues reporting high efficiencies for REEs recovery and metal-slag separation through this system, their results demonstrated a significant amount of metallic Ni and Co remaining trapped in the slag phase [45].The Ellingham diagram in Fig. 5 illustrates the temperature-dependent stability of the various oxides present in the NiMHB smelting process. Since REEs’ lines are positioned below other elements (except for Ca), REOs are more stable than other oxides, which means that metallic REEs can easily reduce the oxides of the elements above their lines. In the diagram, La represents other LREEs as their lines have similar positions. The reduction of NiO and CoO by REEs is thermodynamically favorable due to the significant difference in Gibbs free energy between REOs and Ni–Co-oxides. However, the limited amount of metallic REEs in the anode is not sufficient for complete metallization of NiO and CoO. Therefore, the use of external reducing agents such as C or Fe is necessary in both direct and oxidized battery smelting methods to achieve full metallization.To understand the reaction mechanism of direct and oxidized battery smelting, MHA is considered as the raw material, in which LaNi5, NdCo5, and CeCo5 are the dominant phases. The products of both processes are a slag phase rich in REEs and a Ni-based alloy.The battery (and additives, e.g., fluxing agents) is first heated in the smelting furnace (T ~ 1600 °C [39]) under reducing conditions, where the anode phases are decomposed based on Eqs. (5)–(7). (5) L a N i 5 ( s ) → Δ L a ( s ) + 5 N i ( s ) (6) C e C o 5 ( s ) → Δ C e ( s ) + 5 C o ( s ) (7) N d C o 5 ( s ) → Δ N d ( s ) + 5 C o ( s ) Ni and Co form an alloy bath at the bottom of the furnace. Owing to the high affinity of REEs with oxygen, they may either reduce other oxides present in the system or be oxidized by the purged oxygen (Eq. (8)). The difference in the Gibbs free energy of the oxides is the highest between LREEs and Ni and Co, which means thermodynamically, reduction of their oxides by LREEs is favorable. Due to their low density, oxidized REEs form a slag layer placed on top of the alloy bath. (8) R E ( l ) + O 2 ( g ) → R E 2 O 3 ( s l a g ) The oxidized battery smelting process involves a more complex reaction mechanism, requiring a pre-oxidation step that leads to the formation of various intermetallic compounds. This technique involves the oxidation of Ni and Co prior to smelting, which is achieved through an oxidation roasting step at around 1000 °C, as shown in Eqs. (9)–(13) [40,46]. (9) 2 L a N i 5 ( s ) + 7 O 2 ( g ) → 2 L a N i O 3 ( s ) + 8 N i O ( s ) (10) 2 L a N i O 3 ( s ) → L a 2 N i O 4 ( s ) + N i O ( s ) + 0.5 O 2 (11) L a 2 N i O 4 ( s ) → L a 2 O 3 ( s ) + N i O ( s ) (12) N i ( s ) + 0.5 O 2 ( g ) → N i O ( s ) (13) C o ( s ) + 0.5 O 2 ( g ) → C o O ( s ) During the oxidation stage, REEs present in the raw material, or those that may have formed as per Eqs. (5)–(7), also undergo oxidation and generate intermetallic oxides, as shown in Eqs. (14) and (15). (14) N d ( s ) + 4 C e ( s ) + 4.75 O 2 ( g ) → 5 N d 0.2 C e 0.8 O 1.9 ( s ) (15) L a ( s ) + 2 N i ( s ) + 2 O 2 ( g ) → L a N i O 3 ( s ) + N i O ( s ) The Ni and Co oxides are then reduced to their metallic forms. This reduction can be accomplished through a separate solid-state gas reduction process or by incorporating a reducing agent into the oxide product before it is added to the smelting furnace, as shown in Eqs. (16) and (17). As previously mentioned, carbon-based materials or pure iron can be utilized as reductants in this process. Fig. 6 illustrates a schematic of the slag/metal interface in the smelting furnace, as well as the key reactions between MHA, the slag system, and refractory materials. (16) N i O ( s ) + [ reductant ] → N i ( s ) + [ reductant oxide ] (17) C o O ( s ) + [ reductant ] → C o ( s ) + [ reductant oxide ] Table 3 provides an overview of the various slag systems compositions used for recovery of REEs from NiMHBs via molten slag extraction, along with their respective advantages and disadvantages.In the molten slag extraction method, the REEs isolated in the slag phase are typically in high concentrations (50–60 wt% REEs) [47]. In most cases, the slag undergoes hydrometallurgical processes for further purification or individual separation of the elements [48]. It could also be used as the raw material for molten salt electrolysis. As an example of industrial application of molten slag extraction, Umicore Group, a Belgian mining company, and Rhodia Group (Solvay) in France have jointly developed a process based on Umicore's ultra-high-temperature (UHT) smelting technology for recycling NiMHBs [49,50]. The spent batteries, coke, iron, and fluxes are melted via plasma heating in a shaft furnace. This results in the production of a (Ni–Co–Fe)-based alloy and a REEs-rich slag phase. The slag is then sent to Rhodia's plant for impurity removal and further refinement of the REEs [51].While pyrometallurgical routes have demonstrated promising potential for concentrating REEs from various sources, hydrometallurgical techniques are widely employed for REEs extraction and purification. Precipitation and solvent extraction are the most commonly used hydrometallurgical methods for recovering REEs from end-of-life NiMHBs. Recently, techniques such as the use of ionic liquids and supercritical fluid extraction have also been developed.Like most transition metals, the REEs exist in aqueous solutions as cations, with solubility generally decreasing as pH increases [52]. The REEs cations typically have a charge of +3, but under reducing conditions (and generally high-temperature), europium may have a charge of +2, and under oxidizing conditions, cerium may have a charge +4 [4]. This difference in ionic charge has been used as a basis for separating Ce and Eu from other REEs in multiple studies [53–56]. Common inorganic ligands that REEs tend to combine with include sulfate, carbonate, fluoride, hydroxide, and phosphate [57]. Factors such as pH, temperature, redox conditions, and thermodynamic stability of phases play a role in determining the tendency of REEs to partition between solid (either by adsorption or (co)precipitation) and aqueous phases [4].The hydrometallurgical techniques for recycling NiMHBs typically involve an initial acid leaching step, in which the crushed battery (or electrode material) is dissolved in an acidic solution. The type and concentration of the acid, solution temperature, and agitation time all play a crucial role in determining the efficiency of the leaching process. Mineral acids are widely used in the leaching process of NiMHBs. The most common leaching agents reported in the literature are H2SO4 [58–60], HCl [61–63], and HNO3 [64,65]. There are also reports on the use of organic acids, such as oxalic acid and formic acid as leaching agents [66–68]. However, it should be noted that during this process, the REEs tend to precipitate in the form of organic salts. Although the leaching efficiency and leaching kinetics of mineral acids are significantly higher, organic acids can be more environmentally friendly for greener production [69,70]. Table 4 provides a summary of various leaching techniques and their respective outcomes for extracting REEs from NiMHBs electrode materials.As previously noted, REEs tend to bind with inorganic ligands such as sulfates, fluorides, and phosphates, with oxalates being the most commonly reported among organic ligands [63,68,83]. Fig. 7 illustrates the chemical structures of anhydrous REE double salt, oxalate, and sulfate.NaRE(SO4)2, known as REE double salt, form in the presence of Na+ and SO4 2− in the solution, typically at a pH greater than 1, as shown in Eq. (18) [74–76,84–86]. Under these conditions, the LREEs present in NiMHB leaching solution (La, Ce, Nd, Pr) eventually form mixed crystal REE double salts [74]. This advantage has made double salt precipitation the most commonly reported technique for group separation of REEs. The resulting double salt can be mixed with NaOH solution to produce rare earth mixed hydroxides (Eq. (19)) [74]. These mixed hydroxides can then decompose into REEs mixed oxides through a calcination step. (18) R E 3 + ( a q ) + N a + ( a q ) + 2 S O 4 2 − ( a q ) + x H 2 O ( l ) → R E . N a ( S O 4 ) 2 . x H 2 O ( s ) (19) N a R E ( S O 4 ) 2 . x H 2 O ( s ) + 3 N a O H ( a q ) → R E ( O H ) 3 ( s ) + 2 N a 2 S O 4 ( a q ) + ( x + 3 ) H 2 O ( l ) The data presented in Fig. 8 demonstrates the effectiveness of the precipitation method for extracting La, Nd, Pr, and Ce from a sulfate leachate of NiMHB black mass in the form of double salts. The results indicate a high level of efficiency, with almost complete precipitation of La, Nd, and Ce, and about 90% efficiency for Pr, within 1 h of the reaction.Oxalic acid (C2H2O4) is an effective agent for precipitation of REEs from different acidic solutions. It has also been reported to be an effective leaching agent for MHA [68]. However, it has a disadvantage of co-precipitation of Ni, Co, and Cu with REEs, which requires further purification steps [63]. Additionally, to achieve efficient separation, the amount of oxalic acid added to the system must exceed its stoichiometric amount with the REEs significantly [83]. The reaction of REEs with oxalate ion is shown in Eq. (20). The resulting product, REEs mixed oxalate, is usually subjected to calcination to convert it into mixed oxides. (20) 2 R E 3 + ( a q ) + C 2 O 4 2 − ( a q ) + x H 2 O ( l ) → R E 2 ( C 2 O 4 ) . x H 2 O ( s ) Liquid antisolvent precipitation is a technique based on the addition of a water-miscible organic solvent to an aqueous solution, which creates a supersaturated solution by changing the solubility of the solute. The supersaturation forces the solute to precipitate in the form of a salt [88,89]. In a study, Korkmaz and colleagues [90] achieved a total recovery of up to 86% for REEs from a MHA leachate with negligible co-precipitation of other elements. REEs were precipitated as mixed hydrated sulfates when ethanol and 2-propanol were added as antisolvents, as Korkmaz reported. Although REEs recovery yield was high, a significant volume of alcohol was required to effectively precipitate REEs [90]. The literature on the precipitation of REEs from the leaching solutions of NiMHBs electrode materials is summarized in Table 5 .The mixed REEs compounds or solutions obtained from various processes in most cases require purification or individual separation before being used in high-tech applications. While several techniques such as ion exchange, fractional crystallization, extraction chromatography, and chemical deposition have been reported in the literature, solvent extraction is one of the most commonly used methods for individual separation of REEs [5]. In addition to its high efficiency and cost-effectiveness, solvent extraction has a high potential for scaling up. Nevertheless, there are concerns regarding its environmental impacts, as it requires the use of toxic organic solvents. The basic formulas for solvent extraction are presented in Eqs. (21)–(24). (21) D i s t r i b u t i o n r a t i o : D M = [ M ] O r g a n i c [ M ] A q u e o u s (22) S e p a r a t i o n f a c t o r : β A / B = D A D B (23) O / A = O r g a n i c p h a s e v o l u m e A q u e o u s p h a s e v o l u m e (24) S / L = S o l i d p h a s e m a s s L i q u i d p h a s e m a s s ( v o l u m e ) [ g . g − 1 ] o r [ g . m l − 1 ] The use of cation exchangers is a common method for extracting REEs due to their positive charge in aqueous solutions. The cation exchanger replaces H+ with the REE cation, forming a complex that can be dissolved into the organic phase. The extraction reaction is shown in Eq. (25). (25) R E n + + n H A ‾ ⇌ R E A n ‾ + n H + Neutral extractant's molecule create a neutral complex with the anion of an acid and a REE cation. Eq. (26) represents the complexation of REE3+ with nitrate ions and tributyl phosphate as a neutral extractant [92]. (26) R E 3 + + 3 N O 3 − + 3 T B P ‾ ⇌ R E ( N O 3 ) 3 . 3 T B P ‾ A variety of amines can be utilized to extract REE ions from acidic solutions. Research indicates that amines are primarily effective in extracting REEs from sulfate-based aqueous solutions [92,93]. The extraction of REEs form a sulfuric acid solution by a primary amine can be explained through two steps: 1) acid extraction by the amine (Eqs. (27) and (29)), and 2) complexation of the amine-acid molecule with the REE cation (Eqs. (28) and (30)) [92].At low H2SO4 concentration: (27) 2 R N H 2 ‾ + 2 H + + S O 4 2 − ⇌ ( R N H 3 ) 2 S O 4 ‾ (28) R E 3 + + 1.5 S O 4 2 − + 1.5 ( R N H 3 ) 2 S O 4 ‾ ⇌ ( R N H 3 ) 3 R E ( S O 4 ) 3 ‾ At high H2SO4 concentration: (29) R N H 2 ‾ + H + + H S O 4 − ⇌ R N H 3 . H S O 4 ‾ (30) R E 3 + + 1.5 S O 4 2 − + 1.5 ( R N H 3 . H S O 4 ) 2 ‾ ⇌ ( R N H 3 ) 3 R E ( S O 4 ) 3 ‾ + 1.5 H 2 S O 4 Solvent extraction is an effective method for separating target elements or impurities from aqueous solutions [94]. Table 6 summarizes recent studies on the solvent extraction recovery of REEs from leaching solutions of NiMHBs electrode materials. Petranikova and colleagues [78] successfully extracted Fe and Zn from HCl leachate of MHA with a mixture of Cyanex 923 and TBP, and then extracted REEs using a Cyanex 923-TBP-Decanol system. Similarly, Fernandes and colleagues [62] purified the HCl leachate of NiMHB active material from Zn and Fe using a pure TBP system, and separated Co using a 10 v.% Alamine 336 organic phase. In this work, two different approaches were proposed for recovering lanthanides from the raffinate: solvent extraction with PC 88 A or precipitation using (NH4)2C2O4, among which the solvent extraction yielded a product of high purity.Another approach reported in literature is the group extraction of elements, followed by selective scrubbing of impurities from the loaded organic phase [95,96]. Larsson and colleagues [95] used a mixed Cyanex 923-TBP organic phase to extract REEs, Al, Co, and Mn from the HCl leachate of MHA, while most of the Ni content remained in the raffinate. They selectively scrubbed Co and Mn from the loaded organic phase using a NaNO3 solution, and then stripped the REEs and Al from the organic phase using an HCl solution [95]. In a study by Zhang and colleagues [96], D2EHPA was used for recovering REEs from NiMHB sulfuric acid leachate. The presence of Ni and Co in the stripping solution can negatively impact the process of recovering REEs via precipitation of REEs oxalate, as they may co-precipitate with the oxalate ions. To address this, the co-extracted Co and Ni were scrubbed from the loaded organic phase using a diluted H2SO4 solution prior to stripping [96]. It is important to note that the stripping process of trivalent iron from D2EHPA typically requires concentrated acidic solutions [94]. Therefore, it may be more efficient to use D2EHPA for REEs recovery from MHA leaching solutions, as MHA does not contain iron.Ionic liquids (ILs) are liquids composed of only ions, with a melting point below 100 °C [98]. They have several desirable properties, such as low vapor pressure, broad liquid range, low flammability, high thermal stability, good solubility for both inorganic and organic compounds, as well as controllable hydrophobicity [92,99]. These properties make them suitable for a variety of applications, including in chemical industry, electrochemistry, and separations. The use of ILs for the separation of REEs is a novel approach, in which ILs can be used in solvent extraction as extractant, diluent, or both [92]. The extraction mechanism of ILs is similar to molecular solvents, and the metal cation forms an extractable neutral complex with the IL. Moreover, the approaches mentioned for organic extractants, e. i., group or individual separation, scrubbing, and selective stripping techniques, can be applied to ionic liquids.An article by Prusty and colleagues [100] compares the extraction mechanisms of REEs using ILs and molecular solvents. The unique properties of ILs make them potential candidates for optimizing REEs solvent extraction, as they can potentially be replaced with conventional organic solvents and diluents [101,102]. Furthermore, ILs have a reduced environmental impact compared to typical organic solvents.There are limited reports on REEs extraction from NiMHBs using ionic liquids. Table 7 summarizes recent studies on REEs extraction from NiMHBs electrode materials leaching solutions using ionic liquids. Larsson and Binnemans [103] investigated the separation of REEs from a synthetic chloride solution derived from dissolved NiMHBs electrode materials. The extractant employed was Cyanex 923 dissolved in the nitrated form of Aliquat 336 IL, [A336][NO3]. Although the proposed system was not able to separate REEs from other elements, it demonstrated significant potential for selective scrubbing and stripping. More than 98% of the co-extracted Ni was scrubbed from the extract by an MgCl2 solution. In a stepwise stripping process, Co and Mn were stripped from the extractant using a NaNO3 solution, followed by selective stripping of REEs using a HCl solution. In another study, Larsson and Binnemans [104] employed Aliquat 336 IL to separate Co, Fe, Mn, and Zn, from the synthetic chloride leachate of NiMHBs. REEs were extracted using Cyanex 923 diluted in the nitrated form of Cyphos IL 101. The only impurity, Ni, was scrubbed from the extract and REEs were then stripped from the IL as a group. The nitrated form of Cyphos IL 101 was also used by Hoogerstraete and Binnemans [105] to separate La from a Ni-based solution. The aforementioned studies leveraged the so-called split-anion extraction process, which is well-adaptable to ILs. Nitrate anion has a strong affinity for coordination with the organic phase, while chlorine anions show higher affinity for aqueous phases. Since REEs form extractable nitrate complexes, the substitution of chloride for nitrate in Aliquat 336 and Cyphos 101 molecules allows the extraction of REEs coordinated with the nitrate ligands of the ILs [106,107].Supercritical fluids have high solvating power, low viscosity (during the process), and low surface tension, making them highly efficient extraction media. The low viscosity and high diffusivity of supercritical fluids result in superior mass transport compared to traditional liquids, leading to improved extraction yield and rate when compared to other extraction methods [112]. CO2 is particularly suitable as a solvent for extracting REEs from various resources, as it is cost-effective, has a moderate critical pressure, and low critical temperature [48]. Yao and colleagues [113] utilized supercritical CO2 as a solvent and TBP-HNO3 complex as a chelating agent to extract REEs from MHA. The proposed system involves leaching out REEs from the raw material using supercritical CO2, followed by the complexation of REEs cations with TBP and nitrate ions. High recovery yields for REEs were reported in this study under optimal experimental conditions. One limitation of supercritical fluid extraction is the dependency on high-temperature and high-pressure reactors. Despite this, it is considered a more environmentally friendly process in terms of hazardous waste generation when compared to solvent extraction.The application of adsorption in the separation of REEs is less common compared to solvent extraction and precipitation techniques, due to its lower extraction capacity for REEs and lower selectivity in separating individual REEs [114]. However, there have been a limited number of studies on the use of adsorbents in the recovery of REEs from NiMHBs. Table 8 presents a summary of recent studies on the application of various adsorbents in the recovery of REEs from NiMHBs electrode materials. In a study by Araucz and colleagues [115], Purolite S975 and Diphonix resins were employed to separate La(III) from a synthetic nitrate solution containing La(III) and Ni(II) in the presence of citric acid. The results indicated that while the resins had a high adsorption efficiency, they were not effective for separating the elements individually [115]. Fila and colleagues [116] reported a similar outcome for Diphonix resin, as well. Gasser and Aly [117] suggested a novel synthetic adsorbent named Mg–Fe-LDH-Cyanex 272 for the separation of La(III) and Nd(III) from the sulfate leachate of spent NiMHBs. The new adsorbent displayed superior adsorption capacities compared to Purolite S975 and Diphonix resins in previous literature [115–117]. The adsorbent was stripped using diluted HCl, but its uptake efficiency for La dropped from 87% to 40% after ten cycles of use [117]. Zhi and colleagues [118] developed a new extraction-precipitation method using dibenzyl phosphate (DBP) to recover REEs from waste NiMHBs. By adding DBP to the sulfuric acid leachate of the battery, nearly complete precipitation of all the REEs was achieved, while co-precipitation of Ni, Co, and Mn was less than 1.75%. This method has the advantage of generating a larger particle size of the precipitate compared to other conventional techniques, which facilitates solid-liquid separation. Additionally, the loaded DBP is recyclable and reusable through a simple stripping process, making this technique more environmentally friendly.The aqueous biphasic systems (ABS) are ternary systems composed of water and two water-soluble solutes with distinct hydration entropies, which enable the reversible separation of a mixed-phase into two aqueous-rich phases within a specific composition range [120]. Vargas and colleagues [64] proposed the ABS for the recovery of La, Ce, and Ni from MHA using pluronic triblock copolymer (L35) and dimethylglyoxime (DMG) as the ABS. After a sequential extraction process, Ni, Ce, and La were separated with relatively high purities, demonstrating the potential of this method for the individual separation of these elements. De Oliveira and colleagues [61] suggested the use of ABS for the separation of La(III) from NiMHBs leachate utilizing a combination of PEO1500-Li2SO4-water in the presence of 1,10-phenanthroline (extractant agent) as the ABS. This work reported a high separation efficiency for La(III) over other REEs after three extraction steps.Korkmaz and colleagues [121] proposed a method for the recovery of REEs from MHA based on sulfation of raw material, selective roasting, and water leaching, which resulted in a total REEs recovery of 96%. The anode material was combined with concentrated sulfuric acid and subsequently dried. The sulfated mixture was then roasted at a high temperature, followed by leaching with water. The REEs were efficiently recovered with minimal contamination of Ni and Co. The solid residue of the leaching process is a mixture of nickel and cobalt oxides with trace impurities that may be subjected to further processing.Honda Motor Co., Ltd., in collaboration with Japan Metals & Chemicals Co., developed a hydrometallurgical process to recover up to 80% of the REEs present in NiMHBs [122]. The process involves the acid leaching of the active material and the recovery of REEs in the forms of oxides. The REEs are then metallized through a molten salt electrolysis process and subsequently reutilized to manufacture NiMH battery anode [84].In a patent by Smith and Swoffer [123], the batteries are hammer-milled under a water spray to produce a slurry. Following several physical separation processes, including screening and filtration, the majority of the non-metallic fraction is removed and an intermediate product rich in Fe and Ni is obtained. The product is then passed through a magnetic stripping device to separate Fe and Ni, followed by processing in a froth or foam floatation setup. The sediment of the floatation process is the metallic AB2 or AB5 alloy, which is filtered and recovered. In another patent by Burlingame and Burlingame [124], after battery dismantling, the active material is oxidized at 1000 °C, converting it to NiO and a REEs oxide-nickel oxide compound. The resulting oxide is then mixed with ammonium sulfate and pressed into a slug. The slug is heated at 450 °C and the residue is dissolved in deionized water. As a result, most of the REEs are leached out, while most of the NiO remains in the residue. The REEs in the solution can be precipitated in the form of oxalates and calcined to REEs oxides. Table 9 provides a summary of other methods in the literature for the recovery of REEs from NiMHBs. Fig. 9 illustrates the process flow of the key hydrometallurgical techniques for recovering REEs from NiMHBs. NiMHBs black mass or REEs-rich slag is utilized as the raw material, which can be leached out through conventional leaching using inorganic acids or supercritical fluid extraction technique. The solution obtained from the leaching process, comprising REEs and other ions, is then processed through hydrometallurgical techniques such as solvent extraction, ion exchange, and precipitation to selectively recover REEs from other metallic ions.Molten slag extraction is the primary pyrometallurgical method employed for recycling NiMHBs. The selection of the slag system is a critical step in this process, as the majority of the REEs present in the battery will end up in the slag. The main approaches for battery smelting are direct smelting and oxidized smelting, with the former being more suitable for industrial use as it requires fewer pre-treatment steps. While pyrometallurgy has been demonstrated to be an effective option for the concentration of REEs from various sources, the separation and purification of individual REEs are heavily dependent on hydrometallurgical techniques. Precipitation is a commonly employed technique for the group separation of REEs from different leachates due to its cost-effectiveness, high recovery yield, and the high purity of its product. However, the method exhibits poor selectivity toward individual LREEs. The precipitate formed is typically an organic or inorganic salt, depending on the precipitant and leaching medium utilized. Solvent extraction is the most widely reported method for industrial separation and purification of REEs. Nevertheless, the application of solvent extraction for the recovery of REEs from NiMHBs leaching solutions is limited, possibly due to the significant co-extraction of other metals alongside REEs. In recent times, concerns have been raised over the disposal of hazardous residues generated by the organic solvents used in this process. To address this issue, ionic liquids have been introduced as novel, environmentally friendly compounds that can be used as extractants, diluents, or both in solvent extraction systems. Ionic liquids have demonstrated significant potential for REEs separation on a laboratory scale. However, further research is necessary to evaluate their industrial feasibility [100]. The application of adsorbents is generally limited to leachates with low metal concentrations due to the limited uptake capacity of the conventional adsorbents. Supercritical fluid extraction of REEs from various sources is a novel technique that combines autoclave leaching and solvent extraction processes. The method has several advantages over conventional solvent extraction, including high mass transfer and extraction rate, and the simultaneous leaching and recovery of metals. However, it is more appropriate for the intragroup separation of REEs, and the process parameters must be carefully controlled [125]. As a novel method, the ABS process has demonstrated promising results for the recovery of REEs as an alternative to solvent extraction, enabling the separation of REEs from complex matrices. However, the stripping yield and regeneration of the utilized extractants remains an area of improvement. Additionally, there are other methods that involve combined hydro-pyrometallurgical approaches [79,121,124], some of which have been industrialized.The industrial recovery of REEs from NiMHBs has gained significant attention in recent years due to the increasing demand for REEs in various industries. The market for REEs is highly dependent on their end use, which can be affected by political and economic factors. The price and demand for REEs can fluctuate greatly, making it difficult to predict the profitability of recovering them from NiMHBs. The limitations and challenges of the industrial recovery of REEs from NiMHBs can be classified into several key categories, including. • Complex composition: Ni-MH batteries contain a complex mixture of metals and other materials, making it difficult to separate and recover the REEs. • Economic feasibility: REEs are present in NiMHBs in relatively low concentrations, which makes it challenging to economically recover them. Traditionally, Ni and Co have been the main target elements in an industrial recovery process. However, due to the increased attention to REEs as strategic metals, different industries have modified their processes to efficiently recover these elements from the batteries. • Environmental concerns: The processes used to recover REEs from batteries and the generated wastes can be environmentally damaging, and there are concerns about the potential release of toxic substances during the recovery process. • Technological limitations: There are currently limited technologies available for the efficient recovery of REEs from NiMHBs, which makes it difficult to scale up the process. • Lack of standardization: There are currently no widely accepted or consistent methods for recycling these batteries, which makes it difficult to track and recover rare earth elements. This lack of standardization can also make it difficult for recyclers to identify and separate batteries that contain rare earth elements, which further complicates the recovery process. Additionally, the lack of standardization can also make it difficult for recyclers to ensure that the batteries are being recycled in an environmentally responsible manner. Complex composition: Ni-MH batteries contain a complex mixture of metals and other materials, making it difficult to separate and recover the REEs.Economic feasibility: REEs are present in NiMHBs in relatively low concentrations, which makes it challenging to economically recover them. Traditionally, Ni and Co have been the main target elements in an industrial recovery process. However, due to the increased attention to REEs as strategic metals, different industries have modified their processes to efficiently recover these elements from the batteries.Environmental concerns: The processes used to recover REEs from batteries and the generated wastes can be environmentally damaging, and there are concerns about the potential release of toxic substances during the recovery process.Technological limitations: There are currently limited technologies available for the efficient recovery of REEs from NiMHBs, which makes it difficult to scale up the process.Lack of standardization: There are currently no widely accepted or consistent methods for recycling these batteries, which makes it difficult to track and recover rare earth elements. This lack of standardization can also make it difficult for recyclers to identify and separate batteries that contain rare earth elements, which further complicates the recovery process. Additionally, the lack of standardization can also make it difficult for recyclers to ensure that the batteries are being recycled in an environmentally responsible manner. Table 10 offers a comprehensive comparison of the principles, features, and technical advantages and disadvantages of various methods for recovering REEs from NiMHBs.A techno-economic analysis of the recovery of REEs from NiMHBs involves evaluating the technical feasibility of the process, the involved methods and equipment, as well as the purity and quantity of the recovered REEs. The economic analysis assesses the costs of the process, including the cost of the equipment, materials, and waste management, as well as the revenue generated from selling the recovered REEs, considering the market demand and REEs price. The number of studies on the techno-economic analysis of the recovery of valuable metals from NiMHBs is limited. Furthermore, the cost of the recovery process can vary depending on the techniques employed, making it challenging to conduct a comprehensive analysis that encompasses all aspects of the process.Lin and colleagues conducted a preliminary economic analysis to compare the recovery of valuable metals from NiMHBs through thermal and mechanical processes [126]. According to their work published in 2016, roughly 619 and 821 USD can be obtained in profits by recovering valuable metals from each ton of spent NiMH batteries through thermal melting and mechanical processes, respectively [126]. The diagram in Fig. 10 shows the flow of materials in the recycling process and the potential economic value of the recovered products as per market rates in 2022. Based on revenue potential per unit mass, didymium (Nd + Pr) metal and high-grade nickel metal are the two most valuable co-products which are recovered via recycling of the batteries. Despite comprising less than 1% of the total recovered materials by mass, didymium generates over 14% of the total potential revenue from all products recovered. Negative revenue represents the cost of disposing of waste products. The value of REE products and Ni metal are based on an expected purity of 99% or higher, as determined by the hydrometallurgical separation process used in the study [127].The individual separation of REEs is challenging due to their similar chemical properties, and as such, group recovery of REEs is the most commonly reported practice. The lanthanide contraction, as explained in Section 1.1, makes it more feasible to separate REEs with a large difference in atomic number (e.i., LREEs from HREEs) than adjacent REEs (except for Ce and Eu). Due to the limitations of pyrometallurgical processes, individual separation of REEs is typically achieved through hydrometallurgical techniques. Some methods, such as solvent extraction, require multiple repetitive stages to attain the desired purity of an element, while others necessitate precise attention to the solution's controlling parameters (e.g., for fractional precipitation). Additionally, factors such as process efficiency, environmental impact, investment and operational costs, and industrialization flexibility should be considered when selecting a recovery method.Given the presence of La, Ce, Nd, and Pr in NiMHBs, they can be isolated and purified collectively using methods such as precipitation. The solid products (e.g. REEs mixed oxides) can then be dissolved in acidic solutions for further processing. The current section outlines the latest studies on the individual separation of these elements to complete the recycling cycle of REEs from NiMHBs.Due to the versatility of the solvent extraction process and its scalability, as well as the diversity of options for the extraction media and system, solvent extraction is currently the leading technique for individual separation of REEs. The availability of various organic extractants and the ability to modify them through methods such as saponification, as well as the option to add complexing and auxiliary agents to the aqueous solution, have made solvent extraction adaptable to different target elements and aqueous media. The focus of many recent studies in this field has centered on the difference in REEs’ affinities for complexation with various aqueous species, extractants, or both. These affinities can be derived from differences in thermodynamic stability, hydrophobicity, or the extraction electrochemistry of distinct REEs complexes.Organophosphorous extractants are commonly utilized for the extraction and separation of lanthanides due to their selectivity. However, the use of these extractants alone for individual separation of lanthanides has shown limited success. To improve the outcomes, various auxiliary approaches such as saponification, selective scrubbing, the use of complexing or buffering agents, or taking advantage of the synergistic effect between extractants are often employed. PC 88 A [128–136], D2EHPA [130,132,133,137–141], Cyanex 272 [130,133,135,140,142–144], Cyanex 572 [132,145], TOPO [132,140,144], and TBP [139,140,143–145] are the most commonly used extractants in REEs separation. However, some studies suggested that synthesized extractants may results in better yields and higher separation factors between elements than conventional organophosphorus compounds [54,146]. The majority of published research in this area has focused on REEs separation from synthetic solutions, and the results may differ when using leaching solutions of NiMHBs due to the presence of other elements and associated interferences. Table 11 provides the key physical and chemical properties of conventional organic extractants used in REEs recovery and separation.The extraction of REEs via cation exchangers results in the liberation of H+ which leads to an increase in the acidity of the aqueous solution (as per Eq. (25)). To mitigate this effect, a pre-treatment technique known as saponification may be employed. The technique entails the treatment of the organic phase with an alkaline solution, such as sodium hydroxide or ammonium hydroxide solutions, to reduce the amount of H+ released during the extraction reaction by exchanging the hydrogen in the extractant molecule with the cation of the alkaline agent (e.g. Na + or NH4 +). Saponification can stabilize the process, enhance extraction efficiency, and improve the selectivity of REEs separation [92,133]. However, due to the high volume of wastewater generated, this method is being increasingly replaced by alternative techniques such as making use of synergistic effects between extractants [101] or adding buffering agents.Scrubbing is a technique utilized for the selective removal of targeted elements or impurities from a loaded organic phase (or any other type of adsorbent) using a scrubbing solution. One commonly reported approach in the literature for the separation of LREEs is to scrub impurities from the loaded organic phase using a pure solution of the target element. The target element refers to the element that is being purified. Considering X as the target element and Y as an impurity (where X and Y are REEs that are already loaded into the organic phase), Y can be removed from the loaded organic phases by scrubbing it with a pure solution of X. This process results in the substitution of X for Y in the organic phase and the transfer of Y into the aqueous phase, as outlined in Eq. (31) [129]. (31) X + 3 + Y ( H A 2 ) 3 ‾ ⇌ X ( H A 2 ) 3 ‾ + Y + 3 This approach has been demonstrated to be effective in countercurrent processes, where impurities (such as Y) are scrubbed from the organic phase after multiple sequential steps. Reports in the literature have documented the application of this method for the scrubbing of Pr from Nd, and La from didymium [128,129,140].Carboxylic acids (R–COOH) are among the most commonly used buffering agents in the solvent extraction separation of LREEs. Synthesized complexing agents have also been shown to have great potential in achieving high separation factors between LREEs [138,147]. However, the number of reports on these agents is limited. Carboxylic acids are cheaper and less hazardous to the environment compared to the addition of complexing agents such as EDTA and DTPA. They create a buffer system in the aqueous phase, similar to the effect of saponification, which prevents drastic drops in acidity resulting from the cation exchange mechanism (Eq. (25)), as outlined in Eq. (32) [148]. (32) H + + R C O O − ⇌ R C O O H The use of lactic acid, acetic acid, and citric acid as buffering agents is prevalent in the separation of adjacent lanthanides, among which lactic acid has shown superior performance [133,142,148,149]. The relative effectiveness of saponification versus the use of buffering agents may vary depending on the specific extraction system employed. For example, the application of lactic acid was found to be a more advantageous technique for La/didymium separation using D2EHPA, as compared to saponifying the extractant. However, saponification was determined to be more efficient in terms of improving selectivity for Cyanex 272. In the case of PC 88 A, there was no significant difference between the two approaches [133].In solvent extraction systems, the phenomenon of synergism refers to an increase in extraction efficiency when a combination of extractants is used, resulting in a performance greater than the sum of their discrete efficiencies [5], while in antagonism the opposite occurs. Synergism is a widely-used principle in the solvent extraction of REEs, as it can lead to increased extraction capacity and enhanced separation of the elements through increased hydrophobicity of the metal-extractant complex [5]. Studies have shown that synergism between Cyanex 272 and Alamine 336 is effective in the separation of didymium and La [150]. Similar results have been reported for the combination of D2EHPA and Cyanex 272 [140]. Likewise, the synergistic effect of 8-Hydorquinoline and PC 88 A was found to enhance the selective separation of Nd and Pr, with acetic acid used as a buffering agent [151].Recent advances in the synthesis of novel extractants have garnered significant attention in the individual separation of REEs due to their superior performance. Among the newly developed extractants, TiBDGA showed exceptional separation factor values of about 135 and 60 for the separation of Ce and La from Pr and Nd, respectively [146]. Similarly, DEHAPO has been identified as a promising extractant for the selective separation of Ce from La with a separation factor of up to 167 [54]. Despite these promising results, the number of studies on the application of novel extractants in the separation of adjacent lanthanides is still limited, and further research is required for their industrial application. Fig. 11 illustrates the distribution ratio of lanthanides extracted by TiBDGA as a selective extractant for La and Ce, and DEHAPO as an extractant for selective separation of Ce4+. Table 12 summarizes various solvent extraction approaches for the individual separation of Nd, Pr, La, and Ce. It is important to note that the majority of these studies have focused on the extraction of REEs from chloride media, with relatively fewer studies investigating extraction from sulfate media. This is likely due to the fact that the formation of less-extractable anionic sulfate metallic species, such as Nd(SO4)2 − and Pr(SO4)2 − can result in inferior REEs extraction efficiency in sulfate media [132].Ionic liquids have recently received attention as a potential green alternative for conventional organic materials in solvent extraction, particularly in the separation of adjacent REEs. Along this line, Gras and colleagues [159] investigated the use of [P66614][Tf2N] and [C1C4Pyrr][Tf2N] ionic liquids for the individual separation of Ce from La, Pr, and Nd. The study involved the oxidation of Ce(III) to Ce(IV) under alkaline conditions, while the oxidation state of the other lanthanide ions remained unchanged. The lanthanide mixed hydroxides formed in the oxidation step were then dissolved in a nitric acid solution, followed by selective separation of Ce(IV) from the other elements via ILs.The extraction mechanism of non-functional ILs is based on ion exchange, in which the cationic or anionic ligand of the molecule is released into the aqueous medium. However, this mechanism can result in environmental concerns and additional costs associated with the regeneration and reuse of the ILs [160,161]. In contrast, functional ILs, which contain task-specific coordination ligands, employ a solvation-based extraction mechanism rather than an ion-exchange mechanism, which improves the extraction of metal ions and prolongs the lifetime of the ILs [162]. Khodakarami and Alagha [162] studied the separation of adjacent LREEs from a nitrate solution using two functional ILs, [A336][DHDGA] and [OcGBOEt][DHDGA] with the molecular structures illustrated in Fig. 12 . It was observed that the extractability of [OcGBOEt][DHDGA] was superior to that of [A336][DHDGA], despite the latter exhibiting superior selectivity for the individual separation of LREEs. The extraction equilibrium between REEs (M3+), nitrate ion, and the employed functional ILs can be described by Eq. (33), where Cn and A denote the cation and anion of the IL, respectively. (33) M a q 3 + + x [ C n ] [ A n ] o r g + y ( N O 3 − ) a q ⇌ [ C n ] x M [ A n ] x ( N O 3 ) y Ionic liquids have been widely utilized as solvents to enhance the extraction conditions of conventional organic extractants. Dehaudt and colleagues [163] studied the utilization of TODGA as the extractant in both organic solvent (DIPB) and ionic liquid ([Cnmim][Tf2N] and [Cnmim][BETI], n = 2,4,6,8,10) media to individually separate La, Ce, Pr, and Nd from a synthetic nitrate solution. DTPA was used as the holdback reagent (buffered with citric acid) in the aqueous phase to partition the elements between the aqueous and IL phases. The holdback reagent improves the selectivity of the extractant by retaining certain elements in the aqueous phase. The reagent forms less-hydrophobic complexes with the elements as a function of their acidity and ionic radii. As demonstrated by Dehaudt and colleagues, using TODGA diluted in [C4mim][Tf2N] and DIPB, high selectivity for the separation of La over other REEs could be achieved. The method is also suitable for the individual separation of Ce, Nd, and Pr from their mixed solution [163]. Different approaches for using ILs in the individual separation of Ce, La, Nd, and Pr are reviewed in Table 13 .After solvent extraction, adsorption/ion exchange and fractional precipitation are commonly utilized methods for the separation of REEs. Due to similarities in chemical behavior among REEs within a group (i.e. LREEs and HREEs), fractional precipitation has shown potential for group separation. However, taking advantage of different oxidation states, Eu and Ce can be individually separated from adjacent elements. Various methods for selective oxidation of Ce(III) from acidic media have been investigated, including photooxidation [172] and alkalinization [159], as well as oxidation using wet air [74,173], hydrogen peroxide [53,174–176], hypochlorite [53,176–178], and permanganate [53,176,179,180]. McNeice and colleagues [176] investigated the oxidation of Ce(III) to Ce(IV) in the mixed solution of Ce, La, Pr, Nd, and Te hydroxides using various oxidizing agents, including sodium hypochlorite, hydrogen peroxide, potassium permanganate, and Caro's acid (peroxymonosulfuric acid). Of these, KMnO4 was found to be the most effective oxidizing agent, resulting in a precipitation yield of 98.4% for Ce(IV) hydroxide, with minimal co-precipitation of other elements [176]. The application of adsorption techniques in REEs separation is primarily limited to the use of membranes and resins. Makowka and Pospiech [181] examined the separation of Ce from La (and other metals) using Cyphos 104 IL (as the extractant and ion carrier) from a nitrate solution, through the use of polymer inclusion membranes (PIMs). This study demonstrates the potential of PIMs for the competitive transport of Ce, La, Cu, Co, and Ni from the loaded ionic liquid, with the ionic liquid extracting more than 99% of Ce and 97% of La. However, the transport selectivity of the PIM for Ce and La was 67% and 15.7%, respectively, highlighting the potential of these membranes for individual separation of REEs.Reports on the application of impregnated adsorbents indicate that these materials generally exhibit low selectivity in the separation of individual LREEs. However, they possess substantial potential for the selective desorption of the elements by altering different elution parameters. Lee and colleagues [182] investigated the chromatographic separation of Ce, Pr, Nd, Sm, Zn, Al, Ca, and Fe from La using an Amberlite XAD-7 HP resin impregnated with D2EHPA. The study demonstrated that La could selectively be eluted from the loaded resin with a recovery rate of 90% [182]. In a similar study, a column packed with microcapsules containing PC 88 A extractant was employed to extract La, Ce, and Pr from a chloride solution. Within eight adsorption cycles, more than 95% of Pr was extracted while the concentration of La and Ce in the feed solution remained largely unchanged. Subsequently, La and Ce were co-extracted and then separated through controlled desorption [183]. Ashour and colleagues [184] also reported the application of this approach in using silica nanoparticles and porous microparticles functionalized with a monolayer of DTPA-derived ligands.Metal−organic frameworks (MOFs) are an emerging class of porous crystalline materials which are size-selective and are well suited for individual separation of REEs. Wu and colleagues [185] synthesized a zinc-trimesic acid (Zn-BTC) MOF covered by nanoporous graphene (NG), which resulted in high separation factors between adjacent REEs (e.g., βNd/Pr = 9.8 and βCe/La = 20.05) during extraction. This is due to the fact that the pore size of the MOF is very similar to the hydrated REE ion diameter, which allows their bare ions enter the MOF channels to coordinate with existing oxygen and form stable structures similar to hydrated REEs ions. Moreover, the presence of a 2D nanopore structure with a controlled size and the surrounding oxygen-containing groups, provides NG with higher potential for separation and purification of the elements [185]. Fig. 13 illustrates the separation factor between lanthanides for MOF/NG, ZnO/NG, and MOF alone.Despite their similar chemical properties, REEs can be separated individually through hydrometallurgical approaches. The separation process for REEs is more complex than that of other metals and often involves multiple consecutive steps. Among the REEs present in NiMHBs, only Ce possesses different oxidation states of +3 and + 4, making it more amenable to separation from other REEs. Solvent extraction is the most widely used technique in industry for the individual separation of REEs, as it offers various options for adjusting the process towards specific target elements. Techniques such as saponification of the extractant, selective scrubbing, the addition of complexing or buffering agents, and taking advantage of the synergistic effects between different extractants are commonly employed to adjust solvent extraction for REEs separation. Ionic liquids have recently gained attention as a more environmentally friendly alternative to organic compounds in the separation of adjacent lanthanides by solvent extraction techniques. ILs have been utilized as the primary extractant, solvent, or both in solvent extraction, and have demonstrated exceptional results. Additionally, the same approaches used to adjust organic extractants can also be applied to ionic liquids. The application of adsorbents and membranes in REEs separation is also an area of ongoing research, with various novel synthesized adsorbents showing superior selectivity towards REEs being introduced.The limitations and shortcomings of individual separation of La, Nd, Pr, and Ce are mainly caused by their similar chemical and physical properties, as well as the lack of specificity of the methods that can be used for their separation. Table 14 provides limitations and shortcomings of the individual separation of lanthanides.In addition to above-mentioned points, recycling of the reagents can be difficult, which can potentially increase the cost of the separation process. Furthermore, some of the separation methods are not amenable to scale-up, although they may show high separation efficiency for individual REEs in laboratory scale. In some cases, the separation process may not be fully selective and may result in the presence of impurities in the final product. The challenges of separating La, Nd, Pr, and Ce from one another are significant, and the development of new, more selective and efficient separation methods is an active area of research.Rare earth elements (REEs) are of strategic importance for the world's technological development and play a crucial role in the ongoing efforts towards a more sustainable and environmentally friendly future. However, their limited availability of economic resources and the concentration of production and supply in a few countries pose significant challenges in meeting the increasing demand for REEs. Nickel-metal hydride batteries (NiMHBs) are primarily composed of steel casing and electrode materials containing large amounts of light rare earth elements (LREEs), Ni, and Co. Due to their widespread use in rechargeable devices, recycling end-of-life NiMHBs can make a substantial contribution to addressing the global demand for REEs. Molten slag extraction is the primary pyrometallurgical approach reported for recycling NiMHBs. The efficiency of this method is highly dependent on the properties of the slag system, as almost all of the REEs present in the battery are collected by a slag phase floating on the surface of a molten Ni-based alloy. Among hydrometallurgical methods, precipitation is the most frequently reported technique for group separation of REEs from leaching solutions of NiMHBs, MHA, or REEs-rich slags, owing to its economic, versatile, and scalable nature. Solvent extraction has remained the primary technique for individual separation of REEs, although in some cases, adsorption and fractional precipitation have also shown outstanding results. Studies on the application of ionic liquids (ILs) in solvent extraction have demonstrated the significant potential of these compounds for recovering REEs from various aqueous media. Not only do ILs exhibit exceptional performances on lab scale, but they also present more environmentally friendly alternatives to conventional organic extractants and solvents. Nevertheless, additional research is required to assess their industrial viability. Individual separation of LREEs has been an arduous task for decades due to the highly similar chemical properties of these elements. In solvent extraction, conventional organic extractants are often unable to achieve high separation factors between adjacent lanthanides, and their generated wastes can be harmful to the environment. Despite these limitations, the versatility of solvent extraction for industrial applications, as well as its potential for manipulation through mixing, complexation, and modification with other agents, have made it the preferred technique for individual separation of REEs. Most available adsorption and precipitation methods have shown poor selectivity for individual REEs. However, recent advancements in synthetic materials have led to their applications in REEs recovery and separation. This includes their use as extractants in solvent extraction and adsorption techniques and agents added to the solutions to improve complexation or selective precipitation. In light of the ongoing competition to achieve high degree of separation between the elements through green processes, it is anticipated that future REEs recovery efforts will rely on the development of novel extractants and agents, as well as implementing innovative and sustainable approaches. Hossein Salehi: Conceptualization, methodology, investigation, data curation, validation, writing - original draft, visualization, software. Samane Maroufi: Conceptualization, writing - review & editing, supervision, project administration, funding acquisition. Sajjad S. Mofarah: Methodology, validation, writing - review & editing. Rasoul Khayyam Nekouei: Methodology, validation, writing - review & editing. Veena Sahajwalla: Writing-reviewing and editing, supervision, funding acquisition.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This research was supported by the Australian Research Council's Industrial Transformation Research Hub funding scheme (project IH190100009).

The recycling of nickel-metal hydride batteries (NiMHBs) has garnered significant attention in recent years due to the growing demand for critical metals and the implementation of national and international legislation aimed at achieving zero carbon emissions and reducing environmental impact. Typically, NiMHBs contain 10 wt% of rare earth elements (REEs) including La, Ce, Nd, and Pr. However, the majority of these REEs (>90%) are being discarded in landfills each year. The scarcity of these metals and the concentrated distribution of their ore deposits in only a few countries have prompted significant concern globally. One of the existing strategies to address this issue is extraction of REEs through urban mining. This study provides an in-depth fundamental and systematic review on the existing strategies and technologies for the recovery of REEs from spent NiMHBs. Further, the state-of-the-art approaches for the individual separation of La, Ce, Nd, and Pr from aqueous media are discussed, along with their corresponding challenges and shortcomings as well as the potential future directions. The research aims to provide a transformative understanding of various methods for the recovery of REEs from NiMHBs, the available techniques for the individual separation of REEs from different secondary resources, and potential improvements in the recycling process of spent NiMHBs. The outcome of this work will contribute to the development of more efficient and effective REEs recovery methods and help address the growing concern of REEs scarcity and extraction environmental impact.

This paper outlines a viable method to grow Pt/Ni nanoparticles on TiO2 nanotube surfaces, directly from solutions that resemble those available in industrial nickel processing. Global nickel resources originate from two different types of ores, sulfides and laterites, which in hydrometallurgical Ni recovery are dissolved to either chloride, sulfate or ammonia media, with or without pyrometallurgical pre-treatment steps. Ni recovery from these solutions is typically achieved by solvent extraction-electrowinning (i.e. a pure electrodeposition) route, although hydrogen reduction and different chemical precipitation methods may also be used. As a result, all Ni processing produces hydrometallurgical solutions that not only contain nickel but also a variety of other metals originally present in the ore - like cobalt, copper and precious metals - and the processes are usually optimised also to recover these other metals (Crundwell et al., 2011). In spite of this, such solutions can still contain trace amounts of precious metals, which are not recovered from solutions due to a high concentration ratio between the base metal (such as Ni) and trace metals (such as Pt).This is of particular importance in view of the global scarcity of metal resources that has become an imminent concern (Peck et al., 2015) and thus, improved recycling and recovering of metals from impure process solutions, acidic industrial side-streams and other waste streams is becoming a necessity. At the same time, emerging technologies like fuel cells may increase the demand for Pt as a catalyst material even further (Peck et al., 2015) and therefore, identification of alternative (secondary) raw material sources for platinum group metal (PGM) based catalysts is essential.Ni-Pt combinations have been proved to be extremely effective catalyst materials (Debe, 2012; He et al., 2011, 2013; Kühl and Strasser, 2016; Shao et al., 2016) and thus, process and side-streams associated with Ni metal (McDonald and Whittington, 2008a,b) could be potential raw material sources when trace amounts of Pt is present, especially as annual global nickel production is over 2000 Gg (i.e. 2 million tonnes) and expected to increase up to 140–175% by 2025 (Elshkaki et al., 2017). However, due to the high concentration ratio between Ni and precious metals, the hydrometallurgical solutions in nickel production have not been considered suitable for recovery of precious metals, and even less, for production of a high-added value Pt catalyst.On the other hand, the toxic nature of nickel solutions – which are inevitable in hydrometallurgy in order to fulfil the ever increasing global nickel demand –would require new methodologies of how to fully exploit these solutions and also, the minor components present in them. The results presented here show that utilising redox replacement reactions the formation of Pt/Ni catalysts from such solutions is indeed possible, without any additional chemicals.Usually the redox replacement reaction is used in conjunction with electrochemical deposition and this is the case in our studies (Fig. 1 ). Firstly, Ni is electrodeposited and then, the redox replacement reaction takes place spontaneously between electrodeposited Ni and Pt2+ ions due to the difference in the reduction potentials: the deposited Ni is oxidised by Pt2+ and dissolved back to solution as Ni2+, whilst Pt2+ is reduced to Pt and deposited on the electrode. The half-reaction for the oxidation of Ni is shown in Reaction 1a and the reduction of Pt2+ in Reaction 1b, together with their respective standard electrode potentials (E° vs. SHE). (1a) N i s → N i a q 2 + + 2 e − E° ( ox ) = + 0 . 25 V vs . SHE (1b) P t a q 2 + + 2 e − → P t s E° ( red ) = + 1 . 20 V vs . SHE The overall redox reaction is displayed in Reaction 2, together with the standard reaction potential, which is the driving force for the spontaneous redox replacement reaction: (2) N i s + P t a q 2 + → N i a q 2 + + P t s E° ( RR ) = + 1 . 45 V vs . SHE In contrast to other electrochemical-redox replacement methods like electrochemical-atomic layer deposition (e-ALD) (Gregory and Stickney, 1991; Vaidyanathan et al., 2006) or surface-limited redox replacement (SLRR) (Brankovic et al., 2001), the goal here is not to deposit smooth monolayers but utilise electrodeposition-redox replacement method (EDRR) to create functional Pt/Ni catalysts directly from solutions resembling hydrometallurgical Ni process solutions.Moreover, this one step approach differs from those presented in literature about Pt/Ni surfaces prepared by EDRR (Papadimitriou et al., 2008, 2010; Rettew et al., 2009; Tegou et al., 2010; Wang et al., 2011; Zhang et al., 2012) or electroless deposition-redox replacement (Tamašauskaitė-Tamašiūnaitė et al., 2013, 2014). In these previous studies, all the solutions have been optimised for the application, i.e. Pt and Ni concentrations of synthetic solutions are tailored for the most effective catalyst formation and they do not represent industrial solutions. Conversely, this paper demonstrates – for the very first time – that non-optimised solutions and side-streams could be a potential raw material sources for functional Pt/Ni catalyst surfaces. These solutions are indeed challenging as in hydrometallurgical Ni process solutions the Ni content is high and Pt content extremely low: for a 10 ppm Pt solution, Ni/Pt concentration ratio is typically ≈ 20,000.Earlier, the authors have demonstrated that Ag can be recovered from Zn based hydrometallurgical solutions by EDRR (Halli et al., 2017) and Au from copper based solutions (Korolev et al., 2018) whereas this current paper goes beyond that approach: instead of pure recovery of a precious metal, functional Pt/Ni catalytic surfaces are produced and used without any further modifications for photocatalytic H2 generation.From an industrial perspective, the possibility to utilise under sourced side-streams and process solutions for catalyst production makes EDRR already very attractive method but it has also further advantages: EDRR does not demand any additional chemicals, unlike cementation (precipitation) or solvent extraction traditionally used in hydrometallurgy for metal recovery, no neutralization chemicals are needed either, and when compared to electrowinning (pure electrodeposition), EDRR is more effective with solutions of trace amount of precious metals. This all makes the EDRR method more sustainable than the competing methods.Overall, this research demonstrates a new route for the exploitation underutilised industrial side-stream solutions, which not only leads to the formation of catalytic surfaces for clean energy production but also has the added benefit of reducing/eliminating the presence of potentially toxic material (Pt) from industrial Ni processing. Furthermore, this method provides a platform for the cost and material competitive large-scale catalyst production based on the principles of circular economy.Flat TiO2 surfaces were prepared by anodising Ti foil (99.9%) at 20 V for 15 min in 0.5 M H2SO4. In contrast, TiO2 nanotubes were prepared by the immersion of Ti foil in a tri-ethylene glycol electrolyte consisting of 0.3 M NH4F and 3 M H2O, at 60 V at 60 °C for 15 min. After anodising, all the samples were annealed in air at 450 °C for 1 h.Ni nanoparticles were prepared on TiO2 surfaces by the electrodeposition – redox replacement (EDRR) method (Ivium CompactStat) from a solution containing 60 g/l nickel (from NiSO4∙6 H2O, ACS grade, Sigma-Aldrich) and 10 g/l H2SO4 (95–98%, p.a., Carl Roth), while Pt/Ni nanoparticles were deposited from the same base solution (60 g/l nickel + 10 g/l H2SO4) having 10 ppm or 100 ppm Pt (from 1000 mg/l AAS standard, Sigma-Aldrich). Either flat TiO2 or TiO2 nanotube surfaces – with a geometric area 0.5 cm2 - were used as the working electrode, with a Pt sheet (6 cm2) as the counter electrode and Ag/AgCl in 3 M KCl as the reference electrode. The distance between working and counter electrode was 2 cm and the volume of the solution was 20–25 ml.In EDRR method, the electrodeposition step (ED step) was performed galvanostatically and consisted of a total 74 short cathodic and anodic current pulses. After 37 cathodic-anodic pulse pairs, a redox replacement (RR) step was performed. During this step, no external current or voltage was applied but open circuit potential (OCP) was recorded until a pre-determined time had elapsed. After this, the EDRR cycle was repeated - first with 37 cathodic-anodic pulse pairs followed by a RR step.In the case of the flat TiO2 surfaces, the cathodic and anodic current pulses had a duration of 10 ms each and the current density was −100 mA/cm2 and +20 mA/cm2, respectively. The RR step time was either 10, 30 or 60 s and the number of full EDRR cycles was varied (10, 20 or 30 cycles). For TiO2 nanotube surfaces the conditions were modified, such that the cathodic pulse durations were 4 s at −30 mA/cm2 and anodic pulses 10 ms at +30 mA/cm2, to reflect the higher surface area and lower conductivity of the tubes. The associated RR time was set to be between 60 and 240 s and the number of cycles was either 10 or 20 cycles.Field-emission scanning electron microscope (FE-SEM Hitachi S4800) was used to characterize the morphology of the samples, whereas X-ray photoelectron spectroscopy (XPS, PHI 5600) provided the chemical composition of the samples. In XPS, the signal intensity was divided by a relative sensitivity factor (RSF) and normalized over all of the elements detected. All data processing was performed using MultiPack v.9.6.0 software.Photocatalytic H2 generation measurements were conducted by irradiating the TiO2 samples with an AM 1.5 solar simulator (100 mW/cm2) in a quartz tube containing a 20 vol% ethanol-water solution for 5 h. The amount of produced H2 was measured by using a gas chromatograph (GCMS-QO2010SE, Shimadzu) equipped with a thermal conductivity detector and a Restek micropacked Shin Carbon ST column (2 m × 0.53 mm). The quartz reactor was purged with N2 gas for 10 min to remove O2 prior to the initiation of the photocatalytic experiments.Pt/Ni nanoparticles are formed on TiO2 surface during electrodeposition-redox replacement (EDRR) cycling and a typical EDRR measurement is shown in Fig. 2 : short cathodic + anodic current pulses (during the ED step) are followed by redox replacement (RR step) and the EDRR procedure is cycled a number of times. The EDRR profiles for all samples are shown in Supporting Information (Fig. S1).The ED step (electrodeposition) consists of galvanostatic pulsing between cathodic and anodic currents. Firstly, Ni and possibly some Pt is deposited during the short cathodic current pulse, though simultaneous H2 evolution - that disturbs the deposition - may also take place. This is overcome by applying a short anodic current pulse that not only results in hydrogen desorption, but also makes more surface sites available for deposition in the following cathodic pulse (Kollia et al., 1990; Spanou and Pavlatou, 2010). During the RR step (redox replacement) there is a spontaneous replacement of deposited Ni with Pt, due to the difference in electrochemical oxidation/reduction potentials - Pt2+ oxidises the electrodeposited Ni to soluble Ni2+ while it itself is simultaneously reduced to Pt and deposited to the surface (see Reactions 1–2). Fig. 3 highlights the need of redox replacement (RR) step in the nanoparticle formation. The galvanostatic pulsing (i.e. ED step) alone is not an effective method for the formation of Pt/Ni nanoparticles as is clearly observed when comparing SEM images of a Pt/Ni nanoparticles prepared with galvanostatic pulsing to those prepared by EDRR method. The used parameters were the same in both of these cases, the only difference being that EDRR has an additional RR step/cycle. For comparison, Fig. 3 also shows the deposition of pure Ni nanoparticles by EDRR method and it is evident that although Pt is not necessary for the nucleation of particles on the surface, it improves it. In addition, the presence of Pt results in a characteristically jagged appearance of the nanoparticles (see Supporting Information, S2). This observation is probably a result of Pt growth on the nanoparticles, both during the redox replacement step and possible co-deposition on previously replaced Pt in the subsequent ED steps.The positive effect of redox replacement step on nanoparticle formation is partly due to Ostwald ripening, leading to the presence of larger surface features, and partly due to the replenishing of solution nearby the electrode, reducing the possibility of mass-transfer limitation in ED step. Natter and Hempelmann (2003) have found a similar observation with pulse electrodeposition when varying t off (i.e. short current-off time between deposition pulses) for Au nanoparticles, i.e. the size of nanoparticles grew with longer t off time. It is important to note that - in addition to different materials and solutions (Au in literature (Natter and Hempelmann, 2003) cf. Pt/Ni presented here) - t off time has a different purpose than RR time. In pulse electrodeposition, t off is applied only for milliseconds between short deposition pulses and pulsing is performed in a single metal electrolyte in order to replenish the surface from adsorbed hydrogen, while RR time is clearly longer and performed after electrodeposition step in multi-metal electrolyte in order to redox replacement reaction to take place. As a result, enrichment of the more noble metal on the surface takes place.The effect of the initial EDRR cycles on Pt/Ni nanoparticle nucleation is presented in Fig. 4 , which comprises of Pt/Ni nanoparticles deposited to flat TiO2 surface using a single cycle or 5 cycles (the redox replacement time: 30 s). As can be seen, already after a single EDRR cycle particles have nucleated on the surface, although the particle size is relatively small. When the number of cycles is increased, both the particle density and size of the particles increase substantially and start to show the characteristic jagged appearance. Data from XPS shows that the at-% of Pt is 0.29 after a single cycle and it increases to 5.77 at-% after 5 cycles, whereas of Ni at-% remains low (0.77% after 1 cycle cf. 0.87 at-% after 5 cycles), suggesting that even if Pt may co-deposit with Ni during electrodeposition step, it is deposited primarily during the redox replacement step.The formation of Pt/Ni nanoparticles is further studied as a function of number of cycles and RR time (Fig. 5 – SEM and Fig. 6 - XPS). The associated potential profiles of EDRR (Fig. S1) are shown in Supporting Information.From Fig. 5 it is seen that the size of nanoparticles increases both as a function of number of cycles and RR time. As previously discussed (Fig. 3), the positive effect of RR time on the size and nanoparticle density can be associated with Ostwald ripening and replenishing the solution nearby the electrode: the dissolution of Ni from the surface during RR step leads to a higher local concentration in the vicinity of the electrode, further diminishing the possible mass-transport limitation in the following ED step. Fig. 6(a) shows examples of Ni2p and Pt4f spectra for a sample prepared from 100 ppm Pt solution on a TiO2 surface using the EDRR method (the redox replacement time was 30 s and number of cycles 30). As can be seen, the Pt4f region has well separated spin-orbit components (Δmetal = 3.35eV).The atomic-% (and weight-%) of Pt was determined by considering the doublet peak of Pt4f region, which can be de-convoluted into four peaks. The presence of two main peaks (69.75 eV and 73.12 eV) is ascribed to the Pt0 4f7/2 and Pt0 4f5/2, while the other small peaks at 71.0 eV and 75.26 eV correspond to Pt2+ 4f7/2 and Pt2+ 4f5/2. Ni2p peak, on the other hand, has split spin-orbit components (Δmetal = 17.3eV) that comprise of core level and satellite features, which can be resolved into eight peaks. Two peaks are located at 851.25 eV (Ni0 2p3/2) and 868.5 eV (Ni0 2p1/2), indicating the presence of Ni metal. Another two peaks at 852.40 eV (Ni2+ 2p3/2) and 869.76 eV (Ni2+ 2p1/2) are ascribed to the NiO. The other two peaks at 855.14 eV (Ni2+ 2p3/2) and 872.44 eV (Ni2+ 2p1/2) are due to the formation of Ni(OH)2. The peaks at 860.44 eV and 879.06 eV are the satellite peaks.As can be observed from Fig. 6(b–c), also the atom-% of Pt (and the respective weight-% of Pt, both determined by XPS) increases with longer RR times. This is due to the mass-transfer limitations related to the low Pt concentration: the mass-transfer quickly limits the redox replacement reaction when Pt content is present only in ppm levels and thus, increasing the replacement time provides longer time for Pt to reach the electrode surface, resulting in higher Pt at-% on the surface. In order to demonstrate more clearly the purity of the end-product, the Pt/Ni ratio is calculated from at-% - Fig. 6(b) – and demonstrates that higher RR time results in higher end-product purity.The effectiveness of the EDRR, on the other hand, is best discussed in terms of enrichment, which is calculated by comparing the weight-% of Pt on the particles (determined by XPS) to weight-% of Pt in the solution. It is remarkable how effective EDRR is when compared to pure galvanostatic pulsing (Figs. 3 and 6c). For example for a 100 ppm (1 ppm = 0.0001 wt-%) Pt solution, utilising 60 s RR time results in 42 weight-% of Pt on the surface after only 20 EDRR cycles, and this translates to over 4,200-fold enrichment. In comparison, the reference sample shown in Fig. 3 (30 cycles of galvanostatic pulsing and no RR steps) has a significantly lower Pt content (4.6 wt-%), resulting in a decade lower (460) enrichment, even if the current input in the reference sample is higher (20 cycles in EDRR cf. 30 cycles in galvanostatic pulsing).To further demonstrate the industrial feasibility of the EDRR method, the formation of nanoparticles was performed from a solution containing only 10 ppm of Pt but with the same Ni concentration (60 g/l of Ni), resulting in concentration ratio Ni/Pt = 20,000. As can be seen in Fig. 7 , the formation of Pt/Ni nanoparticles is successful even with such a low concentration of Pt in solution. Moreover, XPS analysis showed that with 60 s replacement time 15 wt-% of Pt was deposited on the surface, a level that is over three time higher than obtained by the pure galvanostatic pulsing (reference sample) from the 100 ppm Pt solution. In the terms of the enrichment, EDRR results in over 10,000-time enrichment of Pt from 10 ppm solutions, suggesting that the method is extremely feasible with the solutions with a low level of precious metals.In order to investigate further the abilities of EDRR in the formation of the high-value added products, experiments were performed using TiO2 nanotube substrates (Fig. 8 – SEM and Fig. 9 - XPS). TiO2 nanotube surfaces are promising candidates for the photo-catalytic applications as the tubular configuration provides a high light absorption pathway and aids the prevention of the recombination of photo-generated electron/hole pairs (Tong et al., 2012). Moreover, TiO2 nanotubes have demonstrated drastically enhanced photocatalytic activity in numerous studies when the nanotubes are decorated with “co-catalyst” metal nanoparticles (Christoforidis and Fornasiero, 2017; Liang et al., 2013; Ni et al., 2007; Papadimitriou et al., 2008; Park et al., 2013). Figs. 8–9 demonstrate that the size and the amount of Pt/Ni nanoparticles on TiO2 nanotube surface can indeed be controlled by RR time and cycling when using the EDRR method (see also Fig. S2b). The results also show that particles not only nucleate on the top of the nanotubes but also on the outer walls, allowing the exploitation of the 1D nature of the tubes for photocatalysis: the main advantages of 1D materials are the enhanced light absorption combined with short reaction paths of photogenerated carriers (Xiao et al., 2015; Altomare et al., 2016; Nguyen et al., 2015, 2016). Fig. 9(a) shows again an exemplar of the XPS data fitted for the Ni2p and Pt4f regions and the detected species were Pt0, Pt2+ and Ni species. Furthermore, the EDRR method allows the control over the Pt/Ni ratio (Fig. 9(b), at-% determined by XPS) which has been shown to be a critically important factor in the electrocatalysis of oxygen reduction reaction in numerous studies (Jiang et al., 2017; Toda et al., 1999; Yang et al., 2004).In order to demonstrate that the EDRR method could be used to produce photocatalytic surfaces from hydrometallurgical base metal streams, proof-of-concept measurements of photocatalytic H2 generation with the prepared Pt/Ni nanoparticle - TiO2 nanotube surfaces were performed. Fig. 10 shows that these surfaces indeed possess significant activity for H2 evolution, the highest being an over 30-fold enhancement (the redox replacement step = 240 s and number of cycles = 20) when compared to a pure TiO2 nanotube surface. When the photocatalytic activity is compared to TiO2 nanotube surfaces covered with pure Ni nanoparticles, the H2 evolution levels are similar for fresh samples (see Supporting Information, S3). However, pure Ni nanoparticles suffers from aging whereas Pt or Pt/Ni are highly stable against oxidation (S3). Moreover, the catalytic activity shown here is comparable with literature, e.g. our results show a similar H2 evolution rate under 1.5 AM solar illumination as that obtained for atomic layer deposited Pt as co-catalysts on TiO2 nanotubes (Yoo et al., 2018). The results also indicate that the Pt/Ni ratio is critical for H2 evolution: the sample with highest H2 production has also the clearly highest Pt/Ni ratio ≈ 8, while for all the other surfaces the Pt/Ni ratio ≈ 2 (Fig. 9(b)). All these samples with a similar Pt/Ni ratio have also similar hydrogen evolution rates (Fig. 10), thus indicating the critical role of Pt/Ni ratio in the particles.As the presented EDRR method is particularly powerful in tuning the Pt/Ni composition, these results are very promising in the view of preparing photocatalytic surfaces directly from sulfate based process streams or side streams of hydrometallurgical Ni metal plants. EDRR is truly attractive approach for the industrial solutions that contain only a small amount of platinum group metals (PGMs), especially as Pt/Ni nanoparticle formation consumes electricity only during the Ni deposition steps while Pt is “enriched” on the surface via an electroless redox replacement reaction, thus enhancing the process economics. It is also worth noting that as EDRR was successful from solutions with Ni/Pt ratio as high as 20,000, the industrial Ni process solutions and side-streams containing trace amounts of PGMs could potentially be “a platinum mine” for clean energy technologies if a future industrial process was developed. For example, in an average size base metal plant 10 m3/h or even 100 m3/h of such solutions are flowing in the processes, and with such amounts the large-scale production of catalysts – which has been identified as one of the main task of catalyst development (Debe, 2012) - can truly become possible.Real hydrometallurgical solutions also contain other metals than Ni and Pt and these will influence both the EDRR process and resultant surface. Previous results of Ag recovery from Zn/Ag solutions [Yliniemi et al., 2018] have shown that although the presence of Fe3+ as in impurity may slightly reduce enrichment efficiency, it may improve product purity. This is most likely due to the selective dissolution of Zn by Fe3+ as the reduction potential of Fe3+/Fe2+ (E° = 0.77 V vs. SHE) is higher than that of Zn2+/Zn. Thus, both Fe3+ and Ag+ can oxidise Zn to Zn2+ but only Ag is enriched on the surface as Fe3+ is reduced to soluble Fe2+. Similar behaviour is expected in Ni/Pt solution and when considering performance, this may actually result in increased catalytic activity. Therefore, EDRR performed in hydrometallurgical solutions has huge potential for the recovery of precious metals like Ag (Halli et al., 2017; Yliniemi et al., 2018) and Au (Korolev et al., 2018). Furthermore, this paper shows that EDRR can also directly produce functionalised surfaces with Pt or other trace metals present in hydrometallurgical solutions. Moreover, EDRR method allows control over not only particle size and density, but more importantly, over the precious metal/base metal ratio (here Pt/Ni) in the particles, i.e. EDRR provides also a control over the catalytic activity of these surfaces.The EDRR process outlined here is an effective method for the production of catalytic surfaces and simultaneously, exploiting the hydrometallurgical solutions fully by utilising also the minor components (such as Pt) present in them. Remarkably, the results presented here show a 10,000-time enrichment level of Pt onto the surface when Pt/Ni nanocatalysts are formed from solution simulating hydrometallurgical process streams on TiO2 surfaces by EDRR. It is also noteworthy, that such an enrichment is possible without any additional use of chemical or further modifications.By adjusting the different EDRR parameters (number of cycles or redox replacement time), the surface characteristics of the resultant catalytic nanoparticles can be tuned to control the desirable properties like nanoparticle size and distribution. Moreover, preliminary results of the produced Pt/Ni co-catalytic surfaces for photocatalytic H2 evolution demonstrated the level of performance that is comparable to the standard procedures for co-catalytic Pt/TiO2 surfaces found in literature.Overall, the findings offer a more sustainable circular economy platform, where minor concentrations of valuable metals present in base metal production solutions are used for the preparation of high-value products to be used as photocatalysts in clean energy sector. Academy of Finland (NoWASTE - project no: 297962), Finland; Technology Industries of Finland Centennial/Jane and Aatos Erkko Foundation (Future Makers: Biorefinery Side Stream Materials for Advanced Biopolymer Materials - BioPolyMet), Finland; ERC, European Union; DFG (the Erlangen DFG cluster of excellence EAM, project EXC 315 (Bridge) and funCOS), Germany.The following is the supplementary data related to this article:Electrochemical – Redox replacement (EDRR) profiles during Pt/Ni nanoparticle formation (Fig. S1), appearance of Pt/Ni nanoparticles (Fig. S2) and photocatalytic activity for H2 generation by pure Ni nanoparticles on TiO2 nanotubes and Pt/Ni nanoparticles on TiO2 nanotubes (fresh and aged samples, Fig. S3). Multimedia component 1 Multimedia component 1 Supplementary data related to this article can be found at https://doi.org/10.1016/j.jclepro.2018.08.022.

Solutions that simulate hydrometallurgical base metal process streams with high nickel (Ni) and minor platinum (Pt) concentrations were used to create Pt/Ni nanoparticles on TiO2 nanotube surfaces. For this, electrochemical deposition – redox replacement (EDRR) was used that also allowed to control the nanoparticle size, density and Pt/Ni content of the deposited nanoparticles. The Pt/Ni nanoparticle decorated titanium dioxide nanotubes (TiO2 nanotubes) become strongly activated for photocatalytic hydrogen (H2) evolution. Moreover, EDRR facilitates nanoparticle formation without the need for any additional chemicals and is more effective than electrodeposition alone. Actually, a 10,000-time enrichment level of Pt took place on the TiO2 surface when compared to Pt content in the solution with the EDRR method. The results show that hydrometallurgical streams offer great potential as an alternative raw material source for industrial catalyst production when coupled with redox replacement electrochemistry.

Data will be made available on request.Carbon dioxide (CO2) emissions to the atmosphere exhibit an economic burden and an environmental threat due to their significant contribution to climate change and global warming [1, 2]. In 2021, according to the International Energy Agency (IEA) the global CO2 emissions were projected to climb by 5% with a total of 1.5 billion tons. This would be the second-highest increase in history and the largest yearly increase in emissions since 2010 [3]. Consequently, different strategies including deploying carbon capture and storage (CCS) and carbon capture and utilization (CCU) were deployed to control the emissions of massive amounts of CO2 to the atmosphere and to replace depleted fossil fuels in the future. Reusing and recycling CO2 can contribute to decreasing the effects of global warming and the production of renewable fuels like methanol.The use of CCS has been a trending topic in industry and literature over the last three decades given the role of this technology in minimizing industrial CO2 emissions to the atmosphere [4–6]. The CCS is an environmental solution supporting the project’s environmental sustainability rather than providing economic value. Consequently, carbon utilization technologies (CCU) have been studied for deployment within oil and gas infrastructures to sustain projects' environmental and economic importance. In CCU, captured and treated CO2 can be utilized for enhanced oil recovery to increase production or as a feedstock in different industries such as food, chemicals, and fuels. Hence, captured and treated CO2 can be either liquefied or compressed for direct selling to different sectors or utilized within the same plant by deploying CO2 monetization technologies to value-added products.Conversion of CO2 to methanol has been considered among the most favorable CO2 utilization processes in the industry in the last ten to fifteen years due to the maturity and stability of the investigated catalytic systems. [7–10]. Methanol is a liquid chemical that can be used as (i) a solvent; (ii) feedstock for producing chemicals such as acetic acid, methyl tert-butyl ether (MTBE), formaldehyde, and dimethyl ether (DME), or (iii) as a cleaner fuel in the transportation sector. In the transportation sector, pure methanol can be used directly as a marine or vehicle fuel, or blended with gasoline for vehicle utilization. Currently, Asia Pacific holds the largest market share of methanol due to the rapid increase in methanol consumption in the automotive, construction, and electronics industries. The global methanol market is projected to reach $26 billion by 2025, with a compound annual growth rate of 6.6% from 2019 to 2025 [11]. Traditionally, methanol has been produced from fossil fuels such as natural gas and coal. However, methanol production from captured CO2 is an emerging route aiming to assure a sustainable production of methanol after the depletion of fossil fuels and to support the efforts to control and mitigate CO2 emissions.The feasibility and efficiency of different chemical, electrochemical, and thermochemical reactors for methanol production have been studied in the literature in the past few years. For instance, Kim et al. [12] assessed the energy efficiency and economic feasibility of a solar-based process for the production of methanol from CO2 and water based on two-step routes. In the proposed process, the first step consists of a thermochemical reactor utilizing solar energy for converting CO2 to CO using a water gas shift reaction. Synthesis gas consisting of CO and H2 is then fed to a methanol catalytic reactor for methanol synthesis. The study concluded that the two-step solar-thermochemical pathway is a promising approach for CO2 utilization to fuels. However, the solar reactor sub-system is capital intensive, and much work must be done to improve the process from an economic perspective. On the other hand, Al-Kalbani et al. [13] modeled and compared the energy assessment of methanol production from CO2-based chemical vs electrochemical production processes. The authors reported that methanol production based on high-temperature CO2 electrolysis has double energy efficiency as CO2 hydrogenation utilizing H2 produced via water electrolysis. Nonetheless, from an economic perspective, CO2 hydrogenation to a methanol-based chemical route is found to be the most feasible solution.Methanol can be produced via catalytic CO2 hydrogenation using homogenous and heterogeneous catalyst systems wherein the reaction pathway mainly depends on the catalyst. To date, different research studies have investigated or developed the conversion of CO2 to methanol on a pilot scale using heterogeneous catalysts [14–17]. In 1996, the first commercial low-pressure methanol synthesis process was patented with process conditions below 150 bar and 300 °C using a Cu/Zn-based catalyst [18]. Since then, Cu/ZnO/Al2O3 catalysts have been widely used in industry and studied in the literature for methanol synthesis due to the superior advantages, where the promotor ZnO provides a high dispersion and stabilization level of Cu active sites and the metal oxide (Al2O3) provides support for the catalyst [19,20]; hence, contributing to enhancing methanol production reaction. Different industrially mature catalyst has been studied for CO2 hydrogenation to methanol from synthesis gas in the temperature and pressure ranges of 210–250 °C and 50–100 bars. In this regard, Van-Dal and Bouallou [21] designed and simulated a CO2 hydrogenation to methanol plant combined with a CO2 capture unit and hydrogen production unit using Aspen Plus software. In the CO2 hydrogenation section of the plant, Cu/ZnO/Al2O3 catalyst was used in the adiabatic reactor. The steam formed in the methanol synthesis unit was utilized as a CO2 capture unit. Atsonios et al. [22] investigated a techno-economic analysis of methanol production from CO2 hydrogenation using a membrane reactor. The study focused mainly on exploring the most economical operation conditions for captured CO2 utilization to methanol and revealed that hydrogen production costs largely influence the economic feasibility of the process. A thermo-economic approach for methanol production from different renewable sources was proposed by Rivarolo et al. [23]. The authors reported two plant configurations, for which CO2 is obtained from biogas or purchased from an external plant. Furthermore, the study mainly focused on electricity generation from renewable hydroelectric, wind, or photovoltaic plants and investigated the option of purchasing electricity if renewable resources are not available.Similarly, Bellotti et al. [8] reported a thermo-economic feasibility study of a power-to-fuel plant for methanol production from CO2. The studied system consists of a methanol production plant, a hydrogen production plant from water electrolysis, and an amine-based CO2 capture unit. The authors concluded that selling the by-product O2 is essential for economic feasibility results. However, neither this study nor the above-mentioned studies have considered the proposed plant's design and simulation. Other studies in the literature either focused on the reaction pathway and kinetics of CO2 hydrogenation to methanol over Cu/ZnO/Al2O3 [24–28], or elaborated on catalyst preparation, catalyst formulation, and reaction mechanisms for CO2 hydrogenation to methanol over different catalysts [29–36]. To the best of the authors’ knowledge, limited studies in the literature provided a comprehensive analysis of the CO2 hydrogenation process to methanol, wherein the majority of these studies focused on thermo-economic aspects of plant configuration or the feasibility of employing different technologies to support the feasibility of methanol production. Moreover, operating conditions such as temperature, pressure and H2/O2 feed ratio, and reactor types have not been intensively examined in previous studies. Consequently, The purpose of this work is to assess the technoeconomic-environmental feasibility of CO2 hydrogenation to the methanol process using the commercial catalyst, Cu/ZnO/Al2O3, with updated operating conditions for improved CO2 conversion. The methanol synthesis process is modeled and simulated using the commercial software Aspen Plus V11. Both isothermal and adiabatic reactors are studied for methanol synthesis under fixed feed conditions and catalyst specifications. Additionally, a sensitivity analysis is considered to investigate the influence of temperature, pressure, and variable hydrogen feed on the methanol yield. The optimized process is finally evaluated under environmental and economic aspects. In comparison with other studies that consider captured CO2 utilization for methanol production, this study emphasizes on the practicality and profitability of deploying CO2 to methanol monetization infrastructure within the biomass value chain for direct utilization of liquid pure CO2 by-product produced from a cryogenic biogas upgrade process. The proposed process could also be employed for the CO2 utilization from petrochemical processes where the raw materials (CO2 and H2) are available.The proposed process utilizes 76.46 kmol/hr of CO2 by-product produced from a cryogenic biogas upgrading process at 12.3 °C and 47.63 bar within the same plant, and 535.22 kmol/hr of hydrogen supplied at 25 °C and 30 bar. Depending on the plant capacity and purchase price of hydrogen, hydrogen can be supplied from a renewable source such as water electrolysis, or a fossil-fuel resource such as natural gas or coal. The deployment of a CO2 monetization process to value-added methanol within the biogas value chain achieves two main targets: (1) minimizing CO2 emissions, and (2) enhancing the economic performance of the biogas value chain. The proposed configuration of the full-scale biogas upgrading process is illustrated in Fig. 1 . Although the proposed CO2 hydrogenation to methanol is connected to the biogas process, it could also be employed for the CO2 utilization from petrochemical processes where the required raw materials (CO2 and H2) are available. Fig. 2 presents the process used for CO2 hydrogenation to methanol. The process consists of a feed preparation section to meet the required conversion conditions; a reactor section where catalytic CO2 conversion takes place and a purification section to produce methanol with purity ≥ 98%. The methodology for modeling the methanol synthesis process was mainly inspired by Van-Dal and Bouallou [21]. For designing and simulating the CO2 hydrogenation process in Aspen Plus, Redlich-Kwong (RKS) equation of state can be optimally used to simulate the process kinetics as reported previously in the literature [37,38]. In contrast, other studies reported using different equations of states, such as Non-Random Two Liquid (NRTL) [7], or employed more than one equation of state, such as RKSMHV2 and NRTL-RK based on the stream pressure [21].The following is a thorough description of each of the sections in this process.In the first section of the conversion process, the feeds (CO2 and H2) are compressed to 78 bar, mixed, and heated to 210 °C to meet the reactor inlet specifications. An advantage of utilizing CO2 produced from the cryogenic biogas upgrade process is that the produced CO2 is at high pressure and does not require high power for compressing compared to other proposed processes in the literature where multi-stage compressors were used [21,37,39,40]. Additionally, the CO2 is supplied with high purity and does not require any treatment before feeding into the conversion process. If CO2 is supplied from other industries the treatment and compression work should be considered.In this section, CO2 is mixed with /H2 then compress to 75.7 bar and heated to 210 °C, and fed to the fixed bed plug flow reactor to produce methanol. Adiabatic and isothermal operating conditions were tested in this study. The reactor system contains 44,500 kg of Cu/ZnO/Al2O3 catalyst. The properties of the Cu/ZnO/Al2O3 are summarized in Table 1 .Two parallel exothermic reactions Eq. (1) and (2) take place inside the reactor to produce methanol alone with an endothermic reverse-water gas shift (RWGS) reaction Eq. (3): (1) C O 2 g + 3 H 2 g ⇌ C H 3 O H l + H 2 O g Δ H = - 87 K J / m o l ( 25 ° C ) (2) C O g + 2 H 2 g ⇌ C H 3 O H l Δ H = - 128 K J / m o l ( 25 ° C ) (3) C O 2 g + H 2 g ⇌ C O g + H 2 O g Δ H = 41 K J / m o l ( 25 ° C ) In the presence of the catalyst, the used kinetic model is based on Langmuir-Hinshelwood- Hougen- Waston (LHHW) mechanism and assumes CO2 as the primary source for methanol production in the presence of the RWGS reaction [41]. The kinetic model parameters were further modified by Mignard and Pritchard [42] to application ranges up to 75 bar as shown in Eq. (4) and (5), where the pressure is in bar and temperature in K. The kinetic constants, Eq. (6), follow the Arrhenius law, while the thermodynamic equilibrium constants, Eq. (7) and (8), are given by Graaf et al. [43]: (4) r C H 3 O H = k 1 P C O 2 P H 2 ( 1 - 1 k e q 2 P H 2 O P C H 3 O H P H 2 3 P C O 2 ) 1 + k 2 P H 2 O P H 2 + k 3 P H 2 0.5 + k 4 P H 2 O 3 m o l k g c a t s (5) r R W G S = k 5 P C O 2 ( 1 - 1 k e q 1 P H 2 O P C O P C O 2 P H 2 ) ( 1 + k 2 P H 2 O P H 2 + k 3 P H 2 0.5 + k 4 P H 2 O ) m o l k g c a t s (6) k i = A i exp B i RT (7) log 10 1 K e q 1 = 2073 T + 2.029 (8) log 10 K e q 2 = 3066 T - 10.592 The equations(1 to 8) were rearranged in alignment with the type of accepted kinetic equations in Aspen Plus software and represented in equations (9 to 11). Table 2 summarizes the model parameters used in the Aspen Plus software [21]. (9) r C H 3 O H = k 5 P C O 2 - k 6 P H 2 O P C H 3 O H P H 2 - 2 1 + k 2 P H 2 O P H 2 - 1 + k 3 P H 2 0.5 + k 4 P H 2 O 3 m o l k g c a t s (10) r R W G S = k 5 P C O 2 - k 7 P H 2 O P CO P H 2 - 1 1 + k 2 P H 2 O P H 2 - 1 + k 3 P H 2 0.5 + k 4 P H 2 O m o l k g c a t s (11) ln k i = A i + B i T Additionally, a multi-tube reactor (# of tubes = 1000 tubes, length = 5 m, and diameter = 1 m) was considered. A pressure drop of 0.6 bar is allowed through the reactor, with an outlet stream leaving the reactor at 75 bar. Exact specifications are applied for the isothermal reactor, with a constant reactor operating temperature of 210 °C.Gases leaving the reaction were collected in the knock-out drum (KO101) to separate the products from unreacted reactants. Unreacted gases are recycled back to the reactor to enhance conversion and part of it is purged to the atmosphere at a split fraction of 0.1 to avoid by-product accumulation. The produced liquid methanol leaves the knock-out drum at 73.4 bar. Two parallel valves are used to reduce its pressure down to 1.2 bar before entering the flash drum (FLT101) for further purifications and removal of unreacted gases. The outlet liquid methanol from the flash drum at 1.2 bar and 14.95 °C is heated up to 80 °C before sending it to the distillation column (D101) for methanol/water separation. A distillation column with 15 stages and a reflux ratio of 2.12 is utilized to produce a high-grade methanol product at 64.92 °C and 1 bar, and water by-product at 101.91 °C and 1 bar. No pressure drop is assumed across the distillation column. The produced methanol is then compressed to 80.29 bar, heated to 80.29 °C, and fed to a final knock-drum (KO102) to increase the purity of the methanol to 99.41 mol%.In this process, all compressors are isentropic and operate at 72% efficiency. Moreover, a stream pressure drop between 0.1 and 2.3 bar is allowed in the heat exchangers. Table 3 presents the main specifications of the designed methanol synthesis process. Fig. 3 illustrates the simulated CO2 to methanol process using Aspen Plus. The proposed simulated process utilizes high-pressure pure CO2 obtained from a cryogenic biogas separation unit within the same plant. As justified by Rivarolo et al. [23], we are utilizing CO2 produced from biogas, which offers more excellent economic performance and a more straightforward plant layout. The use of pure and high-pressure CO2 for methanol production reduces the project's overall costs due to: (1) the employment of a single CO2 compressor, and (2) limited need for CO2 collection and purification devices. Hydrogen was assumed to be purchased from a local market to supply the process.The hydrogenation of CO2 to methanol takes place in the catalytic reactor (R101). The reactor uses a commercial Cu/ZnO/Al2O3 as a catalyst. Consequently, designing and modeling the feed preparation section and the separation section mainly depend on the reactor’s feed specifications and the composition of the outlet stream. The results revealed that feeding the reactor with CO2/H2 mixture at 210 °C and 75 bar achieved CO2 conversion of 99% and a methanol yield of 98%. The productivity of methanol and the conversion of CO2 are enhanced by recycling part of the unreacted gas mixture. This is in good agreement with the results reported by Leonzio et al., [44]. Although the recycled stream of gases contains CO, at lower feed gas temperatures, methanol synthesis from CO2/H2 is faster than CO/H2 [45]. Based on Skrzypek et al. [24] who studied methanol synthesis kinetics over Cu/ZnO/Al2O3 catalyst in a high-pressure fixed bed plug flow reactor, the authors concluded that the surface reaction between CO2 and H2 is the rate-controlling step. The authors further reported that the selectivity is higher for a feed that consists of only CO2 and H2 without any CO. This reveals that CO2 is the primary source for methanol synthesis in the process [24,46]. It was observed that increasing the feed pressure up to 75.7 bar significantly improved methanol production and achieved overall CO2 conversion ≥ 99%. This was aimed at favorable operating conditions of the forward reaction following Chatelier's principle. Kiss et al. [7] simulated the process at 50 bar, which resulted in 100% process conversion using Cu/Zn/Al/Zr catalyst, while Atsonios et al. [22] simulated the process at 65 bar, using a membrane reactor and Cu/ZnO/Al2O3 catalyst, which resulted in 30.5% CO2 conversion. Table 4 summarizes the main specifications of the inlet and outlet streams of the proposed methanol simulation process.Additionally, Table 5 compares the CO2 conversions and methanol yields and/or selectivity of the processes reported by different authors in the literature using adiabatic/isothermal fixed bed flow reactors packed with Cu/Zn or Cu/ZnO2 based catalysts. As indicated in Table 5, the current proposed process's reported CO2 conversion and methanol yield are higher than other studies and relatively close to the results reported by Kiss et al. [7] carried out in an isothermal plug flow reactor operated at 50 bar and 250 °C. Nevertheless, other authors used the fibrous Cu/Zn/Al/Zr catalyst rather than the industrially mature Cu/ZnO/Al2O3 catalyst employed in this study. The obtained results highlight the importance of additional research to demonstrate the impact of reactor type on process conversion and to determine the optimal reactor configuration and operating conditions.Different studies studied the conversion of CO2 to methanol in adiabatic and isothermal reaction system [7,21,47]. Consequently, the productivity and the conversion of CO2 in both reactor types were explored and simulated in this study. Tests were performed at 210 °C and 75.8 bar using same previous flow rate and catalyst loading. The obtained results are illustrated in Table 6 . The adiabatic reactor produced slightly less CO2 but had a higher methanol yield and selectivity. The required heat duty and residence time differed significantly between the two reactors. It was observed that the residence of the adiabatic reactor is 50% less than the isothermal reactor. Confirming that the adiabatic reactor system has more favorable operating conditions. Therefore, the adiabatic reactor was subjected to further analysis to understand the effect of operating conditions (Temperature and pressure) and the molar flow rate of H2 on methanol production and CO2 conversion. It was proved that methanol production is independent of the reactor temperature at various reactor temperatures due to the low activation energy, which was almost zero. Hence, the changing temperature in the Arrhenius equation does not influence the reaction kinetics.The influence of changing adiabatic reactor pressure and H2 molar flow rate on the methanol production is presented in Fig. 4 a and b. Results indicated that increasing the adiabatic reactor pressure is directly proportional to methanol conversion. Maximum pressure of 75 bar can be set as the operating pressure due to kinetic limitations. Moreover, increasing the H2 feed flow rate resulted in higher methanol yield, where a CO2/H2 ratio of 1:7 was optimally selected for the study. As can be seen in Fig. 4 (b), increasing the H2 flow rate beyond 535.22 kmol/hr does not significantly influence the yield, wherein expanding the molar flow rate by 34.6% would only enhance methanol yield by 4.5%.On the other hand, changing the temperature and/or the pressure of the feed disturbs the flash specifications. Hence optimal feed specifications of 210 °C and 75.8 bar were selected to achieve a CO2 conversion of 95.66% in an adiabatic reactor, which is relatively lower than the reported CO2 conversion in the literature. Further optimization of the overall process has been studied to investigate the optimal configuration.The presented process involves different endothermic and exothermic units where heat integration is crucial for improved process efficiency. The pinch analysis method proposed by Linnhoff and Hindmarsh [48] was considered for designing an optimal heat exchanger network (HEN) to (1) improve the overall process energy efficiency; (2) minimize the operational costs and utility consumption, and (3) minimize indirect CO2 emissions due to reduction of fuel consumption for steam generation. The commercial software Aspen Energy Analyzer V11 was used to conduct the pinch analysis where a minimum temperature difference ( Δ Tmin) of 5 °C was selected. The main trade-off when considering a low Δ Tmin in pinch analysis is between energy/external utility reduction and increased capital costs due to extra heat exchangers. A low Δ Tmin decreases utility costs but increases the capital costs for installing additional units.Consequently, the payback time on capital investment was also considered for evaluating the optimal heat integration scenario [49]. The optimal results from the Aspen Energy analyzer were then transferred to the primary Aspen Plus simulation for an updated process flow diagram after heat integration, illustrated in Fig. 5 . The implementation of heat integration resulted in 63.19% energy savings after introducing three additional units, RE101, RE102, and RE103. A comparison of the required external utility requirement under optimized HEN vs total utility requirement in the absence of optimized HEN is illustrated in Fig. 6 . Results after heat integration imply that external hot utilities were reduced from 3.13 to 0.014 GW achieving more than 99.55% of saving. In addition, the external cooling utilities requirements were reduced from 3.62 to 2.11 GW achieving around 41.7% saving. Consequently, deploying an optimized methanol production will significantly reduce the costs associated with utilities purchasing and/or generation.The utilization of high-pressure pure CO2 in the proposed process contributes to reducing both capital costs (Capex) and operating costs (Opex) compared to other models reported in the literature. Under optimized conditions, it was noticed that the proposed process required pure H2 at a pressure of 30 bar and a total specific power of 214 kWh/tMeOH to produce methanol with a purity of 99.41 mol%. This is economically feasible if compared with an operating pressure of 65 bar and specific power consumption of 113 kWh/tMeOH to produce methanol with a purity of 99.3 mol% methanol as reported by Atsonios et al. [50]. The reported value did not consider the power requirement for CO2 feed preparation as it was considered as part of the CO2 capture and treatment unit in the plant.Optimizing the CO2 conversion reduced total utility requirements by 63%, indicating that this process has the potential to generate additional revenue. The plant's economic viability depends on different factors, including Capex, utility costs, electricity costs, the project's lifetime, CO2 taxes in some countries, and methanol price in international markets. The net present value (NPV), equation (12), was used to assist the economy of the methanol production process and determine its profitability: (12) NPV = ∑ t = 1 n C F t 1 + i t where CF represents annual cash flow at any time (t); n is the service life of the project and i is the rate of return on the investment. The profitability of the methanol production process was based on a plant service life of 20 years and a rate of return on investment of 8%. The Capex of the process was determined using the step counting method following the procedure established by Timm’s correlation for similar gas processes [51]: (13) Capex = 13000 N Q 0.615 where Capex is in US Dollars for 1998, N is the number of significant processing units, and Q is the annual plant capacity in metric tons (mt). When counting the significant processing units, only reactors, distillation columns, and compressors are considered to have substantial costs [52]. Moreover, since Timm’s correlation results in Capex were conducted in 1998, cost indices were used to adjust the Capex value to the year 2021 [53].Both fixed and variable Opex must be addressed when estimating the Opex. In this analysis, fixed operating and maintenance costs were taken as 1.04% of Capex assumed previously by Bellotti et al., [8]. On the other hand, variable Opex relays on the production capacity, utility requirement, and fuel and electricity costs for running equipment and generating utilities. All compressors are electrically driven and purchased from an external local supplier in Qatar at a $0.036/kWh [54]. The steam generation total cost ($/lbsteam) was calculated using Eq. (14) [52]: (14) S t e a m c o s t = F u e l P r i c e × H e a t i n g r a t e B o i l e r e f f i c i e n c y where fuel price was taken as a fixed average monthly Henry Hub natural gas price in 2021 of $3.62/MMBTU; the heating rate is the amount of energy needed to heat feed water to saturated low or high-pressure steam in (Btu/ lbsteam) [55], and boilers efficiency of a fixed value of 85.7% was considered [52,56].For cold utilities, sea water and chilled water were considered for cooling process streams down to 35 and 15 °C, respectively. However, the costs of raw water, makeup water, condensate return, water treatment, and power for pumping cooling water were not analyzed. Further detailed calculations can be considered when assessing the plant on a tactical level of project planning. For example, seawater can be used to cool down process streams down to 35 °C, and chilled water can be used to cool down the process stream (S116) entering FT101 to 15 °C. The revenues of the proposed CO2 conversion plant are based on selling the produced liquid methanol at an average price of $692/mt [57]. The process profitability was supported by the availability of CO2 from nearby cryogenic biogas or petrochemical processes and the presence of gray H2. This latter is produced from steam methane reforming in the Middle East and supplied for $0.9/kg [58]. The NPV and payback period of the methanol plant with a production capacity of 23.4 kt/yr was determined to be $6.5 million and nine years, respectively based on 20 years of service life.It is difficult to identify the most competitive methanol production scheme. The economic performance depends on the electricity and/or fuel prices, hydrogen costs, and methanol selling price in international markets. Under fixed Opex and hydrogen supply costs, the profitability of the investment in CO2 hydrogenation to methanol process was investigated based on different methanol production capacities: 23.4 kt/yr, 33.6 kt/yr, and 44.8kt/yr for the project’s lifetime of 20 and 25 years. As shown in Fig. 7 , increasing the production capacity up to 3.36 kt/yr and 4.48 kt/yr for a project’s lifetime of 20 years results in enhancing the NPV of the project by 147% and 411%, respectively. The NPV is further enhanced for all production scenarios when extending the project’s lifetime by five additional years. This reflects the economic attractiveness of deploying a bio-methanol process due to the low Opex and the availability of affordable gray H2 supplied from local steam gas reforming processes. It is worth observing that due to the high requirement of H2 to satisfy the CO2/H2 ratio of 1:7 in the process, a maximum H2 supply price of $0.97/kg is required to break even the NPV for a 20-year project with annual methanol production of 2.34 kt/yr. Consequently, supplying renewable hydrogen from PEM electrolysis at a price between $4.2/kg and $5.2/kg will be economically infeasible for the proposed CO2 hydrogenation to methanol process [58].As shown in the proposed bio-methanol production process, CO2 emissions can be released into the atmosphere directly from the main process equipment or indirectly due to burning fuel for generating thermal energy and/or electricity (in case generated locally). The process releases three streams to the atmosphere containing CO2 for direct emissions: PUR-S111, S-117, and S-125. After heat integration, only burning fuel for generating low-pressure steam contributes to indirect CO2 emissions to the atmosphere. Consequently, the indirect emissions are mainly influenced by the fuel needed to generate steam utilized in the process. According to the US Energy Information Administration [59], burning 1 MMBtu of natural gas emits around 117 lb of CO2. A comparison between direct and indirect CO2 emissions before and after heat integration is presented in Fig. 8 . The figure shows that the indirect emissions were reduced by around 98% after conducting heat integration since the process streams provided sufficient duty for the heaters.Additionally, when considering the optimized process, the estimated direct and indirect emissions are 98.6% and 96%, respectively, less than the values reported by [37], who reported direct emissions of 0.090 tCO2/tMeOH and indirect emissions of 0.136 tCO2/tMeOH for a European bio-methanol plant. The reduction in emissions is mainly attributed to the requirement of a smaller number of compressors and heat exchangers in this process for preparing CO2 feed to feed specifications. This reflects the added value of incorporating a CO2 to methanol process within the biomass supply chain for sustainable bio-methanol production.In the presented process, high-pressure liquid and pure CO2 were considered for methanol utilization. CO2 captured from flue gases is a potential feed that will not impact the thermodynamic properties of the primary catalytic conversion process. However, additional CO2 capture, treatment, and compression units will be needed to meet feed specifications. Hence, overall plant design and economic feasibility should be investigated. On the other hand, it is worth mentioning that different production routes can be utilized for CO2 to methanol production, including CO2 electrochemical reduction to methanol and two-step CO2 catalytic conversion to methanol. In the latter route, CO2 is first converted to CO, and CO is then hydrogenated to methanol in the second step. Both processes are still industrially immature and require further development. Moreover, despite the maturity of the CO2 hydrogenation to methanol technology, catalyst development is still a dynamic area of research where other studies investigated the utilization of Ni/Ga [33,60], ZnO/ZrO2 [61], and InOx/ZrO2 [62] catalysts. The studied catalysts are still under development and have not reached the stage of industrial commercialization yet.A state of art of catalytic conversion of CO2 to methanol is presented in this study. The economic and environmental feasibility of the proposed process under optimized operating conditions was explored. In comparison with previous studies, the assessed process involves less equipment due to the utilization of high-pressure and pure liquid CO2, produced or generated from a former cryogenic biogas separation process or petrochemical industries. Optimized CO2/H2 feed ratio of 1:7 to achieve an overall CO2 process conversion of 99% and methanol yield ≥ 99%. Simulation results indicated better performance for the adiabatic reactor than the isothermal reactor, with a reduced residence time of 48.46% and operating conditions of 210 °C. Overall energy efficiency was further improved by lowering external utilities by 63% after using the heat integration approach. Similar to the economic evaluation, which concluded that the process profitability is highly dependent on H2 supply price, in this analysis, the financial assessment demonstrated the requirement of a maximum H2 supply price of $0.97/kg to break even the NPV for a 20-year project lifetime in the Middle East with annual methanol production of 2.34 kt/yr. From an environmental perspective, the optimized process successfully contributes to reducing total CO2 emissions by 97.8% compared to the baseline process configuration. The catalyst Cu/ZnO/Al2O3 showed excellent efficiency for the industrial commercialization of CO2 hydrogenation to methanol. Future research on process configuration and simulation could involve testing the efficiency and stability of different novel Cu, Pd, or Zn-based catalysts for CO2 hydrogenation to methanol under varied operating conditions and CO2/H2 feed ratios. Noor Yusuf: Methodology, Software, Data curation, Writing – review & editing, Writing – original draft. Fares Almomani: Resources, Supervision, Project administration.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.We would like to acknowledge the fund provided by QAFCO R&D grant # QUEX-CENG-QAFCO-20/21-1. The statements made herein are solely the responsibility of the authors. Open Access funding provided by the Qatar National Library.

The hydrogenation of CO2 to methanol is one of the promising CO2 utilization routes in the industry that can contribute to emissions mitigation. In this work, improved operating conditions were reported for the sustainable catalytic hydrogenation of CO2 to methanol using Cu/ZnO/Al2O3 catalyst operated at 70 bar and 210 °C. The CO2 feedstock used for this process is pure CO2 produced from the cryogenic upgrading process of biogas or hydrocarbon industries and ready-to-use hydrogen purchased at 30 bar and 25 °C. The process was modeled and simulated using the commercial Aspen Plus software to produce methanol with a purity greater than 99% at 1 bar and 25 °C. The simulation results revealed that an adiabatic reactor operated with a CO2/H2 ratio of 1:7 produces methanol with a yield ≥99.84% and a CO2 conversion of 95.66%. Optimizing the heat exchanger network (HEN) achieved energy savings of 63% and reduced total direct and indirect CO2 emissions by 97.8%. The proposed methanol process with an annual production rate of 2.34 kt/yr is economically sound with a payback period of nine years if the maximum H2 price remains below $0.97/kg. Hence, producing or purchasing gray H2 from a steam reforming plant is the most viable economic source for the process.

Since the 1970s societal focus on air quality and pollution control has introduced legislation requiring a continuous lowering of the sulfur content in transportation fuels implying that almost all crude oil (ca. 2500 million tons annually) is to be hydrotreated. In most of the world the maximum sulfur content in gasoline and diesel for road use is now 10 ppm, corresponding to a removal of 99.99% of the sulfur originally present. Hydrodesulfurization is the industrial catalytic process by which sulfur is removed from organosulfur molecules in mineral oil as H2S at high hydrogen pressure and temperature (250–400 °C). The hydrodesulfurization process is sometimes more generally denoted hydrotreating as also organonitrogen compounds, aromatics and unsaturated carbon molecules in the oil undergo simultaneous conversion and/or are hydrogenated [1–3]. Over the last 40 years increasing severity of legislation has called for intense development of industrial hydrodesulfurization processes and catalysts. The main catalytic challenge for production of today’s ultra-low sulfur diesel (<10 ppm sulfur) is the conversion of refractory alkyl-substituted dibenzothiophenes. These are molecules such as 4,6-dimethyldibenzothiophene (4,6-DMDBT) [4] that have alkyl substituents located in positions next to the sulfur atom and where the alkyl groups thus limit access to the sulfur atom.Dr. Henrik Topsøe (1944–2019) [5] was a key figure in the research that established a detailed fundamental understanding of the nature of the industrially employed hydrodesulfurization catalyst, its active site(s), the molecular reaction routes and inhibition phenomena under ultra-deep hydrodesulfurization. Many of his groundbreaking and seminal contributions will be mentioned below including the discovery of the so-called Co-Mo-S structure that is now a commonly accepted model for the working hydrotreating catalyst. Characteristic to his work was a continuous employment of new and more sophisticated experimental methods and the combined use of results from multiple experimental and theoretical approaches, all with the aim to steadily uncover the secrets of hydrotreating catalysis. At Haldor Topsoe A/S, Henrik pursued the fundamental research in close interplay with industrial developments and fostered a synergy for implementation of the findings in industrial practice as visualized in Fig. 1 .In this paper, we will honor Henrik and his enormous impact on research and development in hydrotreating catalysis and acknowledge his unique generosity, mildness, humor, commitment and intellect, which made him a fantastic colleague and inspirational supervisor/mentor. The paper will present new findings for advancing hydrodesulfurization and thus lie at the heart of Henrik’s interests. Although Henrik was not part of the present work, it testifies that the tradition of research and methodologies that he established in the laboratories of Haldor Topsoe A/S are deeply embedded and form the basis for continuous developments of even rather mature catalytic systems.Sweetening of gasoline was in the first part of the 1900s undertaken by non-catalytic processes involving absorption or conversion of the odor-repulsive mercaptans by means of agents such as metallic copper or lead sulfide. Subsequently conversion of organosulfur compounds was claimed using molybdenum-sulfur or tungsten-sulfur materials as catalysts, but actual catalysis was not documented before 1943 when it was discovered that the bimetallic combination cobalt-molybdenum had an extraordinary ability for catalytic hydrodesulfurization [6]. In the context of hydrotreating, cobalt is traditionally termed a promoter (to molybdenum) but as the boost in activity is a factor of 10–20 this terminology is somewhat misleading. The 1943 paper also examined the combinations iron-molybdenum, copper-molybdenum, zinc-molybdenum and aluminum-molybdenum to find that they had much inferior catalytic performance. Soon it was found that the combination nickel-molybdenum also had high catalytic activity and this combination is together with cobalt-molybdenum universally employed in industrial hydrotreating today. The chemical state of the catalytically active metals at process conditions was known to be at least partially sulfidic and in some early models sulfidic Mo was bonded via oxygen to the surface of the high-surface area alumina support employed. It was eventually realized that Mo was in a chemical state like MoS2, i.e., with no or very limited amounts of bonds to alumina surface. The precise role of the promoter element cobalt (also itself in a sulfidic state) was yet unresolved. Two models were considered [1]. The first, a contact synergy model in which separate Co9S8 and MoS2 nanostructures together were responsible for the hydrotreating catalysis with a spill-over of activated hydrogen. The second, an intercalation model in which the promoting Co atoms were located in the van der Waals gap between the layers of the layered MoS2 structure, similarly to how layered MoS2 structures may intercalate e.g., lithium ions. In 1976 a team led by Henrik Topsøe found by in situ Mössbauer Emission Spectroscopy (MES) that cobalt in sulfidic cobalt-molybdenum catalysts occurred in 3 different forms [7]: (i) Co9S8, (ii) CoAl2O4 and (iii) a new form distinctly different from the two-former mentioned. Both Co9S8 and CoAl2O4 may sometimes be present in catalysts but neither of the compounds contribute in any significant extent to the catalytic activity [8–10]. On the other hand, a clear correlation between the size of the MES cobalt signal and the catalytic activity within a series of cobalt-molybdenum catalysts made it possible to assign the catalytically active center to what was named the Co-Mo-S structure [8,9]. It was in 1981 found by Extended X-ray Absorption Fine Structure (EXAFS) [11] that Mo was present as MoS2-like nanodomains and the so-called Co-Mo-S model was proposed which has Co atoms located on the edges of the layered MoS2 structure. A high affinity of Co for the edges of MoS2 has since been confirmed by Scanning Tunneling Microscopy (STM) [12] of Co-Mo-S structures synthesized on Au(111) and by Scanning Transmission Electron Microscopy (STEM) [13] of carbon supported Co-Mo-S synthesized by metals impregnation and subsequent gas phase sulfidation. In parallel with the technical progress in atomic resolution imaging, Density Functional Theory (DFT) methods evolved and have now been able to elucidate the catalytic mechanisms taking place at active sites located on the edges of Co-Mo-S nanocrystals [14–18].Generally, two hydrodesulfurization pathways are in play for HDS of dibenzothiophenes, the direct desulfurization (DDS) and the pre-hydrogenation (HYD) pathways [19]. By the DDS pathway, direct cleavage of the C—S bond(s) in the dibenzothiophene molecule takes place and with a minimum amount of H2 consumed. By the HYD pathway, the dibenzothiophene molecule undergoes an initial hydrogenation in one of its six-rings followed by cleavage of C—S bond(s). In this way the HYD pathway consumes more H2 than the DDS pathway. Whereas molecules such as thiophene and dibenzothiophene are primarily hydrodesulfurized by a DDS pathway then HDS of sterically hindered dibenzothiophenes, such as 4,6-DMDBT, proceeds mainly via a HYD pathway in the model feed studies required to detect and quantify the individual desulfurized product molecules [20–22]. The detailed reaction mechanisms are still debated. Following the DDS pathway, the current view of the catalytic cycle is that sulfur vacancies are created on cobalt atoms (associated with 1 ¯ 00 S-terminated edges of MoS2) by formation of H2S by means of atomic hydrogen formed by activation of H2 at the metallic so-called brim sites at the edge of the MoS2 structure [23–25]. The sulfur atom of an organosulfur molecule attaches to a cobalt atom at a coordinative undersaturated site (CUS) such as a sulfur vacancy and after transferal of more activated hydrogen atoms generated at the nearby brim sites the carbon part of the organosulfur molecule dissociates into a main hydrocarbon part and a single sulfur atom. Molecular hydrodesulfurization has now taken place and subsequently sulfur atom removal as H2S by means of yet more activated hydrogen from the brim sites takes place to close the cycle. STM imaging have revealed organosulfur molecules associated with edge sites for sufficiently long time to capture an STM image, thus providing unprecedented insights and direct visualization of the important first adsorption step in the catalytic process [26–28].Experimental microscopy techniques and DFT calculations give, for Co and Ni promoted MoS2, a mutually consistent image of the nature of the active sites on the MoS2 edges. No other first transition period metals have experimentally to the same substantial degree been found to promote the hydrodesulfurization reaction. For instance, an STM-based study [29] has shown that all four metals Co, Ni, Cu and Zn associate to MoS2 edges in identical ways, i.e., association to the edge is a necessary but not sufficient prerequisite for promoting the catalysis. As Cu and Zn do not have any promoting effect [29], the electronic structure of the potentially promoting element clearly plays a role and only for Co and Ni do the electronic structure seem suited for hydrodesulfurization catalysis. DFT calculations have shown that when Fe is placed on an MoS2 edge in the very same way as Co, then the Fe-Mo-S system is characterized by sulfur binding parameters that suggests low catalytic activity [30]. Going to the second and third transitions periods, Rh-Mo and Ir-Mo catalysts have experimentally [31–37] shown substantially higher activities than a Mo-only catalyst, i.e., the two elements Rh and Ir promote to a comparable extend to their first transition period counterpart Co. Oppositely, the element combinations Ru-Mo, Pd-Mo and Pt-Mo did not provide catalysts with a substantially enhanced HDS activity compared to a Mo-only reference [37–40]. All in all, only Co, Ni, Rh and Ir have been found to substantially promote MoS2.In industrial practice, a particular hydrotreating reactor is designed with a certain size and maximum pressure. To get the best performance, the refiner will choose a catalyst by considering both the specific oil feed (nature of crude, boiling point range) and the achievable hydrogen amount and pressure. In this context Co-Mo is typically the preferred catalyst for relatively low hydrogen partial pressures (<40 bar) whereas Ni-Mo is preferred for high pressure (>60 bar) applications. This has, over the years, led to many catalyst producing companies launching all-in-one catalysts that contain both Co and Ni as promoting elements. Although such element combinations may indeed at certain operating conditions give moderate advantages in terms of catalyst performance, it is certainly not the case that the combination provides the refiner with the very best of both Co and Ni. Rather, these Co-Ni-Mo catalysts have performance characteristics that are probably closer to a kind of average of the Co and Ni catalyst performances, i.e., the performance you would get from loading a physical mixture of Co-Mo and Ni-Mo catalysts. Bimetallic promotion of MoS2 with Ni-Cu has also been examined but with no great industrial breakthrough [41].The complexity of the hydrodesulfurization reaction mechanisms as well as the structural complexity of the bimetallic transition metal sulfide nanostructures have hampered a detailed understanding of all empirical findings of promotional effects. However, with the toolbox at hand today and with the multidisciplinary approach that Henrik was so dedicated to, we will in this paper explore a metal combination Co-Pt that has had very limited attention [42]. Specifically, we will demonstrate how, in our hands, highly activity-enhanced Pt-Co-Mo-S nanocrystals can be prepared by slight modifications of the otherwise standard Co-Mo-S catalyst and we will present an atomistic picture for the synergetic role of this tertiary transition metal sulfide catalyst system.Pt-containing Co-Mo catalysts were prepared in two different ways: (i) by post impregnation of Pt onto a commercial (unsulfided) Co-Mo/Al2O3 catalyst (16 wt% Mo and 3.5 wt% Co) and (ii) to illustrate the feasibility of a more industrially acceptable preparation route, by incipient wetness impregnation of alumina carrier extrudates with a Co-Mo liquor into which Pt had been incorporated. For the former preparation method [Pt(acac)2] (acac = acetylacetonate), corresponding to the desired wt% of Pt on the final catalyst, was dissolved in a volume of dichloromethane (DCM) corresponding to the pore volume, as determined by mercury porosimetry, of the oxidic Co-Mo catalyst to be modified. After 30 min impregnation time, the sample was air dried at room temperature for 30 min followed by drying at 250 °C in air for 1 h. For the latter preparation method, [Pt(NH3)4](HCO3)2 was dissolved into a commercial Co-Mo liquor, followed by pore volume filling impregnation of an alumina carrier. This also yielded catalysts with 16 wt% Mo and 3.5 wt% Co. After preparation the catalysts obtained by one or the other method differed neither in chemical properties nor activity within the uncertainty of the activity measurements. Samples with a specific nominal loading of Pt are in the following referred to as Pt-Co-Mo(wt%).The catalyst samples were tested in a 5-in-1 pilot unit in which 5 reactors are placed closely together thus experiencing the same thermal conditions. The reactors were loaded with a mixture of 15.0 mL neat catalyst volume (whole extrudates) and 40% fine inert SiC particles. In one 5-in-1 test 4 platinum-modified experimental catalysts were tested together with the commercial reference sample that was used as basis for the platinum impregnations. After loading of the reactors, the catalysts were sulfided for 24 h at test condition using a sulfur doped oil. After the sulfidation procedure a switch to the test feed was made and the pilot unit was then run for 110 h at steady conditions. Samples were collected simultaneously from all 5 reactors at run hours 80, 90, 100 and 110. The activities reported are based on the average sulfur level of the 4 samples collected. Sulfur-content variation between the 4 samples obtained at different run hours was minimal. The activities have been calculated using our internal kinetic model to obtain relative rate constants with the industrial reference catalyst used for the impregnation experiments defining activity of 100%. Test conditions: 355 °C, 30 barg H2, 1.5 LHSV, 490 H2/oil, 75/25 w/w blend of straight run diesel (LG) and cracked feed (LC). Feed properties: 1.22 wt% S, 356 wtppm N, SG 60/60 0.8735, calc. D86: 237 °C (10 vol%), 297 °C (50 vol%), 358 °C (90 vol%). The tested Pt-Co-Mo catalysts were collected from the reactor after cooling to RT. Then the samples were sieved and rinsed several times with an excess of pentane to wash away oil residues inside and outside the porous extrudates. All volatiles were removed by drying the sample at 40 °C in vacuum and the samples were subsequently stored in closed containers until characterization by electron microscopy or X-ray absorption was undertaken.Electron microscopy examinations were performed using an FEI Talos F200X (scanning) transmission electron microscope equipped with ultra-bright field emission gun (X-FEG) and Super-X EDX detectors. The microscope was operated at 200 keV in both scanning-beam and broad-beam modes. In the scanning-mode, a high-angle annular dark field (HAADF) detector was employed for imaging concurrently with EDX spectrum acquisition. In the broad-beam mode, a charge-coupled device camera was used for imaging. The combined HAADF and EDX data cube was acquired for 30 min over an area of 1024 pixels × 1024 pixels (0.26 nm pxl−1) with a probe current of 0.7 nA and EDX energy range of 0–20 keV (0.01 keV/channel). Samples for electron microscopy were prepared by crushing the catalyst pellets in a mortar and dispersing the fine powder onto a Cu-TEM grid with continuous carbon (SPI Supplies) in ambient conditions. Thus, the samples have been exposed to ambient conditions for a few days during transportation and storage before examination in the electron microscope.To increase the counting statistics of a single EDX pixel spectrum the data were binned by a factor of x8 to yield an effective pixel size of 2.1 nm and further processed in Brüker software (Esprit 1.9) by a Bremsstrahlung background subtraction, series deconvolution and Cliff-Lorimer quantification to display the net counts. The Pt-Lα (9.435 keV) signal, however, showed too few counts for a proper background subtraction and therefore are displayed as raw counts with an estimated signal-to-noise of 50% as determined from the peak-to-background in the summed EDX spectrum. The Pt-L region was emphasized because it separates from other elemental peaks, in contrast with e.g. the Pt-Mα signal (2.050 keV) that overlaps with Zr Lα (2.044 keV) signals as reminiscence from the microscope. Likewise, the present analysis focuses on Mo-Kα (17.480 keV), Co-Kα (6.931 keV) and S-Kβ (2.465 keV) that all separates peaks stemming from the sample and microscope.High-resolution scanning transmission electron microscopy (HRSTEM) was performed on a JEOL ARM-200F equipped with cold field emission gun (CFEG) and CEOS probe (STEM)-Cs-corrector. The microscope was operated at 200 keV and probe aberrations up to 3rd order was corrected. The illumination system was set with a probe size of ~ 1 Å, with a current of approximately 0.1 nA, and with a pixel dwell-time of 32 µs. The focusing of the sample was done prior to acquisition in an adjacent area as to record an image of a pristine area not previously exposed to the electron beam. Images were generated using a high-annular dark-field detector. Although residual oil is removed by pentane as described in section 2.2, HRSTEM revealed carbon deposition to a degree that varied from area to area. Occasionally areas allowed sufficient cleanliness to acquire an HRSTEM image of a clearly atom-resolved image and the atomic details blurred after a few scans, so it was necessary to move to a new area. It was such images that have been considered in the present study. Even though further optimization of beam energy and current could be pursued, Fig. S8 (Supplementary Material) indicates that some Pt atoms are stabilized in successive images suggesting they reveal their pristine locations. Complementary STEM image simulations were carried out in the QSTEM software suite with experimental details shown in Supplementary Material.The activity tested Pt-Co-Mo(0.5) catalyst was, in its sulfided state as retrieved from the reactor, characterized ex situ by X-ray absorption near edge structure (XANES) at the XAS beamline at the ANKA synchrotron source (Karlsruhe, Germany) using the Pt L3-edge (11.564 keV). The catalyst was crushed and pressed into a 13 mm pellet using polyethylene and measured in transmission mode. The XANES spectra were energy calibrated using a metal reference, background subtracted and normalized. References of PtS (ICSD_31131) and PtS2 (ICSD_41375) were also measured.To obtain total energies we employed the GPAW [43,44] density functional theory code in the finite difference mode with a spacing of 0.18 Å. The exchange and correlation were treated using the BEEF-vdW functional [45]. As in ref. [25], the S-edge and Mo-edge of the Co-Mo-S particle were modelled with 4x4 stripes periodic in the x-direction and separated with vacuum in the y and z directions. The corner of the Co-Mo-S particle was modeled with a step stripped continuous in the x-direction separated by vacuum in the y and z directions, exposing 3xCo and 2xMo on the S- and Mo-edge respectively. A 2x1x1 Brillouin zone sampling was used in the x, y and z direction respectively. All structures were optimized until the maximum force was lower than 0.03 eV/Å. The crystal structures of PtS2, MoS2 and Co8S9 were optimized using the stress tensor method available in ASE [46]. In the following, the free energy of gas phase H2 and H2S has been obtained using the ideal gas approximation [47] G x = E x + ZPE x + Δ H x 0 , T - T S x T + k B T ln p x p where E x is the electronic energy, ZPE x the zero-point energy, Δ H x 0 , T the change in enthalpy from 0 K to T, S x T the entropy at T, p x the pressure of the molecule in the gas, and p the standard pressure, and the free energy of surfaces and bulk sulfides are assumed to be described by the 0 K electronic energy. The uncertainty of the calculation has been estimated using the BEEF-vdW ensemble [45], using an ensemble of 3000 energies. This uncertainty represents how much, e.g., an adsorption energy (at 0 K) can vary within the GGA functionals.We obtain the sulfur equilibrium termination at HDS conditions, 673 K and p(H2)/p(H2S) = 20, for the S-edge, Mo-edge and at the corner by gradually increasing the sulfur coverage and for each step calculating free energy change of the sulfidation according to: (1) CoxMoySz-1 + H2S(g) ↔ CoxMoySz + H2(g) where x, y and z represent the number of Co, Mo and S atoms in the supercell, respectively.The adsorption energy of a molecule on a specific site, Δ E x , has been obtained as follows: Δ E x = E x ∗ - E x - E ∗ where E x ∗ is the energy of the adsorbed molecule on a specific site, E x the energy of the molecule in vacuum, and E ∗ the energy of the site.The stability of single atom Pt incorporation at the Mo-edge is described by the free energy of the following reaction: (2) CoxMoySz + PtS2(b) + nH2(g) ↔ PtCoxMoy-1Sz-n + MoS2(b) + nH2S(g) where CoxMoySz here represents the equilibrium structure of the Mo-edge, PtCoxMoy-1Sz is the equilibrium structure of the Pt doped Mo-edge, where 1 Mo at the edge has been substituted with 1Pt in the unit cell, and PtS2 and MoS2 are the reference metal sulfides of Pt and Mo, respectively. Similarly, at the Co-promoted S-edge and corner site, the stability of Pt is described by calculating the free energy change of the reaction: (3) CoxMoySz + PtS2(b) + (10/9+n) H2(g) ↔ PtCox-1MoySz-n + 1/9 Co9S8(b) + (10/9+n) H2S(g) where CoxMoySz here represents the equilibrium structure of the S-edge/corner of a Co-Mo-S particle, PtCox-1MoySz the equilibrium structure of the Pt doped S-edge/corner site, and PtS2 and Co9S8 the stable metal sulfides of Pt and Co respectively.The hydrodesulfurization (HDS) activity of the Pt-Co-Mo catalysts was measured relative to a commercial Co-Mo reference catalyst under industrially relevant conditions in a pilot unit. The activities are reported as relative volume activities. All (Pt)-Co-Mo catalysts measured have, within the experimental uncertainty of preparation using the same alumina carrier and the experimental uncertainty of subsequent chemical analysis, identical molar loads of molybdenum per volume (and weight) of catalyst. Fig. 2 a shows the influence of increased amounts of platinum in the catalysts. The activity increases approximately linearly up to about 1 wt% Pt (10000 ppm) and here reaches an unparalleled catalytic performance of 146%. Platinum in amounts higher than 1 wt% did not increase activity further and eventually a slight decrease in activity was found. This may suggest a saturation and a corresponding Pt waste of the promotional active sites at high platinum loadings. The activity of catalysts without any cobalt at all, Mo/Al2O3 and Pt-Mo/Al2O3, was found to be at least an order of magnitude lower than that of the Co-Mo reference catalyst of Fig. 2a. Thus, it must be the intimate contact of Co and Pt that is responsible for the boosted performance. Fig. 2a shows a Pt promoted activity for a fixed cobalt amount. Samples with variations in Co amounts were prepared as well. Fig. 2b shows the effect of variations in the amount of both promoters (Pt and Co) while keeping the molar Mo amount constant. The contents of Pt and Co are shown as estimated volumetric concentrations (mmol/L) inside the reactor during catalytic testing. In Fig. 2b, the activity increases with an increased amount of Co (following the x-axis, y = 0). An increase in activity of 10% was obtained by an increase in the Co concentration from 479 mmol/L to 603 mmol/L in the reactor, corresponding to 124 mmol/L or 26% of the total Co in the reference sample. The Pt effect is much more pronounced: a similar activity increase of 10% is obtained by replacing only 10 mmol/L of Co (~2% of total Co) with Pt (following the y-axis). Thereby, in the concentration domains examined, the intrinsic promotional effect of platinum (per atom) is more than 12 times higher than cobalt in a rather large window of promotor concentration variations of the industrial Co-Mo catalyst. Because of the very consistent data, navigating in the activity contour plot by simple extrapolation now allows for tailoring the catalyst activity with the usage of cobalt and platinum in combination. For instance, at a fixed low Pt amount of 10 mmol/L, an activity lift of 20% RVA can readily be obtained, and likewise for 30 mmol/L of Pt an activity boost of more than 35% is expected as indicated by dashed lines in Fig. 2b.Selected Pt-Co-Mo catalysts were after catalytic HDS activity tests characterized by electron microscopy and X-ray absorption spectroscopy in order to unravel the nature of the Pt-promotion. Fig. 3 a shows a STEM image of the activity tested Pt-Co-Mo(0.5) catalyst with the corresponding EDX element maps displaying the distribution of cobalt, molybdenum and platinum. The metals are all uniformly distributed on the agglomerate together with EDX signals of sulfur, aluminum and oxygen in accordance with an alumina-supported (sulfided) Co-Mo catalyst (Fig. S3, Supplementary Material). The elemental composition of various sample agglomerates observed in the STEM imaging appeared surprisingly homogeneous (based on at least 20 agglomerates for the Pt-Co-Mo(0.5) sample) and despite a very little platinum detection signal in a single pixel spectrum (see Fig. S4, Supplementary Material) a Pt peak signal was verified throughout the catalyst from larger EDX sum-spectra areas. In fact, a quantification of the Pt amount from the full area EDX sum spectrum, as well as a selected smaller area, revealed a similar Pt content of 0.5 wt% demonstrating a high homogeneity within the sampled area, and indicates a uniformity on a larger scale as it matches the nominal platinum load in the prepared catalyst (Fig. S5, Supplementary Material). Such a highly dispersed platinum phase matching the nominal Pt weighting was also found in the highly active catalyst Pt-Co-Mo(1.0) of twice the platinum amount. However, in the catalyst with the highest Pt load Pt-Co-Mo(1.9) significant amounts of distinct Pt nanoparticles a few nm in size were observed in addition to the highly dispersed Pt phase (Figs. S5 and S6, Supplementary Material). No other elements from the element maps (e.g. sulfur) could be clearly associated with the Pt nanoparticles indicating likely metallic platinum (Fig. S6, Supplementary Material). Interestingly, a full-frame EDX quantification including both the dispersed Pt phase and Pt nanoparticles still matched the overall nominal Pt load (1.9 wt%), indicating a two-phase Pt distribution with an estimated fraction of the highly dispersed Pt phase corresponding to 1.2 wt% as measured in a selected area free of Pt nanoparticles (Fig. S5, Supplementary Material). The concentration of platinum in the highly dispersed Pt phases, as determined from the STEM-EDX analyses, thus scales with the activity gain of the catalysts at all Pt loadings (see Fig. 2a). On the other hand, the presence of Pt nanoparticles (observed occasionally in all Pt-samples samples but only to a substantial extent in the highest Pt loading samples) appear with limited or no correlation with the HDS activity. Therefore, we exclusively associate the highly dispersed platinum phase with the Pt promotional effect of the Co-Mo catalysts and attribute the activity saturation above 10000 ppm Pt in Fig. 2a to the onset of the separate Pt formation.To shed further light on the local structure of the dispersed platinum, the activity tested Pt-Co-Mo(0.5) catalyst was qualitatively addressed by ex-situ X-ray absorption near edge spectroscopy (XANES). The normalized XANES spectrum recorded at the Pt L3-edge (Fig. 3b) reveals edge features of the Pt-Co-Mo catalyst that in intensity and shape are close to those of PtS2, indicating a similar electronic configuration to platinum(IV) sulfide. Thereby the platinum in the Pt-Co-Mo catalyst is, in average, significantly different from both a metallic (Pt) state or an oxide (PtO2) state (Supplementary Material). This also indicates limited oxidation during the sample storage prior to the measurements.High-resolution TEM imaging was used to visualize the nano-scale structures of the tested catalysts. TEM images of the reference Co-Mo sample without Pt (rel. vol. activity = 100%) and the Pt-Co-Mo(0.5) sample (rel. vol. activity = 124%) are shown in Fig. 3c,d and reveal very similar appearance of elongated dark contrasted features with occasional two such features in pair with a separation of 0.62 nm, corresponding to the lattice distance of MoS2 (002). Thus, the elongated dark features are attributed to MoS2 slab structures viewed along the (001) basal plane, and as a cobalt promoted HDS catalyst the Co-Mo-S structure is assumed consistent with ref. [1,13]. The Co-Mo-S structures viewed along the (001) contour the terminating edges of the support and was not observed as free-standing, unsupported slabs. This indicates that the Co-Mo-S slabs are supported on their basal plane (001) by the alumina substrate. The length and stacking height of the Co-Mo-S slab structures were measured from the TEM images (16 different sample areas each) with very similar size distributions of predominantly single layer structures (~89%), double layers (~10%) or triple layers (<1%). The stacking degree calculated as the average number of layers in the MoS2 (001) direction was 1.12 and 1.13 for the Pt-Co-Mo and Co-Mo catalyst, respectively, revealing only a marginal difference of < 1%. The correspondingslab lengthswerefoundin the rangefrom1.2 nm to7.5 nmwith ameansizeof 2.38 nm and 2.48 nm for the Pt-Co-Mocatalystand Co-Mocatalyst, respectively, asdetermined from alognormalfit to the entire size distribution(Fig. 3e). Thus, the Co-Mo-S and Pt-Co-Mo-S structures are similarly distributed in stacking height and slab length, and this similarity is expected to extend below the detection limit for the slab length of 1.2 nm. Based on the minutechangeinthe average slab length theCo-Mo-S structuresin the Pt-Co-Mo catalystexposeabout4%moreedgesites compared to the Co-Mosample. This minor change in edge dispersion is insufficient to fully account for the activity boost of 24% caused by addition of Pt, indicating that Pt has additional functional effects.Recent work on enhancing the intrinsic catalytic properties of two-dimensional MoS2 for tuning the electronic properties in improved hydrogen evolution reaction (HER) catalysts points to single atom Pt dopants of the (inert) MoS2 basal plane [48]. The location of such Pt single atom has been reported to replace an Mo atom in the MoS2 structure (in which Pt—S bonds are formed), or localized in S vacancies, and various unspecific positions in case of surface (carbon) contamination [49–51]. Therefore, we address possible interactions of single Pt atoms with the Co-Mo-S structure either by adsorption or substitution under HDS sulfiding conditions using density functional theory (DFT).We have with DFT investigated the possible stable sites for Pt substitution into and adsorption on the Co-Mo-S particle. In order to do so we have chosen the 3 model structures from ref. [25] to represent a Co-promoted MoS2 particle, namely the Mo-edge, a Co-promoted S-edge and a corner site. We start out by addressing the equilibrium structures of the 3 model sites by calculating the free energy of sulfidation, H 2 S g + * ↔ H 2 g + S * , at HDS conditions (see Supplementary Material). We find at equilibrium that the Mo- and S-edge is terminated by monomeric S, whereas the corner site has a S vacancy (Fig. 4 a), in accordance with [25].We can now address the most stable sites for Pt substitution into the Co-Mo-S particle. By substitution of one Mo atom with one Pt atom in the Mo-edge model, and one Co atom with one Pt atom in the S-edge and corner models, we first obtain equilibrium structures for the Pt-Co-Mo-S sites by calculating the free energy of sulfidation and desulfidation around the Pt-promoted site at HDS conditions (Supplementary Material). The substitution free energies have then been obtained for reaction (2) and (3) and are presented in the bottom of Fig. 4a together with the Pt-Co-Mo-S edge and corner equilibrium structures. As a reference for our DFT calculations we choose PtS2 (rather than PtS) since this particular platinum sulfide structure is what our XANES results indicate (Fig. 3b). We find it is energetically favorable to incorporate Pt single-atoms into both Mo- and S-edges and corner sites. The calculations indicate a slight preference of the Mo-edge over the corner site, and to a lesser extend the S-edge, however, within the uncertainty of calculations; Mo-edge: −0.50 (+/−0.44) eV, S-edge: −0.16 (+/−0.30) eV, and corner: −0.38 (+/−0.26) eV, the corner and the Mo-edge are probably equally preferred substitution sites. We have also calculated the stability of Pt and PtS adsorbed on several different adsorption sites on the Mo-edge, S-edge and the corner equilibrium structure of Co-Mo-S (see Supplementary Material). We find that Pt substitution into Co-Mo-S is significantly more stable than adsorption on the Co-Mo-S particle.Next, we address the structure of the Co-Mo-S and Pt-Co-Mo-S sites at HDS conditions (as shown in Fig. 4a) in more detail. In Fig. 4b we show how the fraction of sulfur vacancy sites on edges and corners of Co-Mo-S and Pt-Co-Mo-S structures relates to the free energy of sulfidation. We find that the free energy of sulfidation of a vacancy site at the Co-Mo-S Mo-edge is −0.11 (+/−0.19) eV, and for the corner vacancy 0.08 (+/−0.16) eV. Thus, as seen in Fig. 4b, vacancy and monomeric S sites are likely to co-exist at the Mo-edge and corner of the Co-Mo-S particle, however, to which degree is shadowed by the uncertainty of the calculation. On the other hand, platinum single-atoms incorporated into the edges and the corner of the Co-Mo-S particle leads to a weakening of the sulfur binding energy around the Pt atom compared to the non-Pt counterpart (Fig. 4b), such that the Pt sites at the Mo-edge and corner are indeed characterized by inherently stable vacancy sites. In fact, at the Mo-edge, Pt incorporation leads to a double CUS-vacancy around the Pt atom, which is a significant restructuring compared to the non-Pt counterpart. At the S-edge we see a slight shift in the free energy of sulfidation when Pt is incorporated (Fig. 4b), however this does not lead to vacancy sites.Double vacancy sites at the corners of Co-Mo-S have previously been suggested to be attractive for HDS [28], however, such sites are prohibited due to a high formation energy at HDS conditions. Here, we have also calculated the formation energy of such a double vacancy site from the vacancy sites at the corners of Co-Mo-S and Pt-Co-Mo-S (geometry given in Supplementary Material). At HDS conditions, we find a free energy of formation of 1.07 eV and 0.7 eV for Co-Mo-S and Pt-Co-Mo-S corners, respectively. Although Pt indeed lowers the formation energy of a double vacancy by ~ 0.4 eV relative to the corresponding Co corner, also the Pt-corner double vacancy at these sites are unstable at HDS reaction conditions.Visualization of the Pt-Co-Mo(0.5) catalyst at the atom-by-atom scale was approached using high-resolution STEM imaging [52–54]. Previously atomically resolved imaging of Co-Mo-S structures [13,55,56] has required use of a high-surface area carbon as carrier but we have here achieved such resolution for catalysts based on an industrially much more relevant high-surface area γ-Al2O3 carrier. The high-angle annular dark-field (HAADF) image contrast scales approximately as Z1.7 where Z is the atomic number and the Pt atoms (Z = 78) therefore appear much brighter than single Mo (Z = 42) or Co atoms (Z = 27), providing the support is sufficiently light and thin. The industrial Co-Mo-S structures are supported on a few nm thin alumina crystallites, with dimensions and orientations occasionally thin enough for imaging single heavy metal atoms [57]. However, alumina is an insulator with a very poor electrical conductivity and charging of the sample in the electron beam complicates high-resolution imaging. Therefore, we dispersed the alumina-supported catalyst onto a TEM grid with a conducting amorphous carbon layer to compensate for the electrical resistivity of alumina at the expense of image contrast. Fig. 5 a shows a high-resolution STEM image of the Pt-Co-Mo(0.5) sample and reveals a truncated hexagonal shaped nanocrystal about 3 nm in diameter with a periodic array of bright dots indicating heavy atom columns. An intensity profile across the nanocrystal shows four intensity peaks separated by about 0.32 nm consistent with the Mo—Mo distance of MoS2 (Fig. 5b,c). The nanocrystal is therefore attributed a MoS2 structure viewed in (001) projection in agreement with previous studies [55]. We note that the size of the crystal is in accordance with the MoS2 slab lengths viewed edge-on in TEM images (Fig. 3) and we show below by comparison to STEM image simulations that the contrast is consistent with metal atoms in a single-layer slab structure.A wider image field-of-view (Fig. S5, Supporting Material) shows the presence of both hexagonally shaped MoS2 nanocrystals as well as more irregular shaped MoS2 nanocrystalline domains viewed in (001) projection with multiple corner sites and kinks of a concave geometry, and even isolated (metal) single-atoms or few-atom clusters. Such a co-existence of both regular and irregular shaped MoS2 structures in alumina-supported catalysts have previously been reported [53,54]. Furthermore, restructuring of the MoS2 morphologies was observed in successive recorded images as an effect of the electron beam (in line with [58]) under the present imaging conditions. However, such image series also revealed distinctively bright dots associated with the edges of the MoS2 crystals (as indicated by circles in Fig. 5b) in the first acquired images that either remained at the same position or has moved to new positions in the subsequent image (Fig. S6, Supporting Material). To address the atomic arrangement in such nanocrystals, we carry out a detailed image contrast analysis of the MoS2 nanocrystal in Fig. 5, which represents a pristine area not previously exposed to the electron beam.The intensity profile along the edge reveals three distinct contrast levels (Fig. 5c). To ease the interpretation of this contrast pattern, STEM image simulations of a supported MoS2 structure was performed. A single-layer MoS2 slab structure consisting of 16 metal atoms (Mo, Co, Pt) and 36 sulfur atoms was used as model for the STEM image simulations, with an edge length of 4 metal atoms (using the S-edge with 50% sulfur coverage). To overcome the effect of contrast reduction from any (alumina, carbon) support materials, the (Mo,Co,Pt)-MoS2 model structure was placed on top of an amorphous carbon model structure with a varying thickness of 0–15 nm. Not surprisingly, in the simulated HAADF-STEM images of an unsupported or thin (3 nm) carbon supported MoS2 the atomic (Mo) metal lattice as well as the sulfur sub-lattice could be resolved, whereas at thicker supports of 9–15 nm the sulfur columns showed significantly reduced contrast with intensities visually mixed up with the intensity fluctuations of the amorphous support (Fig. S7, Supplemental Material). This is in accordance with a thin, flat (crystalline) support like graphene or graphite being the preferred support materials for obtaining single atom sensitivity imaging of MoS2 [55,58,59]. In contrast, however, the atomic imprints of Mo, Co or Pt single atoms were distinctly resolved in the simulated STEM images up to 15 nm of amorphous carbon support as a result of the high Z-number. In addition, image contrast of single atoms is insensitive towards small sample tilts away from a crystal zone-axis as opposed to e.g., diatomic (sulfur) columns. Thus, the following discussion will only focus on the metal atom positions.The image contrast of the individual supported metal atoms in the simulated STEM images was quantitatively addressed by measuring the intensity maxima of the corresponding atom peak positions to evaluate the confidence level for atom identification. We find that even for the unsupported MoS2 crystal, the peak maxima of the various Mo atoms show a range of intensities rather than a single value as a result of the comprehensive imaging model [60] used to simulate the sampling of the electron probe over the electrostatic potential of the specimen. Hence, a distribution of metal atom intensities measured from the HAADF-STEM images are expected. Importantly, the contrast of a single Pt atom was significantly brighter than any Mo atom or Co atom on a support thickness up to 15 nm (Fig. S9, Supplementary Information). On the other hand, the Co and Mo atoms were convincingly separated by contrast only at very thin supports (3 nm) as for thicker supports (9–15 nm) the contrast levels of Co and Mo started to overlap. The supported Co atoms, however, still appeared with lower contrast than the average Mo atom. Based on the maximum overlap of the peak intensity distributions the Co atoms could be discriminated from Mo with a confidence level of 84% on a 9–15 nm thick amorphous carbon support in the simulations.A comparison of the experimental data and the normalized image simulations is shown in Fig. 5c. The overall contrast levels of the (experimental) MoS2 structure and the adjacent support materials can be well described by the 9–15 nm amorphous (carbon) supported single-layer MoS2 model structure although the (experimental) support might not be completely flat and includes alumina as well. However, the detailed contrast levels could not be fitted with single Mo and Pt atoms only (shown in grey) without exceeding the uncertainty in the image contrast given by the statistical noise in the raw image (error bars). Instead, the simulations show that two Co atoms incorporated on that edge gives a better agreement (Fig. 5c,d). That is the relative contrast level analysis indicates the formation of a single-layer Pt-Co-Mo nanostructure. Moreover, the tertiary Pt-Co-Mo sulfide structure provided by a Pt corner atom in the Co-Mo-S structure is an energetically stable configuration according to our DFT results (Fig. 4). Thereby, we substantiate the argument that the nanocrystal in Fig. 5 is indeed a cobalt promoted single layer MoS2 crystal (Co-Mo-S phase) with a corner Pt atom. An unambiguous atom identification may possibly only be obtained using complementary atom-scale spectroscopy with careful optimized illumination and imaging schemes [13,52]. However, with the atomic assignments in place we point to the Co promoted edge as the S-edge of the MoS2 structure (Fig. 5d) and by symmetry, the opposite edge must be an Mo-edge.More interestingly, from an intensity analysis of the peak maxima over the entire nanocrystal (Fig. S10, Supplementary Material), six atom positions stand out by having a significantly larger intensity more than 3x sigma above the Gaussian distribution fit to all peak maxima and includes the corner Pt atom just identified. These six atom positions are all marked by circles in Fig. 5b and are assigned Pt single atoms with four corner site positions (C) and two Pt atoms associated with the low-indexed edges (E, E’), i.e. the Mo-edge and S-edge, respectively. Based on the location of these six Pt atoms in the nanocrystal in Fig. 5 we thereby justify all the three plausible scenarios of single-atom Pt-Co-Mo-S interactions discovered by DFT calculations (Fig. 4) with indications of the corner site position being the more pronounced.Thus, the Pt promotional effect of the industrial Co-Mo catalysts with Pt added prior to a sulfidation step may be associated with incorporation of Pt in the Co-Mo-S edge structures (referred to as the Pt-Co-Mo-S phase) rather than adsorption or a separate Pt (-sulfide) phase, e.g. PtS which elsewhere have been observed as a stable phase at HDS reaction condition at higher temperatures (400 °C) [61], and also suggested in the case of platinum doping of a pre-sulfided NiW catalyst [62]. The Pt-Co-Mo-S phase, however, is consistent with a Pt(IV)-sulfide resembling structure with Pt—S bonds as revealed by XANES. According to our DFT investigations, the platinum in the Pt-Co-Mo-S structure has a sulfur coordination of 3–4 under HDS conditions (Fig. 4). The linear increase in HDS activity with increased amounts of Pt (Fig. 2a) is consistent with single atoms gradually being incorporated into edge sites of a nanocrystal. In particular, the promotional effect of Pt and Co in combination at various contribution ratios (Fig. 2b) is rationalized from the incorporation of Pt in corner sites or edge sites of a partly cobalt decorated MoS2 in which Mo and Co atoms co-exist at the S-edge as visualized by electron microscopy (Fig. 5).In addition, the amount of Co and Pt relative to Mo per Co-Mo-S slab can be evaluated based on the nominal metal load on the catalysts using simple geometric considerations. Assuming single-layer Co-Mo-S slab structures of length ~ 2.5 nm, this corresponds to a cluster size of N = 61 metal atom sites in a regular hexagonal shape with a side length of 4 Mo—Mo distances (5 atoms) and 9 metal atoms across the longest diameter. We find that 3.5 wt% Co corresponds to Co:Mo = 0.35 (molar ratio) which is equivalent to about 16 atoms out of the assumed average cluster size of 61 metal atoms, in case of full Co incorporation into Co-Mo-S. This would correspond to 66% of the edge metal atoms and agrees with a Co-Mo-S structure with Co decorating the S-edges only. Using a Co-Mo-S slab version with 16 Co atoms and 45 Mo atoms, the corresponding addition of Pt single atoms is Pt:Mo = 0.022, 0.044 or 0.067 (molar ratio) for 1, 2, or 3Pt atoms per Co-Mo-S slab, respectively. Hence the activity maximum at 146% of Fig. 2a occurs at a degree of Pt-promotion corresponding to Pt:Mo = 0.03–0.04 or about 2Pt atoms per Co-Mo-S slab on average. At saturation, Pt separates out into the observed competing second phase of Pt clusters at a high precursor density [50,63,64], which are probably poorly catalytically active. As minor local variations in Pt precursor or metal concentrations at the nano-scale may be expected during synthesis, it can be speculated that the origin of saturation by a few Pt atoms per Co-Mo-S slab on average can be geometrically constrained as the corner sites of the (hexagonal) Co-Mo-S structures are gradually filled and occupied by Pt at increasing load.With the Pt-Co-Mo-S catalyst model in mind we investigate the hydrodesulfurization of the sterically hindered 4,6-DMDBT molecules based on DFT calculations. Such 4,6-disubstituted species are present (see Supplementary Material) in the heavy end of the industrial feed used (a mix of straight run gas oil and light cycle oil) and are, due to sterically hindrance in the vicinity of the sulfur atom, the most difficult sulfur species to convert by hydrotreating. Analytics such as e.g. 2-dimensional gas chromatography show that 4,6-disubstituted DBTs constitute essentially all those organosulfur compounds that remain to be converted when a sulfur level of a few hundred ppm S has been achieved. To reach the desired ULSD target of only 10 ppm S, most of the catalyst volume in the reactor serves the purpose of desulfurizing the relatively very small amount of a few hundred ppm 4,6-disubstitued DBTs (Supplementary Material). At the test conditions employed in this study sulfur concentrations below 20 ppm were obtained in the oil exiting the reactors, i.e. concentrations well within the domain where essentially only 4,6-substituted DBTs are being converted. As the H2 partial pressure was 30 bar, our conditions are suited for evaluating the performance of catalysts aimed at the industrial so-called low-pressure segment, i.e. Co-Mo catalysts. Industrial refiners that have decided on process equipment to produce ULSD in the high-pressure segment would use Ni-Mo rather than Co-Mo catalysts.The exact mechanism for desulfurization of 4,6-DMDBT on Pt-Co-Mo-S is considered to be quite intricated giving the size of the molecule(s) and the complexity of the edge sites. To provide a first preliminary assessment we here address with DFT a possible effect of Pt by investigating the stability of reaction intermediates in two hypothetical DDS and HYD reaction pathways of 4,6-DMDBT desulfurization at Pt-Co-Mo-S and Co-Mo-S sites. We especially focus on the thiolate formation, since thiolate formation has been suggested to be a starting model for which sites participate in the C—S cleavage of a sulfur containing molecule [25]. Fig. 6 shows the hypothetical DDS and HYD pathways we investigate. Species denoted with an asterisk are adsorbed species. For both the DDS and HYD pathway the first step is the adsorption of 4,6-DMDBT on the (Pt)-Co-Mo-S particle. In the DDS pathway, a thiolate intermediate (DDS-thiolate) is formed after C—S cleavage via hydrogenation. This species can then undergo a second round of C—S bond breakage to form 3,3′-DM-BP, which desorbs from the site. In the HYD pathway, the adsorbed 4,6-DMDBT is hydrogenated six times to 4,6-HH-DMDBT, such that one benzene ring has been fully hydrogenated. This species can now undergo C—S bond breakage via hydrogenation of either i) the C—S bond between the S atom and the aromatic ring or ii) the C—S bond between the S atom to the cyclohexyl ring to form a HYD-thiolate. As for the DDS pathway the thiolate undergoes a second C—S bond breakage via hydrogenation, and the desulfurized product, 3,3′-DM-CHB, leaves the site.In Fig. 7 we show the calculated energy diagram for both the DDS and HYD pathway of 4,6-DMDBT on the Co-Mo-S and Pt-Co-Mo-S equilibrium structures of the S-egde, Mo-edge and corner sites. We have chosen also to include the Mo-edge with a single vacancy, since both the 1S terminated Mo-edge and the single vacancies are likely (see Fig. 4b).We start out with the energetics of the DDS pathway in Fig. 7a where adsorption of 4,6-DMDBT is the first reaction step. Beginning with the Co-Mo-S structure, we find the 4,6-DMDBT molecule to adsorb strongest through physisorption at the brim site of the Mo- and S-edges, and the calculations indicate a slightly weaker physisorption to the brim of the defect (vacancy) Mo-edge site or corner site (see molecular configurations in Supplementary Material). At the Co-Mo-S corner, 4,6-DMDBT adsorbs equally strong physisorbed at the brim and in a chemisorbed state aligned in-plane with the particle in agreement with previous work [25] (only configurations free of the support were considered). Introducing the Pt sites in the Co-Mo-S structure slighty weakens the 4,6-DMDBT adsorption strength on both the Mo-edge and the corner site whereas essentially no effect is obtained at the S-edge.Next we address the DDS-thiolate from 4,6-DMDBT on both Co-Mo-S and Pt-Co-Mo-S sites. According to the calculations, the DDS-thiolate is most strongly adsorbed on corner and edge vacancy sites to which it can chemisorb, consistent with that thiolates in generel are found to adsorb strongest on defect sites [25]. However, at all investigated sites except the non-Pt promoted Co-Mo-S corner site we find, interestingly, that the DDS-thiolate formation reaction step is uphill in energy. Thus, since both the adsorption of 4,6-DMDBT and the thiolate formation reaction step is energetically preferred on Co-Mo-S (corner sites) over Pt-Co-Mo-S, these calculations suggest that Pt sites has limited or no promotional effect for the DDS pathway.We now turn to the HYD pathway for which the energetics is shown in Fig. 7b. Here the adsorbed 4,6-DMDBT is hydrogenated to 4,6-HH-DMDBT before the C—S bond is cleaved and the corresponding thiolate is formed. For the Co-Mo-S and Pt-Co-Mo-S structure, as for 4,6-DMDBT we find that 4,6-HH-DMDBT preferentially physisorbs at the brim of the 1S-terminated Mo- and S-edges and with a weaker chemisorption in-plane with the particle for the corner sites (see Supplementary Material for molecular structures). The HYD-thiolate, resulting from 4,6-HH-DMDBT, adsorbs preferentially to the vacancy sites through chemisorption by the S-part of the molecule to the vacancy. This is valid for both the HYD-thiolate a and b (see Supplementary Material). However, on all sites, except on the vacancy site of the Mo-edge, the HYD-thiolate a, formed from breaking the C—S bond between the aromatic ring and the S-atom, is more stable than HYD-thiolate b resulting from breaking the C—S bond between the S-atom and the cyclohexyl ring. Importantly, we find the formation of the HYD-thiolate to be energetically favourable on vacancy sites and unfavourable on 1S-terminated sites, suggesting that vacancy sites could be key sites for the final steps of desulfurization of 4,6-DMDBT. Here, we have not considered energy barriers for thiolate formation, however, it has been found that barriers for thiolate formation specifically from thiophene and dibenzothiophene are significantly lower on defect sites at the Mo-edge and the corner, than on the 1S-terminated S-edge [25]. We also note that the HYD-thiolate formation is significantly more favourable on all vacancy sites of both the Co-Mo-S and Pt-Co-Mo-S structure than the strongest DDS-thiolate formation on the Co-Mo-S corner, indicating desulfurization of 4,6-DMDBT proceeds primarily via the HYD pathway, consistent with experimental findings on a Co-Mo catalyst [21,22]. Overall, this could suggest that different sites may participate in the desulfurization such that 4,6-DMDBT preferentially is hydrogenated at the 1S-terminated brim of the particle, where we find the strongest adsorption of 4,6-DMDBT and the hydrogenated intermediate, and the hydrogenated species then diffuse to a vacancy site, where the final desulfurization of 4,6-DMDBT takes place.The reactivity of corner sites of Co-Mo-S has recently attracted attention [25,65] and specifically adsorption of 4,6-DMDBT molecules has been observed by STM to occur particularly at corner (vacancy) sites [28]. Thus, a corner platinum site in the Co-Mo-S structure, as visualized in the present study, is obviously highly intriguing. Even though intermediates in the HYD pathway adsorb weaker to the Pt-corner in Pt-Co-Mo-S than the corresponding Co-Mo-S corner, the inherent vacancy site of the Pt-Co-Mo-S corner still favors HYD-thiolate formation (Fig. 7b and Supplementary Material). Likewise, at the Mo-edge the single-atom Pt incorporation promotes a double CUS site for which the calculations indicate HYD-thiolate formation is favored. Double CUS sites at the Mo-edge generated in the presence of thiophene was recently observed by STM [66] and suggested to activate the (Mo)-edges of MoS2 towards sterically hindered molecules. Thus, the Mo-edge of the Pt-Co-Mo-S structure provides such double CUS sites available intrinsically, which may also explain enhanced HDS activities [38] of Pt-MoS2 systems without cobalt. Therefore, we propose that the high reactivity of Pt-Co-Mo catalysts is obtained by lowering the sulfur binding energy around the Pt single-atoms at corners and (Mo)-edges in a Pt-Co-Mo-S phase making such vacancy sites more abundant (Fig. 4b). The Pt sites in the catalyst still has ability to bind 4,6-DMDBT and especially the HYD-thiolate sufficiently to favor its formation. This could suggest that despite a weaker S-binding, these sites maintain reactivity towards C—S scission of hydrogenated reaction intermediates from 4,6-DMDBT, although more elaborate studies are needed to address that in detail.Finally, the presence of highly irregular shaped MoS2 crystal morphologies observed in the tested HDS catalysts indicates the availability of other site geometries than the low-index edge structures and corner sites considered in the DFT models in the present analysis, e.g., kink sites of a concave geometry, which can be speculated to mediate hydrogenation and desulfurization favorable sites in combination. Thus, the concentration of catalytically important edges under industrial operations, i.e. the evolution in the number of a particular atomic site geometry, are key to HDS catalysis which in turn depends on the activation procedure of the catalyst (e.g. gas phase sulfidation vs. liquid sulfidation [67,68]), operating conditions, and promoters as well.The present work demonstrates a tertiary transition metal sulfide-based catalyst for hydrotreating processes. Addition of ppm-levels of Pt to an industrial Co-Mo-S catalyst leads to a remarkable increase of up to 46%, as compared to the Pt-unpromoted catalyst, in the activity for low-pressure hydrotreating aimed at producing ultra-low sulfur diesel oil. This boosted performance is attributed to the Pt atoms incorporated into terminating edge and corner sites of the Co-Mo-S nanostructure as resolved by combining X-ray absorption spectroscopy and atomic-resolution electron microscopy. Interplay with density functional theory calculations suggests that Pt acts by reducing the sulfur binding energy under HDS conditions compared to the Co-Mo-S structure only. This mechanistic effect is most pronounced at the Pt-promoted corner site and at the Mo-edge. Since the Pt favors CUS formation and the simple probe HYD-thiolate adsorption is similar on the Pt promoted and unpromoted vacancy sites, the role of Pt is speculated to create sites for favorable adsorption of sterically hindered molecules such as 4,6-DMDBT to undergo desulfurizationutilizing a functionality of adjacent Mo- and S-edges, or corner sites synergistically.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.The authors gratefully acknowledge Christoffer Tyrsted for help with XANES measurements. We acknowledge use of facilities at NMI Natural and Medical Sciences Institute at the University of Tübingen (Germany), Center for Nanoanalysis, for part of this work, and Clementine Warres for kind support. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. SH acknowledges support from Danish National Research Foundation Center for Visualizing Catalytic Processes (VISION) (DNRF146).Supplementary data to this article can be found online at https://doi.org/10.1016/j.jcat.2021.03.008.The following are the Supplementary data to this article: Supplementary data 1

We introduce a tertiary transition metal sulfide nanostructure, Pt-Co-Mo-S, for catalytic hydrodesulfurization (HDS) of sulfur-containing molecules in crude oil distillates with the aim to produce ultra-low sulfur transport fuels. The addition of ppm-levels of Pt to a standard industrial Co-Mo-S catalyst boosts the HDS activity by up to 46%. The promotional effect is examined by combining atomic-resolution Scanning Transmission Electron Microscopy (STEM), Energy Dispersive X-ray Spectroscopy (EDX), X-ray Absorption Near Edge Spectroscopy (XANES) and Density Functional Theory (DFT). It is shown that the Pt-Co-Mo-S catalyst contains predominantly single-layer MoS2 nanocrystals with Co atoms fully covering the S-edge terminations and Pt atoms uniquely attached to corner and edge sites in a platinum(IV) sulfide-like structural motif. Platinum is suggested to reduce the sulfur binding energy and increase the abundance of coordinately undersaturated sites (CUS) and not necessarily changing the reactivity towards 4,6-DMDBT molecules, although more elaborate studies are needed to address this in detail.

Activated carbon fiber, also known as the third generation of activated carbon material, is a new type of activated carbon material after powder and granules, made of the organic fiber. Activated carbon fiber has a strong adsorption capacity due to its large specific surface area, and can adsorb and remove organic matter in water, including some carcinogenic or toxic macro-aromatic substances. Physical adsorption is caused by electrostatic interaction between the surface molecules of the adsorbent and the adsorbate molecules, which is characterized by adsorption without selectivity. Chemical adsorption is due to the chemical bonding between the two molecules, forming a strong chemical bond and also the surface complex. Chemical adsorption has the certain selectivity, generally monolayer adsorption. With the large increase in vehicle usage and the exhaust emissions from equipment such as oil-fired boilers, the pollution of nitrogen oxides and carbon monoxide gas has become an important issue in environmental management. Adsorption or catalytic conversion of nitrogen oxides and carbon monoxide is an important means of treatment. This paper integrates the smart city scenario to construct the novel scenario for the analysis and research. The simulation results prove the effectiveness.

Over the last decades, the use of Bronsted acid catalysts has been increased in both laboratories and industries processes, such as biodiesel production, esterification and hydrolysis because of liquid phase limitations, hard separation, non-recovery acidic waste generation and the corrosion of reactors [1, 2, 3, 4]. Nano-substrates with a larger surface area and high surface-to-volume ratio have been applied in the solid acid catalysis field to increase available catalytic centers and enhance catalytic activity [5, 6]. Among solid nano catalysts, magnetic core-shell nano catalysts have been widely used because of their easy separation and recovery by applying a magnet as well as functionalization possibilities of the inorganic surface. Magnetic field separation is more effective than conventional separation methods such as purification and centrifuge because it can reduce catalyst waste, optimize operational cost and enhance the purity of products [7, 8, 9, 10]. By choosing the appropriate shell, the functionality of the catalyst in the shell is improved. Besides, this shell prevents the accumulation of magnetic nanostructure and enhances its chemical stability [11].Iron oxide magnetic core-shell nanoparticles have been prepared in various ways and have been modified by functional groups with catalytic properties [12]. Silica is one of the best coatings used for the synthesis of iron oxide core-shell nanoparticles because of stability, biocompatibility, functionality and resistance to acid or high temperature [13, 14]. The acid catalytic properties of iron oxide@SiO2 nanoparticles can be exceeded by the modification of the surface of nanoparticles with acid functional groups, such as HBF4 [15], sulfuric acid [16], R-NHMe2][H2PO4] [17] and sulfonic acid [18]. Solid acid catalyst functionalized with sulfonic acid could be used in different organic reaction [19, 20, 21]. The iron oxide coated with silica and functionalized with sulfonic acid (Fe3O4@SiO2–SO3H) nanoparticles has been used as an effective catalyst in the synthesis of pyrimidinones [22], pyrazole [23], quinoline [24], pyrazolopyridines [25] and dihydropyrimidinones/thiones [26] as well as in the reduction of oximes to amines [27] and esterification reactions [28]. However, all the above-mentioned catalysts are nanoparticles as a zero-dimensional nanostructure. It seems that the use of other nanostructures with different morphology (such as one-dimensional nanostructure) as a catalyst in organic reaction is very limited and needs further study.Nanofibers are a new generation of one-dimensional nanostructures frequently used for their different applications like filtration [29], membrane [30], sensor [31], food and packaging [32], medical applications [33] and catalysis [34]. Electrospinning followed by calcination provides a tunable, simple and versatile way for generating ceramic nanofibers with unique properties including high surface area, large porosity, mechanical roughness, superb thermal stability, higher electron transfer and enhancement of the thermo electric merit and magnetic moment [35]. Such new properties lead to the potential application of ceramic nanofibers in lithium batteries, data storage devices, magnetic resonance imaging, sensing devices, energy storage, targeted drug delivery and catalysis [36, 37, 38]. Recently, some researchers have focused on the fabrication of porous ceramic nanofibers and their applications in heterogeneous reactions [39]. Pd–TiO2 nanofibers have been applied in Suzuki coupling reaction with high conversion efficiency due to their high surface area and larger number of active sites [40]. Cu25O/Ni75O nanofibers can be used as anode catalysts for hydrazine oxidation in alkaline [41]. In one study, Pt/TiO2 nanofibers were used during electrochemical reaction for methanol oxidation and were found to effectively facilitate electron transport [42]. In another study, Pt supports composed of graphene sheets decorated by Fe2O3 nanorods and denoted as Pt/Fe2O3/N-RGO displayed higher catalytic activity than free Pt in the 4-nitrophenol hydrogenation reaction [43]. Previously, our research group prepared iron oxide nanofibers and applied them as catalysts in alcohol oxidation [44] and methyl orange degradation as a Fenton catalyst [45]. Also, in other studies, our group examined the fabrication and catalytic application of core-shell Fe2O3@SiO2 nanofibers as novel magnetic nanofibers for the one-pot reductive amination of carbonyl compound [46, 47]. It seems that by functionalization of the surface of Fe2O3@SiO2 nanofibers, this magnetic solid nano catalyst can be used in various organic reactions.Formamides are important crossroad intermediates in organic component synthesis such as formamidines [48], isocyanides [49], Vilsmeier reagents [50], fluoroquinolones [51] and imidazoles [52]. Hence, researchers have focused on the synthesis of formamides via the reaction of amines with various N-formylating agents such as carbon dioxide [53, 54], phenyl chloroformate [55], methyl formate [56], ammonium formate [57] and acetic formic anhydride [58]. Of these reagents, formic acid as a cheap, non-toxic, stable formylating compound is a good candidate for N-formylating amines. Unfortunately, their moderate reactivity demands increased reaction temperatures or catalyst presence, particularly for aromatic and sterically-hindered amines formylation. So, formic acid has been used in the presences of catalysts such as poly ethylene glycol [59], zeolite [60], ZnO [61], ZnCl2 [62], thiamine hydrochloride [63] and cobalt nanoparticles [64]. Nishikawa et al. studied the synthesis of N-formamides with formic acid under mild conditions using tetraethylorthosilicate (TEOS) [65]. However, a long reaction time requires a Dean-Stark trap in reflux conditions and the difficult N-formylation of electron-withdrawing derivatives of anilines can limit the usage of these catalysts. Besides, the above-mentioned catalyst recycles need to be centrifuged or a long-time filtration leading to the unavoidable loss of solid catalyst in the process of separation. Thus, it seems that magnetic nanostructures modified with silica and reinforced with an acid functional group can be a good choice as a catalyst for N-formylation.Formamidines are widely used as effective drug agents and useful intermediate in the synthesis of purine [66] imidazole and quinazoline [67] compounds. Formamidines and related anions can bind transition metals. Recently, formamidines derivatives with Ni (II) [68] Au (I), Ag (I) [69] Co (II) [70] Cu (II) [71] Pt [72] and Pd [73] ions have been synthesized. Amidine functional group has various biological properties, such as anti-virus [74], anti-AIDS [75], anti-degenerative [76], anticancer [77], anti-platelet [78] and antimicrobial [79] activities. Finding simple and moderate conditions for the synthesis of formamidine can be useful in the production of many organic compounds.Continuing our previous research [44, 45, 46, 47], this study sought to prepare Fe2O3@SiO2–SO3H nanofibers and study their use as a heterogeneous acid catalyst in organic reactions aiming at introducing a new generation of one-dimensional nano catalysts in organic reactions. For this purpose, iron oxide nanofibers, prepared by a combination of electrospinning and calcinations, were coated with silica using tetraethyl orthosilicate. Next, the silica surface was reacted with chlorosulfonic acid to obtain Fe2O3@SiO2–SO3H nanofibers. Due to the significance of formamide group, we decided to examine the catalytic performance of Fe2O3@SiO2–SO3H nanofibers as a novel magnetic nano catalyst in the N-formylation of amines using formic acid. We also decided to study the catalytic ability of Fe2O3@SiO2–SO3H nanofibers in the synthesis of formamidine via the reaction of different amines with triethyl orthoformate as a second catalytic reaction.Poly (vinyl alcohol) (PVA) of 88,000 g/mol molecular weight and 88% degree of hydrolysis were obtained from Sigma-Aldrich (USA). The chemicals used in the study were all purchased from Merck Company. No further purification was done to the reagents with deionized water used as a solvent.Fe2O3@SiO2 nanofibers were prepared based on the method employed in our previous study [46]. In summary, 10 ml of 8% W/W PVA solutions were prepared in deionized water as a solvent. Then, 2 g ferric nitrate (Fe(NO3)3.9H2O) was added and stirred for 1 h at room temperature. The electrospinning process was conducted by Electroris (FNM Ltd., Iran, http://www.fnm.ir) that as an electrospinning device can control electrospinning parameters such as high voltage (1–35 kV), injection rate with syringe pump (0.1–100 mL/h), drum rotating speed (0–1000 rpm), drum-to-nozzle distance (0–300 mm), needle scanning rate (0–100 mm/min) and the electrospinning media temperature. The electrospinning condition of the prepared solutions was as follows: 5ml syringe with an 18-gauge needle as a nozzle, a rotating drum with 300 rpm speed covered with an aluminum foil as a collector, the applied voltage of 20 kV, 2 ml/h for the rate of injection and 12 cm for the distance between the collector and nozzle. Next, the electrospun PVA/Fe(NO3)3.9H2O nanofibers were peeled off from foil and then calcinated at 600 °C for 6 h to entirely eliminate the organics with the heating rate of 20 °C/min, which produced iron oxide (Fe2O3) nanofibers.In the next step for coating with SiO2, a mixture of ethanol (15ml), deionized water (5ml) and aqueous ammonia solution (1 ml, 25 wt%) were obtained and stirred completely; then 25 mg of obtained Fe2O3 nanofibers were added to this mixture and stirred at room temperature. Then, the solution of 30 ml ethanol and 1 ml tetraethyl orthosilicate (TEOS) were mixed with dispersed Fe2O3 nanofibers using a syringe pump. Finally, after relaxation for 6 h, the core-shell nanofibers were separated using a super magnet and washed with ethanol and deionized water.200 mg of Fe2O3@SiO2 nanofibers were added into the mixture of diethyl ether (1 ml) and chlorosulfonic acid (0.6 ml) and stirred for 3h at room temperature until the release of hydrochloric acid gas was stopped. The residue was washed with diethyl ether and deionized water and dried for 6h at 100 °C in an oven to obtain the desired nanofibers.The amount of acid on nanofibers was determined using titration with base. For each titration operation, 50 mg of nanofibers was weighed and transferred to a beaker. Then, 2 ml from potassium hydroxide 0.1 M was added. The prepared suspension was stirred by a magnetic stirrer for 15 min. Then, the nanofibers were separated by a magnet, and the clear solution was titrated with 0.1 M hydrochloric acid in the presence of phenolphthalein.The characterization of obtained products was carried out using scanning electron microscope (SEM), energy dispersive x-ray spectroscopy (EDS), transmission electron microscope (TEM), X-ray diffraction (XRD), Fourier transform infrared spectroscopy (FT-IR) and, vibrating sample magnetometer (VSM). SEM images were observed using SEM (PhilipsXL30 and S-4160) with gold coating equipped with EDS. TEM measurements were done at 120 kV (Philips, model CM120). Powder XRD spectrum was recorded at room temperature by a Philips X'pert 1710 diffractometer using Cu Kα (α¼ 1.54056 Å) in Bragg-Brentanogeometry (θ-2θ). FT-IR spectra were obtained over the region 400–4000 cm−1 with NICOLETIR 100 FT-IR and spectroscopic grade KBr. The magnetic properties of catalyst were attained by Vibrating Sample Magnetometer/Alternating Gradient Force Magnetometer (VSM/AGFM, MDK Co., Iran, http://www.mdk-magnetics.com).For each reaction, 1 mmol of triethyl orthoformate was transferred to a small glass flask. Then, 1 mol% of catalyst (0.01 mmol H+ in the catalyst, 25 mg Fe2O3@SiO2–SO3H nanofibers) was added. The obtained mixture was stirred by a magnetic stirrer for 15min. Then, 1 mmol of amine was added slowly to be stirred at room temperature (24–27 °C). The reaction progress was followed by thin-layer chromatography with hexane-ethyl acetate at a ratio of 1:4. After the completion of the reaction, 5 ml of dichloromethane was added, and the catalyst was separated by a magnet. The solvent was evaporated using vacuum distillation. The obtained precipitate in dichloromethane was dissolved and re-crystallized. Then, the isolated product was weighed on an electronic balance and used to compute percent yield based on isolated product weight in terms of the expected weight of the product. The synthesized compounds were identified by FT-IR, Mass and NMR spectroscopy and compared with those reported in the literature. The spectral data for selected products are presented below: N,N′-Bis-(3,4-dicholoro-phenyl)-formamidine (Table 3, entry 7): White solid, mp 134–135 °C; 1HNMR (CDCl3, 400MHz): δ7.25 (m, 2H), 7.28 (d, J = 6.40 Hz, 1H), 7.41 (br s, 1H), 8.39 (m, 2H), 8.5 (s, 1H), 8.85 (s, 1H).FTIR: 697 cm_1, 749 cm_1, 820 cm_1, 866 cm_1, 1048 cm_1, 1100 cm_1, 1147 cm_1, 1296 cm_1, 1396 cm_1, 1468 cm_1, 1524 cm_1, 1587 cm_1, 1665 cm_1, 2895 cm_1, 3077 cm_1, 3242 cm_1, 3395 cm_1. MS m/z: 334(M+). 1H-Benzimidazole ( Table 3, entry 9): Pale yellow solid, mp 169–170 °C, 1H NMR (CDCl3, 400 MHz): δ7.10–7.23 (m, 2H),7.61–7.64 (m, 2H), 8.17 (s, 1H), 10.31–11.1 (br s, NH).FTIR: 619 cm_1, 743 cm_1, 876 cm_1, 950 cm_1, 1001 cm_1, 1125 cm_1, 1202 cm_1, 1243 cm_1, 1303 cm_1, 1405 cm_1, 1456 cm_1, 1591 cm_1, 1703 cm_1, 2736 cm_1, 2806 cm_1, 2854 cm_1, 2933 cm_1, 3061 cm_1, 3425 cm_1. MS m/z: 118 (M+).10 mg of Fe2O3@SiO2–SO3H was added to 1.2 mmol aqueous formic acid. After five minutes of stirring, 1.0 mmol of amine was added and the reaction mixture was stirred at room temperature (24–27 °C). The reaction progress was followed by thin-layer chromatography with hexane-ethyl acetate at a ratio of 1:4. After the completion of the reaction, an external magnet was used to separate the catalysts. The reaction mixture was extracted with CH2Cl2 and H2O. The organic layer was then dried over anhydrous Na2SO4 and identified bythe1H NMR spectroscopy. The percent yield was calculated by the same procedure as mentioned in section 2.4. The spectral data for selected products are presented below: 4- morpholin carbaldehyde ( Table 4 , entry 8): colorless oil, 1H NMR (CDCl3, 400 MHz): δ3.37 (t, 2H, J = 5.1 Hz), 3.55 (t, 2H, J = 5.1 Hz), 3.62 (t, 2H, J = 5.1 Hz), 3.65 (t, 2H, J = 5.1 Hz), 8.04 (s, 1H). N-(tert-Butylamine) formamide ( Table 4, entry 6): colorless oil,1 HNMR (400 MHz, CDCl3): 50:50 (cis/trans) δ 1.32 (s, 9H), 7.28 (d, 1H, J = 2.3, cis), 7.43 (br, 1H, cis), 7.51 (br, 1H, trans) 7.67 (d, 1H, J = 12.2, trans).The generation of well-controlled ceramic nanofibers is typically conducted as follows: (i) the preparation of an electrospinning solution containing a polymer and sol-gel precursor of the ceramic material, (ii) electrospinning the polymeric solution under appropriate conditions and (iii) the calcination of the polymer/precursor composite nanofibers at high temperature to remove polymers and obtain the ceramic phase [35, 36]. Coaxial electrospinning with two immiscible components or polymer in a core-shell nozzle followed by calcination is a conventional electrospinning method used for fabricating core-shell ceramic nanofibers [39]. However, similar to our previous study [46], in this study one-dimensional Fe2O3@SiO2 nanofibers were prepared by a different and new method with the idea of preparing of Fe2O3@SiO2 nanoparticles using a coating of iron oxide (Figure 1 ). To do so, first, polymer/precursor composite nanofibers were fabricated by the routine uniaxial electrospinning of PVA polymer containing Fe(NO3)3 as an iron precursor. Iron oxide nanofibers were then obtained by calcinating polymeric nanofibers. Finally, these nanofibers were coated with silica by the sol-gel method in the vicinity of TEOS. The products of the three steps were characterized using SEM to approve one-dimensional nanostructure formation.The SEM images of PVA/Fe(NO3)3 nanofiber, Fe2O3 nanofiber and Fe2O3@SiO2 nanofibers are shown in Figure 2 . As can be seen in Figure 2 a, b, the successful electrospinning of PVA/Fe(NO3)3 solution led to smooth and fine one-dimensional polymeric nanofibers without bead. However, when composite nanofibers were calcinated at high temperature controlled the rising speed, the surface of Fe2O3 nanofibers appeared with rough surface, bended shape and a few ruptures in the axial direction (Figure 2 c, d). Finally, as shown in Figure 2 e, f, successful and homogenous coating of iron oxide nanofibers with TEOS produced fine Fe2O3@SiO2 nanofibers. The average diameters of nanofibers were calculated by the measurement software based on 15 fibers at random in SEM image. The average diameters of PVA/Fe(NO3)3 nanofibers, Fe2O3 nanofibers and Fe2O3@SiO2 nanofibers were 198 ± 20, 93 ± 16 and 142 ± 35 nm respectively, confirming that the diameter of composite nanofibers diminished during calcinations and the removal of organic phase whereas the diameters increased with the coating of Fe2O3 nanofibers.Continuing the coating, the functionalization of Fe2O3@SiO2 nanofibers was followed in this study in order to increase the acidity of surface to be used as novel heterogeneous acid catalyst. For this purpose, the silica surface was reacted with chlorosulfonic acid to obtain Fe2O3@SiO2–SO3H nanofibers. Figure 3 a, b indicates the SEM analysis of Fe2O3@SiO2–SO3H nanofibers with different magnification. It is observed that nanofibril structures were produced, but some rupture in fibers can be seen due to the increased acidity in the presence of strong acid functional that led to some corrosion in the direction of nanofibers, although the fiber structure with bigger direction than diameter is observable. The mean diameter of Fe2O3@SiO2–SO3H nanofibers was determined using measurement software and the fiber diameter distribution histogram was drawn by Origin software. As can be seen in Figure 3 c, diameter distribution is between 50 and 300 nm with a mean diameter up to 137 ± 12 nm for Fe2O3@SiO2–SO3H nanofibers. The TEM image of Fe2O3@SiO2–SO3H nanofibers is shown in Figure 3 d. As the TEM image shows, Fe2O3 nanofibers were coated with the uniform layer of silica that was functionalized with SO3H. It is observed that the dense layer of SiO2–SO3H as thick as 7 nm exists on the surface of Fe2O3 nanofibers. In contrast to Fe2O3@SiO2 nanofibers (15 nm for SiO2 layer [46]), it seems that the thinner layer of shell surrounded the magnetic core probably as a result of the corrosion of silica surface during the functionalization with chlorosulfonic acid.In order to determine the elemental composition and confirm the functionalization of nanofibers, Fe2O3, Fe2O3@SiO2 and Fe2O3@SiO2–SO3H nanofibers were evaluated by EDS analysis (Figure 4 ).As can be seen, the Fe and O patterns exist as the main elements in the quantitative analysis of the three nanostructures. Furthermore, Si element can be seen in Fe2O3@SiO2 and Fe2O3@SiO2–SO3H nanofibers patterns. The presence of S in the EDS analysis of Fe2O3@SiO2–SO3H nanofibers confirmed the successful functionalization of Fe2O3@SiO2 nanofibers surface with sulfonic acid. Figure 5 shows the FT-IR spectra of PVA/Fe(NO3)3 nanofibers, Fe2O3 nanofibers, Fe2O3@SiO2 nanofibers and Fe2O3@SiO2–SO3H nanofibers. The FT-IR of PVA/Fe(NO3)3 nanofibers in Figure 5 a exhibited various transmittance peaks at 1300-1800 cm−1 which were attributed to the functional group of PVA. The main peak in 1721 cm−1 corresponding to the (C =O) residual from primary vinyl acetate can be seen in Figure 5 a, but this peak decreased significantly in Figure 5 b, c and d due to the removal of PVA during calcinations at high temperature. In Figure 5 b two peaks appeared at 463 and 548 cm−1 which could be assigned to the Fe–O vibration of the Fe2O3 nanofibers [46] shifting to 461 cm−1 and 581 cm−1 in Fe2O3@SiO2–SO3H nanofibers. In all of Fe2O3@SiO2–HA Bronsted acids, the band at 400–650 cm−1 is assigned to the stretching vibrations of (Fe–O) bond [15, 80].The strong peak at 1031 cm−1 in Fe2O3@SiO2 nanofibers spectra (Figure 5 c) corresponds to the vibrations of the Si–O bond that appeared as shoulder of broad peak in 1000–1100 cm−1 at Fe2O3@SiO2–SO3H nanofibers spectra (Figure 5 d) as well as the peaks around 800 cm−1 correspond to Si–O–Si [81]. The broad peak that appeared at about 3409 could be shown the stretching of the OH group in the SO3H moiety [16]. Based on the results of previous studies on Fe2O3@SiO2–SO3H nanoparticle synthesis, it can be understood that functionalization with SO3H lead to a wider peak than Fe2O3@SiO2 nanoparticles at about 3400 cm−1 [16,23-28]. The same result could be found in Figure 5 by comparing the spectra. The peak at 1631 cm−1 represents the physical absorption of the lower amount of water [28]. The peaks of 1089 cm−1 and 1384 cm−1 are likely hidden under the wide peak at 1000-1400 cm−1 that are related to O=S=O stretching vibration in –SO3H groups [82,83].Diffraction peaks at around 23.97°, 28.29°, 29.97°, 33.09°, 35.49°, 40.29°, 49.41°, 53.97°, 62.37° and 63.81° which related to the (012), (031), (411), (104), (110), (512), (024), (116), (214) and (300) are readily recognizable from the XRD pattern (Figure 6 ). The observed diffraction peaks agree well with the structure of Fe2O3 (1999 JCPDS file No 24-0072 and 16-0653). It seems that the silica sheath of core-shell nanofibers was in amorph structure without the distinct sharp peak in the XRD pattern. However, the broad peaks at position 10-20° could be assigned to the small amount of silica.The magnetic properties of Fe2O3@SiO2–SO3H nanofibers were characterized by a vibrating sample magnetometer (VSM) (Figure 7 ). The magnetic saturation was about 13.4 emu g−1 that is a suitable amount for nanostructure as a magnetic catalyst. The amount of the magnetic saturation of Fe2O3@SiO2–SO3H nanofibers was very close to the value of Fe2O3@SiO2 nanofibers (14 emu) [46], which can indicate that functionalization with SO3H had no effects on it. Also, the hysteresis loop was not observed confirming that the catalyst is a super paramagnetic material.The number of acids functionalized on the surface of core-shell nanofibers was detected using the reverse titration method. It was observed that 0.4 mmol−1 g of acid existed in the Fe2O3@SiO2–SO3H nanofibers that could be enough for catalytic application. The data can be used to specify the meaning of mol% H+ in catalysts. Mole percent H+ is the percentage of the moles of H+ in proportion to the total number of moles in a mixture. If we use 1mmol of reagent, it is about 0.01 mmol H+ that equals 25 mg Fe2O3@SiO2–SO3H nanofibers according to the calculation of the number of acids in the surface of Fe2O3@SiO2–SO3H nanofibers.Although sulfuric acid acts industrially as an effective catalyst, its aqueous environment can interfere with many reactions because many reactants are sensitive to water that can accelerate the rearrangement and production of by-products, reducing the efficiency of the desired reaction. In addition, it can lead to high corrosion in industrial reaction devices. Using catalysts situated on the surface of a mineral solid, such as silica, can be very effective in avoiding the presence of water in the catalyst as well as raising its specific surface area. Core-shell magnetic silica with sulfuric acid as an insoluble, non-corrosive catalyst with high acidity brings ease of working and separation at the end of the reaction by the magnet; it also makes it possible to use it again in a similar reaction without reducing the catalytic power.Formamidines can be prepared in several ways using different starting materials [15] in most of which toxic solvents are used. High temperature, long reaction time, severe acidity conditions, low yield, hard separation and using additional amounts of reactant and catalyst are the problems of these reactions. Triethyl orthoformate is one of the good starting materials that reacts with amines to produce formamidine. In this study, the reaction was done for the first time in acetic acid under reflux at 140–150 °C for 1.5–94 h [23]. The use of liquid acid, high temperature and long reaction time are the three disadvantages of this reaction. Due to the importance of formamidines, the catalytic ability of Fe2O3@SiO2–SO3H nanofibers in the preparation of formamidines was studied in mild condition and without solvent. For this purpose, the reaction between triethyl orthoformate and aniline was selected as the model reaction (Figure 8 ).First, the reaction progress of triethyl orthoformate and aniline at room temperature was followed by thin-layer chromatography in the presence of Fe2O3@SiO2–SO3H nanofibers. After three hours, the product achieved a yield of 86%. Proper efficiency at low temperature indicates the catalytic activity of Fe2O3@SiO2–SO3H nanofibers in the direct reaction of formamidine preparation. Next, the effect of an increase in the temperature on the model reaction was investigated (Table 1 ). The increase in temperature could increase the speed and efficiency of the reaction. The addition of 1 mol% of catalyst at 70 °C synthesized 92 % formamidine during 1 h (Table 1, entry 3). The reaction was remarkably fast. Next, the effect of catalyst amount on the reaction efficiency was investigated. By increasing the amount of catalyst to 3 mol%, the reaction efficiency increased to 96% (Table 1, entry 4). Without a catalyst, no product was prepared at room temperature. At 70 °C, a crystalline amine was formed, but formamidine was not obtained meaning that the reaction needed the catalyst (Table 1, entry 5).Moreover, the catalytic effect of iron oxide nanofiber with silica coating (Fe2O3@SiO2) was investigated in this reaction (Table 1, entry 6). By adding the same values of this nanofiber to the reaction and performing it in optimal conditions, it was observed that after four hours, the product was obtained with a yield of 27%, indicating the effectivity of functional group on reaction efficiency, which increased by increasing the acidity of the catalyst surface.The comparison of the catalytic activity of some catalyst in formamidine synthesis was shown in Table 2 . Sheykhan et al. used Fe2O3@SiO2-HA as a magnetic solid acid catalyst to synthesize formamidine (Table 2, entry 3,4) [15]. As observed in Table 2, the yield of reaction was low in the presence of Fe2O3@SiO2-HCLO4 nanoparticles (entry 3); however, better results were achieved with Fe2O3@SiO2–SO3H nanofibers (entry 1). When Fe2O3@SiO2–HBF4 nanoparticles (entry 4) were used as catalysts, the yield of reaction exceeded, but the reaction time was still more than Fe2O3@ SiO2–SO3H nanofibers. It seems that catalysts that are functionalized with SO3H, such as sulfonated rice husk (RH-SO3H), had very good results (Table 2, entry 5) but with a difficult catalyst separation process in spite of magnetic catalyst [84]. Archibald et al. synthesized formamidine complex in reflux conditions and low yield of product with acetic acid as a liquid acid catalyst (Table 2, entry 6) [85]. The reaction with cyclodextrin (CD) as a catalyst (Table 2, entry 7) [86] was done at more time (14 h). It was revealed that Fe2O3@SiO2–SO3H nanofibers as a novel magnetic catalyst performed this reaction in a shorter time and with higher efficiency. Also, all the above-mentioned studies used more amount of reactant (2 mmol amine) and catalyst (0.05 mmol Fe2O3@SiO2–HBF4 nanoparticles), but the new procedure of this study included using 1mmol of amine and 0.03 mmol of Fe2O3@SiO2–SO3H nanofibers.Finally, the optimum conditions were extended to all types of aniline derivatives. The results of the study revealed that in all cases, high-yield products were obtained and the procedure was applicable in the synthesis of a wide range of formamidine derivatives (Figure 9 , Table 3 ). The reaction in the presence of aniline derivatives was carried out conveniently with the electron donation, withdrawing group and more than one substitution. This method proves to be effective with p-bromoaniline, p-fluoroaniline, 2,4-dichloroaniline and p-nitroaniline as derivatives of aniline with the electron-withdrawing group on the aromatic ring (Table 3, entry 5, 6, 7, 8). As shown in Table 3, the withdrawing electrons groups led to a decrease in efficiency. The yield of reaction with p-bromoaniline was 80% in 3h while the same product achieved in70% in 4h in the presence of Fe2O3@SiO2–HBF4 nanoparticles [15]. In contrast to Fe2O3@SiO2–HBF4 [15] nanoparticles and RH-SO3H [84], the novel magnetic nanofibers were successful at the synthesis of electron deficient rings such as 2,4-dichloroaniline and p-nitroaniline (Table 3, entry 7,8).To test the possibility of inter-molecular reaction, the reaction of orthophenylenediamine with triethyl orthoformate was done and led to produce benzimidazole in a very high yield (Table 3, entry 9). Among different derivatives of formamidine, benzimidazole had the highest yield due to the high speed of intermolecular reaction. Aiming to prepare unsymmetrical diaryl orthoformamidines, we decided to do a reaction using a mixture of 50:50 mol/mol of aniline and p-toluidine with triethyl orthoformate under the same reaction conditions. The result was a product with 84–86 °C meting point and three new stain in TLC (Table 3, entry 10). It seems that the product is a mixture of unsymmetrical formamidine and two corresponding symmetrical formamidines which is consistent with Robert's finding in the same reaction in the presence of acid [87]. We also tried to synthesize aliphatic formamidine using this procedure. For this purpose, the reaction of t-butyl amine and triethyl orthoformate under the same reaction conditions was examined (Table 3, entry 11). It has been observed in the literature that the formamidines with an aliphatic residue can quickly decompose to the corresponding amines when exposed to silica, and they are less stable than aryl formamidines [88].The magnetic recycling of catalysts was also investigated. Thus, the reaction between triethyl orthoformate and aniline was studied in the presence of 3 mol% of Fe2O3@SiO2–SO3H nanofibers at 70 °C during 1h reaction time. After the completion of the reaction, an external magnet was employed to separate the mixture, and recovered Fe2O3@SiO2–SO3H nanofibers were reused in a subsequent reaction without any significant decrease in activity even after five runs. The isolated product yield decreased slowly from 96% to 92% after five cycles (Figure 10 ). In conclusion, we can introduce Fe2O3@SiO2–SO3H nanofibers as a new and powerful recyclable magnetic catalyst for the conversion of amines to formamidine derivatives under solvent-free conditions. The notable advantages of this catalyst including industrial-scale production, reusability, low-cost separation process of the catalyst by using a magnet and the operational simplicity of the method and its solvent-free condition can present this catalyst as an important alternative to previously reported methods.The methods previously used for amine N-formylation [53, 54, 55, 56, 57, 58] have some disadvantages such as high temperature, long reaction time, additional amounts of reactants, the use of reactors, toxic and harmful solvents for the environment, low efficiency and use of expensive and harmful compounds. Because of the importance of the N-formylation reaction and the problems mentioned, the N-formylation reaction of amines was performed in this study using Fe2O3@SiO2–SO3H nanofiber catalyst and N-formyl products were obtained under room temperature conditions and in short time with high yields.A variety of amines were used to investigate the magnetic Fe2O3@SiO2–SO3H nanofibers catalysis in the formylation reaction (Figure 11 ) and the results are summarized in Table 4 . The formylation of different types of amines, including aliphatic, acyclic, aromatic and heterocyclic compounds, were considered in this study. As shown in Table 4, aromatic amines such as aniline and p-toluidine reacted in excellent yields to produce the corresponding N-formyl compounds (entries, 1, 2). The most effective result was gained with the amine/formic acid molar ratio of 1–1.2 with 0.01 g of the catalyst at room temperature in solvent-free conditions (Table 4, entry 1,3). The same result can be found in the synthesis of formamide from the N-formylation of amines in the presence of Natrolite zeolite as catalyst [60]. However, more catalyst demand (0.02 g) and hard separation process are the disadvantages of the Natrolite zeolite catalyst. The sterically-hindered amines (entry 6) and poorly reactive ones such as p-nitroaniline (entry 4) were easily formylated providing corresponding formamides in good yields. The conversion of p-nitroaniline to the corresponding amides was carried out in 94% with Fe2O3@SiO2–SO3H nanofibers as a catalyst while the yield of reaction was 83% and 90% in the presence of the Natrolite zeolite [60] and RH-SO3H [84] catalysts, respectively. In addition, Nishikawa et al. were not able to do the N-formylation of p-nitroaniline probably due to the low acidity of TEOS asa catalyst [65]. It is worth mentioning that the main problem of TEOS is the hard recovery of the liquid catalyst. The method is effective for the formylation of aliphatic amines (entries 6, 8) with high yield. This method can be applied to convert orthophyllen-diamine to benzimidazole in high yields (entry 9) as an inter-molecular reaction. Imidazole as a heterocyclic amine reacted with formic acid with a 94% yield (entry 10). The chemo selectivity of reaction can be seen in the N-formylation of amine functional group in phenyl hydroxyl amine without O-formylation (Entry 11).The reaction of aniline and formic acid in the presence of Fe2O3@SiO2–SO3H nanofibers was used to study the possibility of the magnetic recycling of catalysts. The catalyst was easily separated from the product using an external magnet due to the magnetic properties of the nanofibers. The separated catalysts were applied in the sample reaction over four cycles in high yields.In conclusion, Fe2O3@SiO2–OSO3H nanofibers were prepared in the following steps: electrospinning, calcinations, coating with silica layer and functionalization with chlorosulfonic acid. The modified, novel, one-dimensional nanostructure was characterized by SEM, TEM, EDS, VSM and FT-IR approving the formation of a core-shell nanofibrous structure. The results revealed that magnetic core-shell nanofibers modified with acid can act as a novel catalyst for the synthesis of formamide and formamidine derivatives with high efficiency. This reaction normally requires high temperature, high acid amount, long reaction conditions and often a difficult catalyst separation process whereas in the method used in this study, most of the above-mentioned disadvantages were removed. The catalyst can be separated by magnet and reused several times in organic reactions without the significant reduction of the yield.Hakimeh Ziyadi: Conceived and designed the experiments; Analyzed and interpreted the data; Wrote the paper.Mitra Baghali: Performed the experiments; Wrote the paper.Akbar Heydari: Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data.This work was supported by Islamic Azad University, Medical Sciences Tehran, Iran (IAUMST).Data included in article/supplementary material/referenced in article.The authors declare no conflict of interest.No additional information is available for this paper.The authors would like to thank the Active Pharmaceutical Ingredients Research Center (APIRC) and Chemistry Department of Tarbiat Modares University for equipment and laboratory services.

Over the past several decades, the fabrication of novel ceramic nanofibers applicable in different areas has been a frequent focus of scientists around the world. Aiming to introduce novel ceramic core-shell nanofibers as a magnetic solid acid catalyst, Fe2O3@SiO2–SO3H magnetic nanofibers were prepared in this study using a modification of Fe2O3@SiO2 core-shell nanofibers with chlorosulfonic acid to increase the acidic properties of these ceramic nanofibers. The products were characterized by scanning electron microscope (SEM), transmission electron microscope (TEM), energy dispersive X-ray spectroscope (EDS), vibrating sample magnetometer (VSM), X-ray diffraction (XRD) and Fourier transform infrared spectroscopy (FT-IR). The prepared nanofibers were used as catalysts in formamide and formamidine synthesis. The treatment of aqueous formic acid using diverse amines with a catalytic amount of Fe2O3@SiO2–SO3H nanofibers as a reusable, magnetic and heterogeneous catalyst produced high yields of corresponding formamides at room temperature. Likewise, the reaction of diverse amines with triethyl orthoformate led to the synthesis of formamidine derivatives in excellent yields using this novel catalyst. The catalytic system was able to be recovered and reused at least five times without any catalytic activity loss. Thus, novel core-shell nanofibers can act as efficient solid acid catalysts in different organic reactions capable of being reused several times due to their easy separation by applying magnet.

Data will be made available on request.Lignocellulose, the natural composite of cellulose, hemicellulose and lignin, is the most abundant form of biomass [1]. Different from cellulose and hemicellulose, lignin is composed of aromatic units, such as sinapyl (S), coniferyl (G) and p-coumaryl (H) alcohols as well as ferulic (FA) and p-coumaric (pCA) acids [1]. Nowadays, cellulose and hemicellulose are fully utilized in pulping and the second-generation (2G) bioethanol industries. The 2G biofuel industry produces enzymatic hydrolysis lignin (EHL) as a low-value and large volume byproduct. As a renewable resource of aromatic molecules, EHL is an ideal feedstock for the sustainable production of commercial aromatic chemicals and fuels.Catalytic lignin solvolysis (CLS) is a promising route to directly produce aromatic chemicals at relatively mild reaction conditions [1]. In CLS reaction, lignin is firstly dissolved and depolymerized in a solvent before contact with a catalyst [2]. However, due to its complex and highly cross-linked structure, lignin cannot be efficiently dissolved in most of the solvents at ambient temperature, and a high reaction temperature (∼300 °C) is often needed for CLS to achieve complete lignin conversion. Nevertheless, high reaction temperature also promotes the repolymerization/condensation of active monomers and self-conversion of solvent, resulting in the formation of char and complex products [3–7]. Yan and his co-workers have examined CLS at relatively low temperature (100 ∼ 200 °C) in water with noble metal catalysts, but, due to the low lignin solubility in water, the monomer yields are only around 7 ∼ 8 wt% [8,9].Ethylene glycol is a green solvent that can be produced from the conversion of cellulose and has been widely used for the production of chemicals and fuels [10]. Recently, ethylene glycol was used as a solvent for lignin depolymerization and fractionation, due to its high lignin solubility. For example, Song et al. [11] depolymerized lignosulfonate with a Ni/C catalyst in ethylene glycol at 200 °C under 5 MPa H2, and achieved 68 wt% conversion of lignosulfonate, with 4-propyl guaiacol and 4-ethyl guaiacol as the main monomer products. Ren et al. [12] used ethylene glycol as a solvent for fractionation of lignin in poplar sawdust with Ru/C and H2SO4 as co-catalysts, and obtained 24.1 wt% phenolic monomers at 185 °C for 6 h. Nevertheless, the mechanism of lignin dissolution in ethylene glycol is still not clear and the steps of lignin depolymerization at low temperatures also need to be further elucidated.Herein, EHL dissolution at room temperature and solvolysis at 200 °C in different solvents are examined. Ethylene glycol achieves complete EHL dissolution and gives the highest monomer yield among the solvents examined. The interaction between ethylene glycol and lignin molecules is investigated with 1H and 13C NMR and Gaussian simulation. The roles of Ni and NaOH catalysts in EHL solvolysis are discussed based on the GPC and HSQC-NMR results as well as molecular dynamics simulation. Based on these results, the mechanisms of EHL dissolution and solvolysis in ethylene glycol are proposed.The EHL was provided by Shandong Long Live biological technology Co., Ltd. which was obtained from the microbial enzymatic hydrolysis of corncob to produce ethanol. The composition of EHL, 91.2 wt% lignin, 0.12 wt% residual carbohydrate and 1.42 wt% ash, has been reported in our previous work [13]. The solvents (AR), including cyclohexane, ethyl acetate, isopropanol, ethanol, methanol and ethylene glycol, were purchased from VWR Chemicals. NiCl2·6H2O (>99.9%), NaOH (>99.9%) and NaBH4 (>98%) MgO (>99.9%), ZrO2 (>99.9%) and Al2O3 (>99.9%) were purchased from Sigma Aldrich. Ferulic acid (>99.9%) and coniferyl alcohol (>98%) were also purchased from Sigma Aldrich. Anisole (99%) was supplied by Acros Organics.The Ni catalyst was prepared via the reduction of NiCl2·6H2O with NaBH4. NaOH (0.5 g) and NaBH4 (1 g) were dissolved in 30 mL deionized water and the solution formed was dropped into the solution of NiCl2 (4.05 g NiCl2·6H2O in 50 mL deionized water) with magnetic stirring at room temperature. The black precipitate, i.e., the Ni catalyst, was washed with 100 mL deionized water for 4 times and preserved in deionized water before use.Solid bases, including NaOH/MgO, NaOH/ZrO2, NaOH/Al2O3, were prepared through incipient wetness impregnation technique with prescribed 20 wt% NaOH loading. After drying at 100 °C overnight, the simples were calcined at 450 °C for 4 h.The mixture of EHL (1 g) and solvent (20 mL) was treated with ultrasound for 30 min at room temperature and then left to stand at room temperature for 72 h. The EHL solution and insoluble EHL were separated with filtration, and the insoluble EHL was dried at 80 °C for 24 h. The amount of dissolved EHL was calculated with Eq. (1). (1) T h e a m o u n t o f d i s s o l v e d E H L % = T h e w e i g h t o f a d d e d E H L - T h e w e i g h t o f i n s o l u b l e E H L T h e w e i g h t o f a d d e d E H L × 100 31P NMR spectra of methanol-soluble (M-soluble) and methanol-insoluble (M-insoluble) EHL were measured according to a method in literature [14]. The sample (40 mg) was dissolved in the mixture of pyridine and deuterated chloroform (1.6:1 v/v, 0.4 mL). Cholesterol (19 mg/ml, 0.2 mL) was added as an internal standard, while chromium-III-acetylacetonate (19 mg/ml, 0.05 mL) was applied as a relaxation reagent. 2-chloro-4,4,5,5-tetramethyl-1,3,2-dioxaphospholane (0.1 mL) was employed as a phosphitylation reagent. After phosphitylation for 2 h, the sample was moved into an NMR tube and measured with a Bruker AVANCE III HD 400 MHz instrument. The heteronuclear single quantum coherence nuclear magnetic resonance (HSQC-NMR) spectra of M-soluble and M-insoluble EHL were recorded with the same instrument. For HSQC-NMR, the sample (50 mg) was dissolved in DMSO‑d 6 (0.6 mL) as the deuterated NMR solvent.The 1H NMR and 13C NMR spectra of EHL, the mixture of EHL and methanol (EHL-MET) and the mixture of EHL and ethylene glycol (EHL-EG) were acquired with the same instrument. 50 and 100 mg EHL were dissolved in 0.6 mL DMSO‑d 6 for 1H NMR and 13C NMR spectra, respectively. For EHL-MET and EHL-EG, methanol (0.1 mL) and ethylene glycol (0.1 mL) were added into EHL and the DMSO‑d 6 mixtures, respectively.EHL solvolysis was carried out in a 50 mL batch reactor (Parr 4597, Hastelloy C-276) equipped with a temperature controller and a pressure sensor. In a typical test, EHL (1 g), Ni (1 g), NaOH (0.5 g) and ethylene glycol (25 mL) were added into the reactor. The reactor was sealed and purged with nitrogen for six times, and then purged with hydrogen for three times, and finally pressurized to 3 MPa H2 at room temperature. The reactor was then heated to 200 °C and remained for 6 h with a fixed stirring rate of 600 rpm.After reaction, the mixture was filtrated to separate the solid catalyst and liquid product. NaOH in the liquid product is neutralized with HCl, and then the liquid product was extracted with deionized water (60 mL) and dichloromethane (30 mL). Floccule formed at the interface of two phases during extraction and was separated with a filtration technique. The monomer products were extracted into the dichloromethane phase and were analyzed qualitatively with Shimadzu GC–MS (QP2010SE with Optic 4) and quantitatively with an Agilent 7890 GC equipped with an FID. For both GCs, the working conditions were the same. The oven temperature program was set from 45 to 250 °C at 10 °C/min and then held at 250 °C for 7 min. The solvent delay was set as 2 min for the MS detector. HP-5 MS capillary column (30 m × 0.25 mm × 0.25 µm) and a split ratio of 50 were used. The mass detector was set to scan the m/z range from 10 to 500. Anisole was used as an internal standard to quantify the products. The total monomer yield was calculated with Eq. (2): (2) T h e t o t a l m o n o m e r y i e l d = T h e w e i g h t o f t o t a l m o n o m e r s t h e w e i g h t o f E H L a d d e d i n t o t h e r e a c t i o n The average molecular weight of the EHL and floccule were determined with an Agilent HPLC system with Phenogel (5 μm–5 nm and 100 nm) columns and a UV detector at 280 nm·THF was used as an eluent at a rate of 1.0 mL min−1 and the analysis was carried out at room temperature. Calibration was performed using polystyrene standards ranging from 30300 g mol−1 to 208 g mol−1. The EHL and floccule samples were acetylated before analysis to make them soluble in THF [15]. The HSQC-NMR spectra of EHL and floccule were recorded with the same Bruker instrument. The sample (50 mg) was dissolved in DMSO‑d 6 (0.6 mL) as the deuterated NMR solvent.The infrared spectra of liquid ethylene glycol as well as adsorbed ethylene glycol and ethylene glycol-lignin monomers mixture were obtained with attenuated total reflectance-Fourier transform infrared spectrometer (ATR-FTIR, PerkinElmer Co.) The scan number was 200 and the spectral resolution was set as 4 cm−1. Adsorbed samples were prepared through heating Ni catalyst (0.5 g) in pure ethylene glycol (10 mL) or ethylene glycol-lignin monomer (0.1 g ferulic acid or coniferyl alcohol in 10 mL ethylene glycol) at 200 °C for 1 h. After cooling, these samples were washed with acetone and dried at 60 °C.The Gaussian simulation was carried out with the Gaussian16 package using M062X simulation method in conjunction with the 6–31 g(d) basis set [16]. The interaction energies between solvent and phenol or benzene are calculated according to Eq. (3), where H is the enthalpy and BSSE is the Basis Set Superposition Error acronym. (3) H int = H A + B - H A + H B + E BSSE The non-covalent interaction (NCI) analysis is carried out with software Multiwfn and VMD [17]. Before NCI analysis, structures are firstly optimized with Gaussian simulation.The Forcite module in Material Studio was used for molecular dynamics simulation of the competitive adsorption of lignin dimer (1-(4-hydroxyphenyl)-2-phenoxypropane-1,3-diol (Fig. 4 (b)) and ethylene glycol over Ni surface. The Ni (111) facet with a (10 × 14 × 4) supercell was chosen as the Ni surface model, above which are 400 ethylene glycol and 2 lignin dimer molecules built with amorphous cell module in Material Studio. Before the simulation, the unit cell and atomic position of the model are firstly optimized. In order to quickly obtained the optimized model, it was quenched for 5 cycles from 26.85 to 226.85 °C, and then underwent an isobaric (NPT) molecular dynamics simulation for one picosecond (ps). After that, this model underwent NPT simulation at 200 °C for 500 ps, and then underwent isothermal molecular dynamics (NVT) simulation at 200 °C for 1000 ps. In all of these simulations, Nose and Berendsen were used as temperature and pressure control algorithm, respectively.The amount of EHL dissolved in 20 mL solvent was examined at room temperature, and the results are shown in Fig. 1 (a). In cyclohexane, ethyl acetate, isopropanol, ethanol and methanol, the amount of dissolved EHL shows a positive linear relationship with the solvent polarity (δH). Nevertheless, this relationship does not fit EHL dissolution in ethylene glycol. The solvent polarity of ethylene glycol is close to that of methanol, but ethylene glycol achieves the complete dissolution of EHL, while methanol only dissolves 47.9% EHL. When the EHL-ethylene glycol solution is diluted with 60 mL ethanol, no EHL is precipitated. Increasing the dosage of EHL to 3 g with keeping the volume of ethylene glycol unchanged (20 mL), the residue remaining on filter paper is not the original EHL solid particles, but a viscous liquid (Fig. S1).The methanol insoluble EHL cannot be dissolved with fresh methanol, indicating that the soluble and insoluble parts have different structures. Therefore, the contents of hydroxyls in M-soluble and M-insoluble EHL were determined with 31P NMR, and the results are depicted in Fig. 1 (b). The contents of aliphatic-OH, aromatic-OH and carboxylic-OH in M-soluble EHL are 1.00, 1.37 and 0.68, respectively, much higher than that of M-insoluble EHL, which are 0.65, 0.74 and 0.35, respectively. The linkages in M-soluble and M-insoluble EHL were determined with HSQC-NMR, and the spectra are illustrated in Fig. 1 (c). In the spectrum of M-insoluble EHL, the intensity of the signal of β-O-4 linked structures (Aγ) obviously decreases, and more intense signals of C-C linkages are detected, compared to those in the spectrum of M-soluble EHL, indicating that M-insoluble EHL has more C-C linkages and less β-O-4 linkages than M-soluble EHL.Non-catalytic EHL solvolysis reactions in different solvents were examined at 200 °C under 3 MPa H2 for 6 h. As depicted in Fig. 2 (a), the monomer yields obtained show a positive correlation with the amount of dissolved EHL in these solvents, and ethylene glycol gives the highest total monomer yield, i.e., 5.5 wt%. After that, catalytic EHL solvolysis reactions in ethylene glycol were examined with Ni catalyst, with NaOH, and with both Ni and NaOH as co-catalysts under the same reaction conditions. The Ni catalyst was prepared via the reduction of NiCl2 with NaBH4, which is a classic catalyst that has been widely employed in many hydrogenation reactions (its characterizations are shown in Fig. S2) [18,19]. Fig. 2 (b) and (c) show the total monomer yields and monomer structures obtained, and Scheme S1 gives the yields of individual monomers obtained. With Ni catalyst, the total monomer yield is 8.2 wt%, lower than that obtained with NaOH, which is 14.6 wt%. When Ni and NaOH catalysts co-existed, the total monomer yield is up to 18.8 wt%. Without a catalyst, most of the monomers obtained have carbon–carbon double bonds in their side chains, and para-alkyl phenols (para-ethyl phenol, para-ethyl guaiacol and para-propyl guaiacol) and phenols without para side chains (phenol, guaiacol and syringol) are also detected. With Ni catalyst, carbon–carbon double bonds are hydrogenated, and para-propanol syringol appears. With NaOH as a catalyst, C2-ketone and C3-ketone substituted syringol appear, which are typical products formed in soluble-base catalyzed lignin conversion reactions [20]. When Ni and NaOH catalysts co-existed, the product distribution is similar to that obtained with only NaOH.The effects of NaOH, with keeping 1 g Ni catalyst unchanged, and Ni, at 0.5 g dosage of NaOH, amounts on the total monomer yield are plotted in Fig. 2 (d). The total monomer yield is only 12.2 wt% when 0.25 g NaOH is added and increases to 18.8 wt% with 0.5 g NaOH. Further increasing the amount of NaOH results in the decrease of the total monomer yield. With the increase of the amount of Ni catalyst from 0.25 to 0.75 g, the total monomer yield increases from 16.4 to 18.9 wt%, and is not obviously changed when the Ni catalyst amount further increases to 1 g.The recyclability of the Ni catalyst is shown in Fig. S3 (a). The Ni catalyst (1 g) was separated from the liquid products with filtration and then directly reused with fresh NaOH (0.5 g). During 4 runs of the Ni catalyst, the total monomer yield slightly decreases from 18.8 to 17.0 wt%. Nevertheless, the XRD pattern of the used Ni catalyst indicates that the Ni catalyst has transformed from an amorphous phase (Fig. S2(a)) to a crystalline phase (Fig. S3 (b)) after the first time run. Hence, the phase transition of the Ni catalyst does not obviously affect its activity on EHL solvolysis.The activities of solid bases, including MgO and NaOH supported on different metal oxides (NaOH/MgO, NaOH/ZrO2, and NaOH/Al2O3), are also examined (Fig. 2 (e)). When 0.5 g solid bases are added with 1 g Ni as co-catalysts, the total monomer yields obtained are around 10 wt%, much lower than that obtained with 1 g Ni and 0.5 g NaOH as co-catalysts, indicating that all the solid bases examined show much lower activities than NaOH. Increasing the amount of MgO and NaOH/MgO from 0.5 g to 1 g with keeping 1 g Ni catalyst unchanged, total monomer yields are not obviously improved, slightly increasing from 9.1 to 10.7 wt% and from 11.2 to 12.4 wt%, respectively. Hence, these solid bases cannot efficiently catalyze EHL depolymerization at a low reaction temperature (200 °C).Although EHL was completely dissolved in ethylene glycol before and after the reaction with or without a catalyst, a floccule appeared between the water and CH2Cl2 phases during product extraction. The floccule is composed of lignin fragments that cannot be dissolved in water and CH2Cl2. The weight average molecular weights (Mw) and β-O-4 linkage contents of EHL and the floccule samples were analyzed with GPC (Fig. S4) and HSQC-NMR (Fig. S5), respectively, and the results are summarized in Fig. 2 (f). The Mw of EHL is 4333 g/mol. The Mw of the floccule obtained without a catalyst is much lower than that of EHL, which is 2216 g/mol. When a catalyst is added, the Mw decreases in an order: Ni (1768 g/mol) > NaOH (642 g/mol) > both Ni and NaOH (464 g/mol). The intensities of the peaks of β-O-4 linkages in HSQC-NMR spectra were normalized with the peak of DMSO. For EHL, the relative intensity of β-O-4 linkage signal is 0.22. These values of floccule obtained without catalyst and with Ni are similar, which are 0.15 and 0.13 respectively. Nevertheless, this value of floccule obtained with NaOH is only 0.04. When Ni catalyst and NaOH co-exist, the signal of β-O-4 linkage disappears and this value turns to 0.EHL, EHL-MET and EHL-EG samples were analyzed with 1H NMR and 13C NMR to reveal the interactions between EHL and solvent, and signal assignment is based on the published works [21–27]. Fig. 3 (a) plots the 1H NMR spectra obtained. In the spectra of EHL-MET and EHL-EG, the peaks of H in phenolic hydroxyls (10–8 ppm) shift to higher field (lower δ value), compared to those in the spectrum of EHL. This is due to the cleavage of original intramolecular hydrogen bonds in lignin and the formation of new hydrogen bonds between solvent and phenolic hydroxyls [28]. Nevertheless, in the spectra of EHL-MET and EHL-EG, the positions of the peaks of H in phenolic hydroxyls are similar, indicating that the phenolic O-H⋯O hydrogen bonds in EHL-MET and EHL-EG have similar strengths. In addition, the strong peaks of H in aromatic ring (7.5–6.3 ppm) and aliphatic chain –CH3/–CH2 (1.4–0.6 ppm) also shift to higher field in the spectrum of EHL-MET, and further shift to the same direction in the spectrum of EHL-EG, compared to those in the spectrum of EHL. In the 13C NMR spectra (Fig. 3(b)), the peaks ascribed to C4 in G unit (G4, 145.7 ppm), C2/C6 in pCA and H units (pCA2/6 and H2/6, 130.5 ppm), C3/C5 in pCA and H units (pCA3/5 and H3/5, 116.0 ppm) as well as C in CH2 in the aliphatic side chain (29.6 ppm) shift to lower field (higher δ value) in the spectrum of EHL-MET, and further shift to the same direction in the spectrum of EHL-EG, compared to those in the spectrum of EHL. The shifting of these 1H and 13C NMR peaks in the spectra of EHL-MET and EHL-EG suggests the existence of interaction between aromatic and aliphatic C-H in EHL and O in the solvent.Gaussian simulation was employed to verify the existence and strength of phenolic O-H⋯O and aromatic C-H⋯O interactions in EHL-MET and EHL-EG. Phenol and benzene were used to represent the phenolic OH and benzene ring in EHL, respectively, to exclude the influence of other functional groups. Fig. 4 (a) illustrates the stable structures and interaction energies of phenol-methanol (P-MET), phenol-ethylene glycol (P-EG), benzene-methanol (B-MET) and benzene-ethylene glycol (B-EG) complexes. For P-MET and P-EG, the interaction energies are similar, i.e., −0.44 eV and −0.59 eV, respectively. Hoverer, the interaction energy of B-EG is −0.36 eV, much lower than that of B-MET, which is only −0.06 eV. Therefore, the strengths of phenolic O-H⋯O interaction formed in P-MET and P-EG are similar, but the aromatic C-H⋯O interaction in B-EG is much stronger than that in B-MET.The strength of O-H⋯O hydrogen bonds between aliphatic OH in a lignin dimer (1-(4-hydroxyphenyl)-2-phenoxypropane-1,3-diol) and OH in ethylene glycol and methanol are also calculated. As shown in Fig. 4 (b), the interaction energies of Cγ-OH⋯O and Cα-OH⋯O in lignin dimer-ethylene glycol (D-EG) are −0.55 and −0.69 eV, respectively, slightly lower than these value in lignin dimer-methanol (D-MET), which are −0.37 and −0.41 eV, respectively. This indicates that the aliphatic O-H⋯O hydrogen bonds in D-EG are slightly stronger than those in D-MET.The NCI of this lignin dimer, lignin dimer-ethylene glycol (D-EG) and lignin dimer-methanol (D-MET) are further analyzed, and the results are shown in Fig. 4 (c). Blue and green clouds indicate the hydrogen bond and van der Waals interactions, respectively. In the lignin dimer, the van der Waals interaction between the two benzene rings in lignin dimer is ascribed to π-π stacking interaction, which results in the overlapping of two benzene rings [29]. In D-EG, overlapping benzene rings are opened due to the van der Waals interaction between the lignin dimer and ethylene glycol, which are ascribed to the lone pair⋯π interaction between the lone pair of O in ethylene glycol and π electrons in benzene rings [30]. Nevertheless, in D-MET, methanol prefers to form a hydrogen bond with phenolic hydroxyl in lignin dimer, and the benzene rings are still stacked.The interaction between ethylene glycol and a lignin model molecule (Fig. 5 (a)) consisting of five benzene rings and C-O and C-C linkages, i.e., α-O-4, β-O-4, 5-O-4, and β-1, was further examined. As shown in Fig. 5 (b) and (c), without ethylene glycol molecule, the lignin model molecule is aggregated due to the intramolecular hydrogen bond and π-π stacking interactions. When seven ethylene glycol molecules are added, the aggregated lignin model molecule is stretched (Fig. 5 (d)), and both hydrogen bond interaction and van der Waals interactions, including C-H⋯O and lone pair⋯π, form between ethylene glycol and the lignin model molecule (Fig. 5 (e)).The adsorption of ethylene glycol and lignin monomers, i.e., ferulic acid (FA) and coniferyl alcohol (CA), over Ni catalyst at 200 °C was examined with ATR-FTIR, and the spectra are shown in Fig. 6 . In the spectrum of liquid ethylene glycol, the broad band at 3290 cm−1 is ascribed to the stretching vibration of –OH, and the bands at 2936 and 2869 cm−1 are ascribed to the stretching vibration of –CH2–, whose bending vibrational bands appear in the range of 1455–1200 cm−1, and the bands at 1029 and 866 cm−1 are ascribed to the stretching and bending vibration of –C–O–, respectively. In the spectrum of ethylene glycol adsorbed over Ni catalyst, the bond of –OH shifts to 3664 cm−1 and its intensity is significantly weakened compared to that of liquid ethylene glycol, and the stretching vibrational bonds of –CH2– and –C–O– also shift to higher wavenumber, appeared at 2985, 2894 and 1054 cm−1, respectively. The shift of these bonds results from the transformation of electrons from ethylene glycol to Ni atoms. When the mixtures of 10 mL ethylene glycol and 0.1 g FA or CA are adsorbed over Ni catalyst, the spectra obtained are the same as that of pure ethylene glycol adsorbed over Ni catalyst, indicating that these lignin monomers in ethylene glycol can not be adsorbed over Ni catalyst.The competitive adsorption of the mentioned lignin dimer (1-(4-hydroxyphenyl)-2-phenoxypropane-1,3-diol) and ethylene glycol molecules over the Ni surface was investigated with molecular dynamics simulation. Fig. 7 (a) is the initial state of the model, in which one lignin dimer (Dimer A) is surrounded with ethylene glycol molecules and another one (Dimer B) is over the Ni surface. After 1000 ps simulation (Fig. 7 (b)), Dimer A is still in the liquid phase, while Dimer B remains adsorbed over the Ni surface. The distribution of ethylene glycol molecules along the z-axis (Fig. 7 (c)) indicates that ethylene glycol molecules are enriched over the Ni surface. The trajectory of the mass center of Dimer A and B during 1000 ps (Fig. 7 (d)) shows that Dimer A cannot go through the ethylene glycol molecular layer to adsorb over the Ni surface, while Dimer B steadily adsorbs over the Ni surface.EHL dissolution is the first step of EHL solvolysis. We found that ethylene glycol achieves complete EHL dissolution at room temperature, while other solvents, such as methanol, only dissolve part of EHL with high content of hydroxyls and β-O-4 linkages. Previous articles thought that lignin dissolution is mainly attributed to the O-H…O hydrogen bond between solvent and lignin [28,31–33]. Nevertheless, for lignin molecules with low content of hydroxyls, the main obstacle to its dissolution is the π-π stacking interaction [34–37]. Gaussian simulation indicates that ethylene glycol forms stronger C-H⋯O and lone pair⋯π interactions with benzene rings of EHL than methanol dose, because one ethylene glycol molecule contains two O atoms. These strong van Der Waals forces break original π-π stacking in EHL and achieve complete EHL dissolution.The monomer yields obtained from non-catalytic EHL solvolysis show a positive correlation with the EHL solubility of solvents, and the highest monomer yield is obtained in ethylene glycol. As revealed with the GPC and HSQC-NMR results, linkages in EHL are already partly broken in ethylene glycol even at 200 °C without a catalyst, forming monomers and lignin fragments. We speculate that the strong van Der Waals forces between ethylene glycol and EHL may result in the shift of electrons in the benzene ring of EHL, reducing the bond energy of β-O-4 linkages in EHL, as shown in Scheme 1 (a).It has been generally accepted that a large lignin molecule cannot be directly adsorbed on the surface of the solid catalyst due to its large three-dimensional structure [1,2,4]. Our results of molecular dynamics simulation further reveal that the adsorption of ethylene glycol hinders the adsorption of lignin dimer from the liquid phase to the surface of the Ni catalyst. Hence, catalytic EHL hydrogenolysis is not the main reaction pathway, and lignin depolymerization mainly occurs through solvolysis reaction.The comparison of the results of blank reaction without catalyst and catalytic reaction with Ni catalyst reveals that the Ni catalyst does play a role in the hydrogenation of carbon–carbon double bonds in the side chains of lignin monomers. Nevertheless, the adsorption of lignin monomers is also hindered by the adsorption of ethylene glycol. Hence, the hydrogenation reaction may occur in the liquid phase. we guess that O atoms in ethylene glycol may attract adsorbed H atoms over the Ni surface due to their strong electronegativity, forming a hydrogen-ethylene glycol complex, which may desorb from the Ni surface and involve in the hydrogenation reaction in the liquid phase (Shame 1(b)). As reported, even at around 200 °C, lignin depolymerization and product repolymerization occur simultaneously in non-catalytic lignin alcoholysis, and intermediates with carbon–carbon double bonds in their side chains more readily undergo repolymerization reactions [38–42]. The Ni catalyst stabilizes active monomers through hydrogenation reactions, suppressing repolymerization reactions. Hence, the addition of the Ni catalyst improves monomer yield and reduces Mw of floccule.As discussed above, Ni catalyst just plays a role in the hydrogenation reaction of carbon–carbon double bonds in EHL solvolysis reaction. The hydrogenation of carbon–carbon double bonds is relatively easy [43,44], and does not require a Ni catalyst with high hydrogenation activity. Hence, the activity of the Ni catalyst is insensitive to the phase transition of Ni from the amorphous phase to the crystalline phase.NaOH is soluble in ethylene glycol and serves as a homogeneous catalyst that directly promotes the breakage of C-O linkages in lignin [45]. Hence, NaOH is more efficient than solid base catalysts for catalyzing lignin depolymerization. When NaOH was used as the catalyst, the content of β-O-4 linkage and Mw of floccule were significantly reduced. When Ni and NaOH were used as co-catalysts, the product distribution obtained is similar to that obtained with only NaOH as a catalyst, indicating that the reaction mainly follows the soluble-base catalyzed route. Nevertheless, NaOH also promotes repolymerization/condensation of active monomers and intermediates, and hence too high NaOH amount results in the decrease of total monomer yield [45,46]. As mentioned above, the Ni catalyst stabilizes active monomers and intermediates, suppressing repolymerization reaction, and hence the combination of Ni and NaOH obtains a higher monomer yield than a single catalyst [47–49].Based on the presented results, the pathways of EHL solvolysis in ethylene glycol at 200 °C with Ni and NaOH as co-catalysts are proposed and presented in Scheme 2 . Agglomerated lignin molecule is firstly dissolved and partly depolymerized in ethylene glycol, exposing more functional groups, e.g., –OCH3 and –OH. NaOH depolymerizes dissolved lignin fragments through attacking these functional groups, forming active monomers and intermediates with carbon–carbon double bonds [45,50]. Active hydrogens are transformed from the Ni surface to the liquid phase with ethylene glycol as a porter, and involve into the hydrogenation of carbon–carbon double bonds. After several cycling of base-catalyzed dehydroxylation and hydrogenation reactions, stable monomers are produced.Ethylene glycol shows high EHL solubility and achieves complete EHL dissolution at room temperature, while methanol only dissolves part of EHL with high content of hydroxyls and β-O-4 linkages. The Gaussian simulation results indicate that ethylene glycol forms strong Van Der Waals interactions with EHL, including C-H⋯O and lone pair⋯π interactions, and these interactions break original π-π stacking in EHL, achieving complete EHL dissolution.The total monomer yields obtained from non-catalytic EHL solvolysis at 200 °C under 3 H2 for 6 h show a positive correlation with EHL solubility, and ethylene glycol gives the highest total monomer yield, i.e., 5.5 wt%, among the solvents examined. With Ni and NaOH as co-catalysts, the total monomer yield in ethylene glycol reaches 18.8 wt% under the same reaction condition.EHL is partly depolymerized at 200 °C without a catalyst due to the strong interactions between EHL and ethylene glycol. NaOH as the homogeneous catalyst directly attracts C-O bonds in EHL and depolymerizes EHL into active monomers and intermediates. The adsorption of lignin fragments over Ni catalyst via C-O bond is hindered with ethylene glycol, and Ni catalyst mainly plays a role in supplying active hydrogen atom to stabilize the active intermediates, suppressing repolymerization side reactions.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This work has received funding from the European Union’s Horizon 2020 research and innovation program, (BUILDING A LOW-CARBON, CLIMATE RESILIENT FUTURE: SECURE, CLEAN AND EFFICIENT EN-ERGY) under Grant Agreement No 101006744. The content presented in this document represents the views of the authors, and the European Commission has no liability in respect of the content. Y.S. Sang and G. Li would like to express their gratitude to both the China Scholarship Council (202006250156, 202208320030) and the EU-101006744 project.Supplementary data to this article can be found online at https://doi.org/10.1016/j.cej.2023.142256.The following are the Supplementary data to this article: Supplementary data 1

The dissolution and solvolysis processes of enzymatic hydrolysis lignin (EHL) in ethylene glycol are investigated. Ethylene glycol exhibits high EHL solubility and achieves complete EHL dissolution at room temperature. Gaussian simulation reveals that van de Waals interactions between ethylene glycol and EHL, including C-H⋯O and lone pair⋯π interactions, break the π-π stacking in EHL, achieving complete EHL dissolution. EHL is partly depolymerized in ethylene glycol at 200 °C even without a catalyst due to the strong van de Waals interactions. When NaOH and Ni are used as co-catalysts, EHL is efficiently depolymerized at 200 °C, and the overall monomer yield reaches 18.8 wt%. Fourier transform infrared spectroscopy (FT-IR) and molecular dynamics simulation results indicate that the adsorption of ethylene glycol over Ni surface hinders the adsorption of lignin fragments and monomers. Hence, EHL catalytic solvolysis in ethylene glycol occurs in the liquid phase, where OH− of NaOH promotes the EHL linkage breakage and active hydrogen atoms formed on Ni surface stabilize the active monomers.

Energy and the environment are among the most important concerns of the current era. However, most of the energy we are using still comes from nonrenewable fossil fuels obtained from reserves that are ultimately unsustainable and that result in environmental pollution. The conversion of energy from renewable energy sources could reduce the dependence on fossil fuels significantly [1–3]. Among these renewable energy methods, electrochemical catalytic water-splitting is considered to be one of the most effective.Efficient electrolysis of water usually requires developing high-performance electrocatalysts with high stability, fast kinetics and low overpotential [4]. Noble metal-based materials such as Pt and Ru-based catalysts are the most widely used catalysts in the hydrogen evolution reaction (HER) and the oxygen evolution reaction (OER), respectively. However, optimum operating conditions and reaction mechanisms are different for HER and OER electrocatalysts. Catalysts that are good for HER do not perform well in OER, and vice versa [5]. Pt and Ru-based catalysts, although they are effective bifunctional electrocatalysts, perform poorly in overall water splitting. Moreover, the expense and scarcity of these precious metals severely limits their large-scale application, and it is not economical to produce single-function electrocatalysts for each HER and OER, as this would increase manufacturing costs [6–10]. The development of efficient bifunctional electrocatalysts for water splitting will greatly reduce the preparation cost and simplify the electrolytic system as well. However, designing and preparing a catalyst that promotes both HER and OER in the same electrolyte remains a major challenge.In recent years, transition metal phosphides (TMPs) [9,11], carbides [12], nitrides [13,14] and sulfides [15,16] have been widely reported as potential electrocatalysts for OER or HER [17]. In particular, TMPs have attracted great interest from researchers due to their outstanding catalytic performance [18]. The negatively charged P atoms in the catalyst are capable of attracting protons and acting as active sites for H2 evolution. In the oxygen evolution reaction, transition metal phosphates are transformed into transition-metal oxyhydroxides on the surface of the catalysts which can act as active catalytical sites for O2 evolution [19]. Additionally, TMPs have been considered as advanced catalysts because of their superior electrical conductivity [20]. These properties make TMPs promising electrochemical catalysts for water splitting [21]. However, many non-noble metal phosphates do not exhibit satisfactory bifunctional properties. For instance, NiCoP catalysts have been reported as promising electrocatalysts of non-noble metals for overall water splitting in recent years [22,23], but they show inferior HER catalytic activity relative to the noble metal-based catalysts in the alkaline condition.For this reason, different strategies have been implemented to improve their bifunctional performance, such as developing single atom catalysts [24], doping other transition-metals and heteroatoms [25] and modulating the structure and composition [26]. It has been proved that the electrocatalytic performance of catalysts is tremendously affected by their morphology, active surface sites and electronic structure [27], which can be tailored by designing their heterostructures. Generally, heterostructured materials demonstrate better electrocatalytic performance than their individual building units because of the benefits from the strong coupling between different building components. It is an efficient approach to introduce a “collaborator” that can form heterostructures to further improve HER and OER activity.CoMoO4 is considered an ideal choice as a “collaborator” to provide active hydrolytic dissociation sites because Co and Mo have hydrogen adsorption energy close to Pt, and their binary metal oxides have higher constitutive activity than monomeric metal oxides, which significantly facilitates the proton supply. Nevertheless, due to the weak intrinsic conductivity of CoMoO4, its application in alkaline HER is limited. The synthesis of oxide catalysts with the incorporation of a P atom to tailor the electronic structure is an efficient strategy to address the problem of low conductivity [28]. Yaqiong Gong et al. [28] utilized phosphorus-doping modulation to fabricate monoclinic P-CoMoO4 with an optimized electron structure supported on nickel foam (P-CoMoO4/NF) for alkaline HER via a facile hydrothermal method, followed by low-temperature phosphidation. Ivan P. Parkin et al. [29] synthesized a series of P-doped CoMoO4 nanostructures on Ni foam by facile hydrothermal annealing and phosphidation modification, which enhanced their electrochemical performance significantly.Herein, we report a novel P-CoMoO4@NiCoP/NF heterostructured nanoarray catalyst as an efficient bifunctional electrocatalyst to promote the overall water splitting. P-CoMoO4@NiCoP/NF nanoarrays have been successfully prepared through the phosphatization of the CoMoO4@NiCo2O4 nanoarray precursor with NaH2PO2·H2O as the P source under heat treatment in the N2 atmosphere. Heterostructured P-CoMoO4@NiCoP rooted on Ni foam possesses a novel tree-like 3D structure, in which P-CoMoO4 nanosheets (leaves) are assembled on the surface of NiCoP nanowires (trunk). Due to the unique heterostructure, more exposed active sites and coordinated electronic structure, the P-CoMoO4@NiCoP/NF presents excellent HER and OER catalytic activity and shows outstanding overall water splitting performance.Analytical grade Ni(CH2COOH)2·6H2O (≥98%), Co(CH2COOH)2·6H2O (≥99%), Na2MoO4·2H2O (≥99%), NH4F (≥96%), NaH2PO2·H2O (98%–103%), KOH (≥85%), urea (≥99%) and ethanol were ordered from the Sinopharm Chemical Reagent Co. Commercial ruthenium dioxide (RuO2, 99.9%) and platinum on activated carbon (Pt/C, 20 ​wt%) were ordered from Aladdin Chemical Reagents Co. Ni foam (denoted as NF) was ordered from Shenzhen Meisen Electromechanical Equipment Co., Ltd. All chemical reagents were used without further purification. The experimental water was deionized water.A piece of Ni foam (2 ​cm ​× ​4 ​cm) was degreased in an acetone solution, ultrasonicated in 3.0 ​M HCl solution for 3-5 ​min, then thoroughly washed with deionized water and ethanol alternately to clean the surface. Urea (10 ​mmol), NH4F (8 ​mmol), Ni(CH2COOH)2·6H2O (0.333 ​mmol) and Co(CH2COOH)2·6H2O (0.667 ​mmol) was dissolved in 36 ​mL of deionized water and stirred continuously to form a clear solution. The pre-treated NF was then transferred to a Teflon lined stainless steel autoclave (50 ​mL) containing the solution and maintained at 120 ​°C for 3 ​h. After natural cooling, the NiCo-based precursor of NF (denoted as NiCo-OH/NF) was taken out, rinsed with deionized water until there was no residue, washed with ethanol three times, and finally dried at 60 ​°C for 12 ​h NiCo2O4 nanowire arrays grown on Ni foam (denoted as NiCo2O4/NF) were obtained after annealing the NiCo-OH/NF sample at 450 ​°C for 2 ​h in air.The prepared NiCo2O4/NF nanoarrays were immersed in a solution containing 36 ​mL of deionized water, 1 ​mmol Na2MoO4·2H2O and 1 ​mmol Co(CH2COOH)2·6H2O. The reaction was carried out at 100 ​°C in a Teflon-lined stainless steel autoclave for 12 ​h. After natural cooling, the sample was taken out, washed with deionized water until there was no residue, washed with ethanol three times, and finally dried at 60 ​°C for 12 ​h to obtain CoMoO4@NiCo2O4/NF.P-CoMoO4@NiCoP/NF composite nanoarrays were obtained through phosphatization with NaH2PO2·H2O as the P source. The prepared CoMoO4@NiCo2O4/NF composite nanoarrays and NaH2PO2·H2O were put separately in a porcelain boat with the NaH2PO2·H2O powder at the upstream side, and then heated in a tube furnace at 300 ​°C (ramp rate of 5 ​°C min−1) for 180 ​min under a N2 atmosphere.As comparison samples, NiCoP/NF and P-CoMoO4/NF nanoarrays based on Ni foam were prepared separately by the same method, as detailed in Supporting Information 1. Fig. 1 illustrates the process of synthesizing P-CoMoO4@NiCoP/NF nanoarrays. Briefly, NiCo-OH/NF (Fig. S1) was synthesized through hydrothermal synthesis, followed by annealing to form NiCo2O4/NF nanowire arrays (Fig. S2). The CoMoO4 nanosheets were then synthesized on the NiCo2O4 nanowires to form CoMoO4@NiCo2O4/NF (Fig. S3). Finally, heterostructured P-CoMoO4@NiCoP/NF nanoarrays were synthesized through phosphatizing the prepared CoMoO4@NiCo2O4/NF sample. Fig. 2 a shows the X-ray diffraction (XRD) pattern of the P-CoMoO4@NiCoP powder sample which was peeled off the Ni foam substrate. The reflections in the XRD pattern could be indexed to NiCoP (PDF No. 71-2336) and CoMoO4 (PDF No. 21-0868), indicating the partial phosphatization of the CoMoO4@NiCo2O4/NF sample. Under the experimental conditions, NiCo2O4 was completely phosphatized into NiCoP, while CoMoO4 did not show an obvious conversion to a phosphatized product. The 2θ values of 41.14° and 45.06° corresponded to the (111) and (201) crystal planes of NiCoP, and it was found that the 2θ values were shifted positively about 0.15° compared with the standard card of NiCoP (PDF No. 71-2336). This was ascribed to the higher amounts of Co atoms and the mixed valence states of the Co ions [6,30]. Meanwhile, the 2θ values corresponding to the (002) and (021) crystal planes of CoMoO4 were 26.66° and 23.54°, exhibiting 0.15° and 0.21° deviations from the standard values of CoMoO4 (PDF No. 21-0868), probably due to the incorporation of P in it. Further element mapping characterization (line scanning of P-CoMoO4@NiCoP in Fig. S4) revealed that the P elements were highly concentrated in the core nanowires, but also with a uniform distribution of relative low content in the CoMoO4 nanosheets, indicating that the P element was partially incorporated into the CoMoO4 matrix [11]. Combining it with the XRD analysis, the phosphatized sample was denoted as P-CoMoO4@NiCoP/NF.The morphological and structural characteristics of the synthesized samples were investigated through field emission scanning electron microscopy (FESEM). First, the FESEM images of NiCoP/NF (Fig. S5) with different magnifications showed that the NiCoP nanowire arrays were uniformly grown on the Ni foam. The length of the NiCoP nanowires was about 1.3 ​μm, and the diameter of the nanowires was ~100 ​nm. Each nanowire was directly in contact with the Ni foam, which can ensure efficient electron transport between the electrocatalyst and the Ni foam substrate, thereby contributing to the water splitting [31]. After a second hydrothermal reaction and phosphatization, the P-CoMoO4 nanosheets were closely and homogeneously covered on the Ni foam (Fig. S6). Fig. 2b shows the FESEM image of the P-CoMoO4@NiCoP/NF sample. It can be found that the P-CoMoO4 nanosheets with a thickness of about 40 ​nm (Fig. S7) were uniformly assembled on the NiCoP nanowire arrays, constructing heterostructured P-CoMoO4@NiCoP/NF nanoarrays. The heterostructured P-CoMoO4@NiCoP anchored on the Ni foam exhibited a novel tree-like 3D structure in which the P-CoMoO4 nanosheets were assembled like leaves on the NiCoP nanowire trunk. The transmission electron microscopy (TEM) image of the P-CoMoO4@NiCoP (Fig. 2c) also attests to the heterostructure, with the P-CoMoO4 nanosheets tightly aggregating around the NiCoP nanowires.The high-resolution TEM (HRTEM) image (Fig. 2d) on the nanosheet (marked in red circle) reveals two clear lattice distances of 0.224 ​nm and 0.187 ​nm, corresponding to the (003) and ( 1 ¯ 33) crystal plane of CoMoO4, respectively. The corresponding selected area electron diffraction (SAED) image (Fig. 2e) shows several bright rings with discrete spots, which match well with the ( 1 ¯ 31), (003), ( 1 ¯ 33) and ( 3 ¯ 52) planes of the CoMoO4. The corresponding elemental mapping image of P-CoMoO4@NiCoP (Fig. 2f) illustrates that the Ni element is only distributed on the nanowires and Mo only on the nanosheets, while the P and Co elements are uniformly distributed on the NiCoP nanowires and P-CoMoO4 nanosheets. This confirms that the P-CoMoO4 nanosheets are uniformly grown on the NiCoP nanowires. These results demonstrate the successful preparation of the heterostructured P-CoMoO4@NiCoP/NF composite.The chemical composition and valence states of the elements were studied by XPS for the heterostructured P-CoMoO4@NiCoP/NF, along with those of bare NiCoP/NF and P-CoMoO4/NF. XPS survey spectra of P-CoMoO4@NiCoP/NF (Fig. S8a) shows that P-CoMoO4@NiCoP/NF is mainly composed of Ni, Co, Mo, P and O elements. The peaks of Ni in P-CoMoO4@NiCoP/NF are weaker than those in NiCoP/NF because there is a dense vegetation of P-CoMoO4 nanosheets wrapped on the surface of the NiCoP nanowires, thus weakening the Ni peak intensity of the P-CoMoO4@NiCoP sample. The XPS spectrum of the Ni 2p in P-CoMoO4/NiCoP displays two peaks at 875.08 and 856.58 ​eV, attributed to Ni2+ 2p1/2 and Ni2+ 2p3/2 (Fig. 3 a) [32]. The peaks at 870.98 ​eV and 852.98 ​eV are assigned to the peaks of P-CoMoO4@NiCoP/NF, related to Ni0 2p1/2 and Ni0 2p3/2, respectively. Compared to those of NiCoP/NF, the Ni2+ peaks of P-CoMoO4@NiCoP/NF are negatively shifted about 0.1 ​eV. Likewise, Fig. 3b demonstrates the Mo 3d spectrum. The appearance of Mo4+ and Mo6+ can be attributed to the CoMoO4 and Mo-P species. In Fig. 3b, the Mo 3d signal in the P-CoMoO4@NiCoP/NF exhibits a positive shift of about 0.3 ​eV compared with that in the P-CoMoO4/NF. Generally, the increase of the valence electron charge will lead to the decrease of binding energy, and vice versa [33]. Therefore, the negative shift of binding energy for Ni 2p and the positive shift of binding energy for Mo 3d strongly demonstrate that there are strong electronic interactions between the NiCoP nanowire and P-CoMoO4 nanosheet, which can significantly accelerate the charge transfer. The Co 2p core-level spectrum (Fig. 3c) exhibits Co2+ species at 797.98 ​eV and 781.78 ​eV. As can be seen, there are no peaks of Coδ+ species in CoMoO4@NiCo2O4/NF, indicating that Coδ+ (793.88 ​eV and 778.88 ​eV) is due to the formation of Co-P bonds (Fig. S8b) [4]. Moreover, the peak intensity of the Coδ+ species in NiCoP/NF and P-CoMoO4/NF is weaker than that in P-CoMoO4@NiCoP/NF, indicating that the formation of the heterogeneous interface facilitates the formation of Coδ+ species, thereby improving the OER or HER performance [19]. In Fig. 3d, the high-resolution P 2p spectra of P-CoMoO4@NiCoP/NF, with the peaks of P-M 2p3/2 and P-M 2p1/2, locate at a binding energy of 129.58 and 130.43 ​eV, respectively. Compared to those of bare NiCoP/NF and P-CoMoO4/NF, the peak intensities of P-Metal (P-M 2p3/2 and P-M 2p1/2) demonstrate that the redistribution of the charge is caused by the strong interaction at the NiCoP and P-CoMoO4 interface in P-CoMoO4@NiCoP/NF, which could significantly enhance the electrocatalytic activity [34]. The peak at 133.9 ​eV represents P-O species oxidized (PO4 3−, etc.) due to air exposure [22].Generally, a large number of air bubbles will be generated on the surface of the electrode during the electrocatalytic reaction under high current density. The gas bubbles produced in-situ, if not released immediately, will seriously hinder the electrolyte diffusion and produce dead areas that cannot participate in the catalytic reaction, thus hindering the mass transfer process and causing the catalytic performance to deteriorate [35].In order to study the hydrophilic and aerophobic properties of the samples, the contact angle of the water droplet and the underwater contact angle of the air bubble were tested. Dynamic testing showed that the water droplet spread immediately on the surface of the P-CoMoO4@NiCoP/NF, as well as the NiCoP/NF and P-CoMoO4/NF, once in contact with them, while the water droplet remained on the surface of the blank Ni foam (see the testing video in Supporting Information 2). As shown in Fig. S9, the P-CoMoO4@NiCoP/NF, as well as NiCoP/NF and P-CoMoO4/NF, displayed a liquid contact angle of 0°, showing its superhydrophilic property, while the blank Ni foam showed its hydrophobic property, with a large static liquid contact angle of 120°. Fig. S10 displays the measurement images of the underwater air bubble contact angles for those samples. The P-CoMoO4@NiCoP/NF demonstrates superaerophobic properties, and the air bubble shows complete estrangement from the surface of the contact film, which is beneficial to the release of the gas bubbles produced during the electrocatalytic water splitting. Similarly, the components comprising NiCoP/NF and P-CoMoO4/NF also exhibit superaerophobic properties, and the air bubbles become estranged from the surface of the film upon cessation of contact, while the air bubble can retain a contact angle of 134° on the surface of the Ni foam, implying the underwater aerophilic property of the bare Ni foam surface. The videos about the underwater air bubble contact angle testing are provided in Supporting Information 3. The superhydrophilic and superaerophobic properties of P-CoMoO4@NiCoP/NF make it conducive to water dissociation and bubble separation in the electrocatalytic process, thus improving the catalytic performance.Supplementary video related to this article can be found at https://doi.org/10.1016/j.nanoms.2021.05.004 The following are the supplementary data related to this article: Video 1 Video 1 Video 2 Video 2 To determine the HER performance, a standard three-electrode cell was used to perform linear sweep voltammetry in a 1.0 ​M KOH electrolyte. The iR-compensated polarization curves of P-CoMoO4@NiCoP/NF, NiCoP/NF, P-CoMoO4/NF, Pt/C/NF and Ni foam at 5 ​mV ​s−1 are displayed in Fig. 4 a, and their corresponding overpotentials at current densities of 10, 50, and 100 ​mA ​cm−2 are presented in Fig. 4b. The P-CoMoO4@NiCoP/NF offers outstanding HER performance with a low overpotential of 66 ​mV at 10 ​mA ​cm−2, which is very close to that of the commercial Pt/C/NF catalyst (45 ​mV). In contrast, the NiCoP/NF and P-CoMoO4/NF exhibit inferior HER performance with high overpotentials of 139 ​mV and 125 ​mV at 10 ​mA ​cm−2, respectively, suggesting significantly enhanced effects of the heterostructure catalyst on the HER performance. Furthermore, the P-CoMoO4@NiCoP/NF delivers a high current density (j ​= ​209 ​mA ​cm−2) at an overpotential of 200 ​mV (Fig. S11), which is about 4 times higher than those of NiCoP/NF (j ​= ​55 ​mA ​cm−2) and P-CoMoO4/NF (j ​= ​56 ​mA ​cm−2), indicating strong synergic effects derived from the P-CoMoO4@NiCoP/NF heterostructured interface [36]. In addition, Tafel plots were investigated to determine the HER rates of these catalysts (Fig. 4c). The P-CoMoO4@NiCoP/NF presents the smallest Tafel slope of 75 ​mV dec−1 among the three catalysts, including NiCoP/NF (135 ​mV dec−1) and P-CoMoO4/NF (103 ​mV dec−1), demonstrating the fastest HER kinetics of the P-CoMoO4@NiCoP/NF catalyst [37]. Compared with some reported catalysts in recent literature, the P-CoMoO4@NiCoP/NF catalyst shows superiority in HER performance (Table S1).Double-layer capacitance (C dl ) reflects the electrochemical surface area of the catalyst. The capacitive current density difference between the anode and the cathode is proportional to the scan rate [15]. Cyclic voltammetry (CV) was used to study the electrochemical surface area (ECSA) at different scan rates (Fig. S12). The C dl of the three catalysts was calculated based on their current density under different scan rates, as shown in Fig. 4d, with 34.79 ​mF ​cm−2, 27.15 ​mF ​cm−2 and 16.26 ​mF ​cm−2 for P-CoMoO4@NiCoP/NF, P-CoMoO4/NF and NiCoP/NF, respectively. The heterostructured P-CoMoO4@NiCoP/NF possesses the largest electrochemically active surface area, which endows it with the best electrocatalysis activity. The excellent performance of the hydrogen evolution is also related to the superhydrophilic and superaerophobic properties of the P-CoMoO4@NiCoP/NF. The superhydrophilicity of the P-CoMoO4@NiCoP/NF can facilitate electrolyte diffusion and make full contact with the electrolyte in the reaction process. In addition, the superaerophobic property is conducive to the release of gas bubbles for exposing more active sites and clearing the pathways for electrolyte diffusion, thus accelerating the hydrogen evolution reaction.Electrochemical impedance spectroscopy (EIS) was investigated in an alkaline medium to further reveal the relevant properties. The fitting circuit consisted of Rct (charge transfer resistance between electrolyte and catalyst interface) in parallel with CPE and then in series with Rs (the intrinsic resistance of the electrode and electrolyte). According to the equivalent circuit diagram shown in Fig. 4e, the Rs of P-CoMoO4@NiCoP/NF (1.286 ​Ω) is smaller than that of NiCoP/NF (1.350 ​Ω) and P-CoMoO4/NF (1.409 ​Ω). Similarly, the Rct of P-CoMoO4@NiCoP/NF (0.353 ​Ω) is also smaller than that of NiCoP/NF (0.406 ​Ω) and P-CoMoO4/NF (0.435 ​Ω). The smaller Rs and Rct of the P-CoMoO4@NiCoP/NF electrode resulted from the tree-like 3D architecture with a heterostructured interface, which reduced the internal resistance and promoted electron transfer, thus improving the electrocatalytic performance [38]. As shown in Fig. S13, the Rs (1.723 ​Ω) and Rct (0.918 ​Ω) of the CoMoO4@NiCo2O4/NF are significantly larger than those of P-CoMoO4@NiCoP/NF, indicating that phosphide and the incorporation of P can effectively increase the charge transfer property of the electrocatalyst.In addition, the P-CoMoO4@NiCoP/NF shows high durability during the 60 ​h chronopotentiometry test at 10 ​mA ​cm−2, and the overpotential displays a negligible increase (Fig. 4f). It basically maintains the original tree-like 3D heterostructure (Fig. S14). This is attributed to the superaerophobic property of the P-CoMoO4@NiCoP/NF. The superaerophobic property of the P-CoMoO4@NiCoP/NF accelerates the release of bubbles, and will not produce a dead zone in the catalytic reaction, maintaining good performance in the stability test [35]. Fig. S15 shows the XPS spectra of the P-CoMoO4@NiCoP/NF before and after the 60 ​h chronopotentiometry test for HER at 10 ​mA ​cm−2 in 1.0 ​M KOH solution. Of note, the XPS spectra of the Ni 2p region does not change, but the Coδ+ peaks disappear. The XRD of the P-CoMoO4@NiCoP/NF after the chronopotentiometry test for HER in 1.0 ​M KOH solution (Fig. S16a) shows the presence of Co(OH)2, confirming the partial change from P-CoMoO4@NiCoP/NF to Co(OH)2 [9]. The high-resolution TEM (HRTEM) image (Fig. S16c) on the nanosheet (marked in red circle in Fig. S16b) reveals a clear lattice distance of 0.237 ​nm, corresponding to the (101) crystal plane of Co(OH)2. The corresponding selected area electron diffraction (SAED) image (Fig. S16d) shows several bright rings with discrete spots, which match well with the (101), (102) and (111) planes of Co(OH)2. Phase transformation may cause the disappearance of Coδ+. And the overall decrease in the peak intensity of P 2p and Mo 3d may be caused by the partial leaching out of P and Mo from the P-CoMoO4@NiCoP sample during the reaction process [39]. The existence of Coδ+ related to the formation of the Co-P bond has been proved in Fig. S8b and it has been reported in some literature [4,40]. The decrease of P-M bonds may be one of the reasons for the disappearance of Coδ+.The OER catalytic activities of these samples were also investigated in 1.0 ​M KOH solution. The LSV curves are shown in Fig. 5 a. To reach a current density of 100 ​mA ​cm−2, the heterostructured P-CoMoO4@NiCoP/NF requires an overpotential of only 252 ​mV, which is much lower than that of NiCoP/NF (287 ​mV) and P-CoMoO4/NF (262 ​mV). To reach even higher current densities of 200 and 300 ​mA ​cm−2, only 292 and 313 ​mV overpotentials are required for the P-CoMoO4@NiCoP/NF catalyst. The capacitance behavior of the non-Faradaic capacitance current range for P-CoMoO4@NiCoP/NF, NiCoP/NF, P-CoMoO4/NF and Ni foam was measured (Fig. S17). The results show that the P-CoMoO4@NiCoP/NF, NiCoP/NF and P-CoMoO4/NF samples exhibit very strong capacitive behavior in comparison with the Ni foam sample. This may be one of the reasons that the polarization curve is raised in the non-Faradaic capacitance current range. In addition, the formation of Ni and Co oxidation peaks also raises the current, to some extent, which causes an obvious peak in the non-Faradaic region of the polarization curve [41,42]. Moreover, as shown in Fig. 5b, it has the smallest Tafel slope of the three samples, P-CoMoO4@NiCoP/NF (126 ​mV dec−1), NiCoP/NF (150 ​mV dec−1) and P-CoMoO4/NF (148 ​mV dec−1), demonstrating the fastest OER kinetics of the P-CoMoO4@NiCoP/NF catalyst. Similarly, the P-CoMoO4@NiCoP/NF achieves a higher current density (j ​= ​231 ​mA ​cm−2) at an overpotential of 300 ​mV than those of NiCoP/NF (j ​= ​124 ​mA ​cm−2) and the P-CoMoO4/NF (j ​= ​183 ​mA ​cm−2) (Fig. S18), indicating the synergistic effect of NiCoP and P-CoMoO4 for improving the OER performance [36]. P-CoMoO4@NiCoP/NF's superhydrophilicity enables it to adsorb water molecules well and promotes the wettability of the electrolyte, thus promoting the surface activity of the catalyst, showing excellent oxygen evolution performance. In the EIS spectra (Fig. 5c), the heterostructured P-CoMoO4@NiCoP/NF exhibits the lowest semicircle. The Rs of P-CoMoO4@NiCoP/NF (1.041 ​Ω) is lower than that of NiCoP/NF (1.181 ​Ω) and P-CoMoO4/NF (1.200 ​Ω), indicating that the P-CoMoO4@NiCoP/NF presents the lowest intrinsic resistance of the electrode and electrolyte among the three catalysts. Similarly, the Rct of P-CoMoO4@NiCoP/NF (0.646 ​Ω) is significantly smaller than that of NiCoP/NF (0.993 ​Ω) and P-CoMoO4/NF (0.858 ​Ω), indicating a faster charge transfer between the P-CoMoO4@NiCoP/NF electrode and the electrolyte [19]. Compared with the Rs (1.615 ​Ω) and Rct (1.717 ​Ω) of CoMoO4@NiCo2O4/NF (Fig. S19), the charge transfer can be increased through phosphating, and then improve the electrical conductivity of the electrocatalyst [20]. In addition, a minimal increase in overpotential is observed even after continuous chronopotentiometry testing over 50 ​h at 100 ​mA ​cm−2 (Fig. 5d), confirming the high stability and durability of the P-CoMoO4@NiCoP/NF. Its superaerophobic property enables P-CoMoO4@NiCoP/NF to release bubbles rapidly to maintain stability. The morphology observation in Fig. S20 shows that the original heterostructure is basically retained.Generally, the proposed mechanism of OER under alkaline conditions are considered as follows (M, surface metal sites)[43]. (3-1) M ​+ ​OH– → M ​− ​OH ​+ ​e– (3-2) M ​− ​OH ​+ ​OH– → M ​− ​O ​+ ​H2O ​+ ​e– (3-3) 2M ​− ​O → O2 ​+ ​2 ​M (3-4) M ​− ​O ​+ ​OH– → M-OOH ​+ ​e– (3-5) M-OOH ​+ ​OH– → M ​+ ​O2 ​+ ​H2O ​+ ​e– This indicates that oxides and hydroxides are all intermediates in the oxygen evolution reaction. It was also reported that the OER electrocatalytic activity of NiCo phosphides are attributable to the Ni-Co oxo/hydroxo species, which are key OER intermediates during oxygen evolution and are partially derived from the oxidization of Ni and Co atoms on the surface of the catalyst [4,9,44]. Fig. S21 shows the XPS spectra of P-CoMoO4@NiCoP/NF before and after the 50 ​h chronopotentiometry test for OER at 100 ​mA ​cm−2 in 1.0 ​M KOH solution. Of note, the peaks of Ni0 and Coδ+ disappeared after 50 ​h OER testing, indicating Ni and Co have been oxidized. Meanwhile, the sole presence of the P-O bond and the disappearance of the P-M bonds are in close correlation with the oxidation of Ni and Co during the catalytic processes [45]. In order to further determine the surface oxidation, the samples were characterized by XRD after the OER stability test (Fig. S22a), and the presence of NiO was found, indicating that surface oxidation did exist on the P-CoMoO4@NiCoP/NF sample. The high-resolution TEM (HRTEM) image (Fig. S22c) on the nanosheet (marked in red circle in Fig. S22b) reveals a clear lattice distance of 0.205 ​nm, may correspond to the crystal plane of Co oxo/hydroxo species. The selected area electron diffraction (SAED) image (Fig. S22d) on the nanowire shows several bright rings with discrete spots, which match well with the (111), (200) and (220) planes of NiO. These results elucidate that an additional oxide catalyst layer is gradually formed on the surface of the [email protected] above experimental results showed that the heterostructured P-CoMoO4@NiCoP/NF electrocatalyst presented a superior bifunctional electrocatalytic performance on HER and OER. Therefore, a two-electrode overall water splitting electrolyzer was constructed using P-CoMoO4@NiCoP/NF as both the anode and the cathode. As shown in Fig. 6 a, P-CoMoO4@NiCoP/NF electrocatalysts are used as anode for OER and cathode for HER. In view of the superior bifunctional characteristics of P-CoMoO4@NiCoP/NF, its possible mechanism for overall water splitting can be illustrated by Fig. 6b. During OER, electrons are transferred from the P-CoMoO4 nanosheets to the NiCoP nanowires via interface action. Then they are transferred from NiCoP to the Ni foam substrate, while the electron transmission path of HER is the opposite [15]. The superhydrophilic and superaerophobic properties of P-CoMoO4@NiCoP/NF make it conducive to water dissociation and bubble separation in the electrocatalytic process, thus improving the catalytic performance and stability. In addition, the heterostructured nanostructures ensure strong electronic interactions of the P-CoMoO4@NiCoP/NF [27]. Because of these advantages, this electrolyzer only needs a low potential of 1.62 ​V to achieve the current density of 20 ​mA ​cm−2 (Fig. 6c), which could be maintained well with almost no degradation after the 50 ​h chronopotentiometry test (Fig. 6d), suggesting an impressive durability of overall water splitting. Such outstanding activity and durability enable the heterostructured P-CoMoO4@NiCoP/NF to be a potential alternative to noble metal electrocatalysts for energy-efficient and cost-effective water splitting [46]. Table S1 compares the overall water splitting performance of P-CoMoO4@NiCoP/NF with some representative catalysts reported recently. It can be seen that it is comparable to or even outperforms its counterparts.In summary, we have developed an efficient strategy for improving the electrocatalytic activity of water splitting by engineering a 3D tree-like heterostructure of P-CoMoO4@NiCoP/NF, which could promote electron and charge transfer and provide abundant active sites. In addition, the superhydrophilic and superaerophobic properties of the P-CoMoO4@NiCoP/NF can facilitate good contact between the catalysts and electrolyte, which is very conducive to water electrolysis. In the half-cell evaluation, P-CoMoO4@NiCoP/NF exhibits excellent HER and OER performance with low overpotentials of 66 ​mV at 10 ​mA ​cm−2 and 252 ​mV at 100 ​mA ​cm−2. Furthermore, it also displays small Tafel slopes of 75 and 126 ​mV dec−1 in alkaline media, as well as high stability, even in the chronopotentiometric testing of 50–60 ​h. Moreover, as both the cathode and the anode, P-CoMoO4@NiCoP/NF exhibits good overall water splitting performance. To reach a current density of 20 ​mA ​cm−2, P-CoMoO4@NiCoP/NF only needs a low potential of 1.62 ​V with 50 ​h durability. This excellent performance indicates that P-CoMoO4@NiCoP/NF is a promising bifunctional electrocatalyst for overall water splitting, which may make it possible to realize large scale, high efficiency catalytic electrolysis of water under high current density.The authors declare no competing financial interests.The authors acknowledge the National Natural Science Foundation of China (NSFC 91834301, 21808046 and 21908037) and Anhui Provincial Science and Technology Department Foundation (201903a05020021 and 202003a05020046) for funding support.The following are the Supplementary data to this article: Multimedia component 1 Multimedia component 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.nanoms.2021.05.004.

Improving catalytic activity and durabilty through the structural and compositional development of bifunctional electrocatalysts with low cost, high activity and stability is a challenging issue in electrochemical water splitting. Herein, we report the fabrication of heterostructured P-CoMoO4@NiCoP on a Ni foam substrate through interface engineering, by adjusting its composition and architecture. Benefitting from the tailored electronic structure and exposed active sites, the heterostructured P-CoMoO4@NiCoP/NF arrays can be coordinated to boost the overall water splitting. In addition, the superhydrophilic and superaerophobic properties of P-CoMoO4@NiCoP/NF make it conducive to water dissociation and bubble separation in the electrocatalytic process. The heterostructured P-CoMoO4@NiCoP/NF exhibits excellent bifunctional electrocatalysis activity with a low overpotential of 66 ​mV at 10 ​mA ​cm−2 for HER and 252 ​mV at 100 ​mA ​cm−2 for OER. Only 1.62 ​V potential is required to deliver 20 ​mA ​cm−2 in a two-electrode electrolysis system, providing a decent overall water splitting performance. The rational construction of the heterostructure makes it possible to regulate the electronic structures and active sites of the electrocatalysts to promote their catalytic activity.

Direct methanol fuel cells (DMFC) are a type of efficient, ‘green’ power source in which chemical energy can be converted to electric energy via the oxygen reduction reaction (ORR) on the cathode and the methanol oxidation reaction (MOR) on the anode [1,2]. It is known that the performance of a DMFC is mainly determined by the MOR kinetics. The oxidization of methanol to CO2 is highly favoured since this oxidation mechanism is accompanied by the simultaneous release of 6 electrons [3]. Unfortunately, this reaction is sluggish. More seriously, CO intermediates are likely to be generated during the MOR. This so-called CO path can easily inactivate noble-metal catalysts with serious consequences [4]. Since the oxidation of CO to CO2 requires the presence of OH*, the adsorbed OH species is indispensable to facilitate CO desorption or convert CO to CO2 [5–7]. Therefore, the design of catalysts with improved ability to adsorb OH species is of great importance in boosting the MOR and eventually enabling the assembly of high-performance DMFCs.One alternative to a Pt catalyst for the MOR is a Pd catalyst, which exhibits better CO resistance. This is due to its better oxophilic nature [8]. To facilitate the increased adsorption of OH species on the Pd catalyst for the MOR, other metals (e.g., Mn, Rh, Ni, Ag) with a high ability to adsorb OH species have been combined with the Pd catalyst. For example, a number of PdM alloys (M = Mn, Rh, Ni, Ag) have been synthesized [9–13]. The drawback of these alloys is the reduced utilization of the Pd catalyst, since the exposed atomic Pd sites in these alloys tend to be occupied by the introduced metals [14].A core–shell structured Pd catalyst is expected to be better for MOR than an alloy, as a core–shell catalyst would effectively have an increased number of atomic Pd sites in the shell. More importantly, it is possible to optimize the electronic structure of the Pd atoms via the strain effect [14–17]. This is because the d-band centre of Pd atoms in the shell can be modified according to the strain state formed [18–22]. Furthermore, the strong interaction between the metal core and the Pd shell facilitates electron transfer between them, which dramatically affects the electronic structure of the Pd shell [23,24]. Consequently, improved OH adsorption on the Pd shell is expected, eventually leading to increased activity for the MOR.In this work, a series of Ag-core/Pd-shell (Ag@Pd x ) catalysts are synthesized. The Ag metal was selected as the core because, with its large lattice parameter, an Ag core can provide a tension strain for a Pd shell. Meanwhile, the thickness of the Pd shell is varied. In particular, the strain and electronic state of a Pd shell can be adjusted by changing the thickness of the Pd shell. The detailed MOR performance of the as-designed Ag@Pd x (x = 1,3,5) and related density functional theory (DFT) calculations are reported in this contribution.Silver nitrate (AgNO3), methanol (CH3OH), triphenylphosphorous (TPP), palladium acetylacetonate (Pd(acac)2), tri-n-octyl oxyphosphorous (TOPO), and borane tert-butylamine complex (BTB) were purchased from Aladdin Chemical Reagent Co. Ltd. Ketjen Carbon was supplied by Shanghai Cuike Chemical Technology. All reagents were of analytical grade and used as received without further purification.Ag@Pd x (x = 1,3,5) catalysts were synthesized by a seed-growth method as described previously [14]. A Pd/C catalyst and AgPd alloy were also synthesized for comparative purposes. Prior to employing these catalysts for the MOR, 10 mg of the catalyst and 90 mg of Ketjen Carbon were dispersed in hexane individually with the aid of sonication for 2 h. Subsequently, the black mixture was centrifuged and dried in a vacuumoven at 60 °C for 12 h. An ink was then prepared by dispersing 5 mg of this black product in 1 mL of ethanol and 20 µL of Nafion solution (Sigma Aldrich, 5 wt%) under sonication for 30 min. To form a working electrode, 5 μL of the resulting ink was coated onto a glassy carbon disk electrode (GCE, 5.0 mm in diameter). After that, the electrode was dried at room temperature for about 10 min.The phase characterization of the fabricated Ag@Pd x catalysts was carried out on a Bruker D8 advance X-ray diffraction (XRD) system (Karlsruhe, Germany) with Cu Kα radiation in the 2θ range from 25° to 90° at a scanning rate of 5° min−1. Transmission electron microscopy images of these catalysts were recorded on a JEOL JEM-2100F (Tokyo, Japan), operated at 200 kV. High-resolution high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images were also recorded. The EDX analysis was performed in the STEM mode using an aberration-corrected JEOL 2200FS electron microscope (Tokyo, Japan) equipped with a Bruker-AXS SDD detector (Karlsruhe, Germany). The X-ray photoelectron spectroscopy (XPS) was carried out on a Kratos Axis Ultra DLD 5 spectrometer (Manchester, UK) with an Al Kα (hv = 1486.6 eV) X-ray excitation source. All XPS spectra were calibrated using the C1s peak located at 284.6 eV.The MOR performance was evaluated in a three-electrode system using a CHI850 electrochemical workstation (Shanghai, China). A Pt sheet (1.5 × 1.5 cm2) and a saturatedcalomelelectrode (SCE) were selected as the counter electrode and the reference electrode, respectively. The working electrode was the ink-coated GCE. The electrochemical active surface area (ECSA) was calculated by the equation: ECSA = Q / (m∙C), where Q is the calculated charge required to reduce PdO to Pd on the catalyst surface, m is the Pd atomic mass of the catalyst dropped on the GCE surface, and C is the theoretical charge (420 μC cm−2) needed to reduce a layer of PdO to Pd [25].All calculations were performed using the Perdew − Burke − Ernzerhof method implemented in the Vienna ab initio Simulation Package [26–28]. The interactions between ion cores and valence electrons were calculated with a projector augmented wave method [28]. The valence electronic states were expanded in the plane wave basis sets within a cut-off energy of 400 eV. The Pd(111) surface was modelled as a periodic slab with a p(3 × 3) unit cell with four Pd layers. The two layers at the bottom were fixed in the slab while the two layers at the top, together with the adsorbates, were relaxed throughout the geometry optimization. For these surface slabs, the Brillouin zone of the surface calculations was sampled with 4 × 4 × 1 Monkhorst-Pack mesh. The geometric optimizations converged when the energy difference was smaller than 10-5 eV and the forces were less than 0.01 eV Å−1.The solvent reduction method used is illustrated schematically in Fig. 1 a. The Ag core is 3.2 nm in size (Fig. S1a, b). The thickness of the Pd shell grown on the surface of the Ag core is varied by altering the amount of the Pd(acac)2 precursor used. In this way, three Ag@Pd x nanoparticles were synthesized with Pd shells of different thicknesses. From the TEM images, one can see that they are uniformly distributed on the carbon support (Fig. 1b–d). Their sizes (Fig. S2a–c) are 3.7, 4.6, and 5.4 nm, respectively. In other words, their size increases when the shell is thicker or the number of Pd layers in the shell is increased. These particles are named Ag@Pd1, Ag@Pd3, and Ag@Pd5 catalysts, respectively, throughout this paper. As control experiments, well-dispersed Pd/C (2.6 nm) and AgPd alloy (7.6 nm) catalysts were also synthesized (Fig. S3). As an example, the HAADF-STEM image of a Ag@Pd3 nanoparticle (Fig. 1e) reveals an interplanar spacing of 0.234 nm in the inner layer and an interplanar spacing of 0.228 nm in the outer layer. These are indexed to Ag (111) and Pd (111), respectively. Note that the interplanar spacing of the Pd shell is larger than that of the standard Pd (111) (0.224 nm) [29]. In other words, the Pd shell is stretched by the Ag core. EDX element mapping of two Ag@Pd3 nanoparticles was then conducted (Fig. 1f), illustrating their core–shell structure, in that the Pd element is distributed around the Ag element. In addition, the estimated thickness of the Pd shell is about 0.8 nm. According to the EDX line profiles (Fig. 1g), a Ag@Pd3 nanoparticle has nearly 3 Pd layers in the shell.XRD analysis of these nanoparticles was then performed (Fig. S4). The Ag core exhibits a typical FCC structure with diffraction peaks located at 38.2°, 44.4°, 64.3°, 77.5°, 81.4°, corresponding to (111), (200), (220), (311) and (222) of Ag, respectively [14]. As a control experiment, the visible diffraction peaks of the Pd/C catalyst are located at 39.6°, 46.9°, 67.5°, and 80.1°, associated with (111), (200), (220), (311) of Pd, respectively. After the growth of a Pd shell on this Ag core, no obvious changes are found in the XRD peaks of the Ag core. Meanwhile, no Pd diffraction peaks are seen in the XRD patterns of the Ag@Pd x catalysts, indicating the ultrathin nature of the Pd shell.In order to investigate the electronic interactions between the Ag core and the Pd shell in the Ag@Pd x nanoparticles, an XPS analysis was carried out. In the Ag 3d XPS spectra of metallic Ag and the Ag core, both Ag (0) and Ag (+1) are observed and most of the Ag is shown to be in the zero-valence state (Fig. 2 a). A very tiny shift in the binding energy of Ag is observed. For example, the binding energies of Ag (0) 3d5/2 for the Ag@Pd1 (367.43 eV), Ag@Pd3 (367.46 eV) and Ag@Pd5 (367.44 eV) nanoparticles are slightly more negative than that of the Ag core (367.73 eV). Moreover, the binding energies of Ag 3d5/2 in these Ag@Pd x particles are very similar. With respect to the Pd 3d XPS spectra of these nanoparticles (Fig. 2b), most of the Pd present is shown to be in the zero-valence state. Only a small amount of Pd (+2) is seen, probably due to surface oxidation. Compared with the Pd/C catalyst, the binding energy of Pd 3d5/2 in the Ag@Pd x nanoparticles shifts negatively by about 0.6 eV. The shift in this binding energy demonstrates electron transfer from the Ag core to the Pd shell. When this Pd shell becomes thicker, the negative shift in the related binding energy turns out to be weaker. This phenomenon confirms the reduced electronic effect in these Ag@Pd x nanoparticles.The electrochemical MOR performance of these Ag@Pd x nanoparticle-based catalysts was then tested in 1 M KOH solution (Fig. 3 a). The reduction peak of AgO to Ag is seen at ~0.05 V in the cyclic voltammogram (CV) of the Ag nanoparticles (or the Ag core). It is weakened and eventually disappears when the Ag core is coated with a Pd shell. Simultaneously, a new reduction peak appears at about −0.4 V when a Pd shell is coated on the Ag core. Obviously, this peak is due to the reduction of PdO to Pd. Using the area of this cathodic peak, the ECSAs of the catalysts were calculated and further utilized as an index of active sites inside these catalysts [25]. The ECSA of the Ag@Pd1 nanoparticle is 113 m2 g−1, larger than that of Ag@Pd3 (98 m2 g−1), Ag@Pd5 (81 m2 g−1), and Pd/C (61 m2 g−1) nanoparticles (Table S1). In other words, a thinner Pd shell exposes more active Pd sites. The MOR activity of these catalysts was then tested in 1 M KOH + 1 M CH3OH solution. Note that for all these catalysts it is necessary to activate them after several voltammetric cycles (Fig. S5). An example of such activated CVs is shown in Fig. 3b. Surprisingly, the MOR catalytic activity of the Ag@Pd3 (2369 mA mg−1 Pd) catalyst is higher than that of Ag@Pd1 (1322 mA mg−1 Pd), Ag@Pd5 (1703 mA mg−1 Pd), and Pd/C (571 mA mg−1 Pd) catalysts (Fig. 3b). This is probably due to the fact that the Ag@Pd3 catalyst has the best intrinsic activity among these catalysts. The specific current was further normalized by its ECSA. The Ag@Pd3 catalyst exhibits the highest intrinsic activity in that the peak current on the Ag@Pd3 catalyst is about 2.4 mA cm−1 larger than that of the Ag@Pd (1.5 mA cm−1), Ag@Pd5 (2.1 mA cm−1) and Pd/C (0.94 mA cm−1) catalysts (Fig. 3c, Table S1). Moreover, the Ag@Pd3 catalyst shows better MOR performance than the AgPd alloy with respect to the mass activity (985 mA mg−1 Pd) and the specific activity (1.4 mA cm−1) (Fig. S6, Table S1). In fact, the AgPd alloy exhibits enhanced MOR activity compared to the Pd/C catalyst, probably originating from a bifunctional mechanism [30]. Specifically, the OH radical can be adsorbed on the Ag sites of the AgPd alloy, leading to improved MOR activity. When compared with recently reported Pd-based catalysts, the mass activity and specific activity of the Ag@Pd3 catalyst are found to be greatly superior (Table S2).The cycling stability of the catalysts was also evaluated. After 1000 cycles of cyclic voltammetric tests, the mass activity of the Ag@Pd3 catalyst still retained 43% of its initial value (Table S1). This is much higher than the corresponding values for Ag@Pd1 (23%), Ag@Pd5 (18%), AgPd alloy (21%), and Pd/C (4%) catalysts. Moreover, the Ag@Pd3 catalyst still retains a pair of typical MOR peaks even after 1000 cyclic voltammetric cycles. This performance is much better than that of the Pd/C catalyst (Fig. S7). Actually, aggregation of the Ag@Pd3 catalyst occurring during the course of such a stability test is obviously reduced when compared with the Pd/C catalyst (Fig. S8). In order to further investigate the change in the active sites of the Ag@Pd3 catalyst during the stability test, the Pd 3d XPS spectrum of the Ag@Pd3 catalyst was recorded after such a stability test (Fig. S9). A much more pronounced Pd (+2) peak is visible. Consequently, the Pd (0) is oxidized to Pd (+2) during the MOR. This is the reason why its MOR activity decreases as a function of running time. Moreover, the atomic ratio of Ag:Pd in the Ag@Pd3 catalyst changes from 1:2 to 1:1.5, which suggests Pd dissolution during the MOR. This is another source of the decreased MOR activity.The MOR mechanism on the Ag@Pd x catalysts was further examined by DFT calculations using a VASP code. To evaluate the electronic and strain effects in the Ag@Pd x catalysts, simple Ag@Pd (111) and tensional Pd (111) models were constructed (Fig. S10). Compared with Pd (111)-0%, both d-band electron energy ranges of Pd (111)-3% and Pd (111)-5% are decreased. On the other hand, the d-band centre of Pd (111)-5% is −1.44 eV (Fig. 4 a). This is closer to the Fermi level than that of Pd (111)-3% (-1.47 eV) and Pd (111)-0% (-1.55 eV). As for Ag@Pd(111), its d-band electron state is more negative and narrower, although the strain state of Pd in Ag@Pd(111) is close to that in Pd(111)-5%. All these results confirm that there is an electronic effect between Ag and Pd. The projected partial density of states (PDOS) of Ag after being coated with a Pd layer exhibits several new weak peaks in the range from −0.3 to −2 eV. They are in line with those in the PDOS energy range from 0 to −3 eV for Pd in Ag@Pd (111) (Fig. 4b). Consequently, the tensional strain effect makes the d-band centre shift closer to the Fermi level, while the electronic effect causes the d-band centre to move far away from the Fermi level in the Ag@Pd system. Since the Pd atom plays an important role in the adsorption action, the adsorption energy of CH3OH on Pd (111) is increased when the Pd (111) is stretched, but the adsorb energy of CH3OH on Ag@Pd (111) is still smaller than the Pd (111)-0% (Table S3).The difference in the adsorption energies was further elucidated using these catalyst models. According to the PDOS of free and adsorbed CH3OH (Fig. 4c), the electronic energy of CH3OH is decreased after the adsorption of CH3OH on Pd (111)-0%. This indicates that the system of CH3OH-Pd (111)-0% is stable. Moreover, a higher reduction in energy is seen when CH3OH is adsorbed on Pd (111)-3% and Pd (111)-5%. In other words, CH3OH is more easily adsorbed on the tensional Pd (111). However, when CH3OH is adsorbed on Ag@Pd (111), the corresponding energy is higher than that on Pd (111)-0%. This result reveals that CH3OH has a weaker adsorption energy on Ag@Pd (111). On the other hand, the adsorption energy and PDOS of OH are similar to the action of CH3OH on the Pd-based models (Table S3, Fig. 4d). These results show that the strain effect in the Ag@Pd x catalysts is beneficial for the adsorption of CH3OH and OH, while the electronic effect in the Ag@Pd x catalysts is harmful for the adsorption of CH3OH and OH. In short, the Ag@Pd3 catalyst exhibits superior activity to its counterparts, which is mainly ascribed to the optimal combination of strain and electronic effects.In summary, a number of core–shell Ag@Pd x catalysts have been synthesized with a Pd shell of different thicknesses. Optimizing the thickness of the Pd shell leads to enhanced MOR activity. As confirmed by DFT simulations, this high MOR activity is attributed to an optimal combination of strain and electronic effects in the Ag@Pd3 catalyst. This work offers new insight into strain and electronic effects in core–shell structured electrocatalysts. It thus provides a novel approach to the design of various high-performance MOR electrocatalysts. Such catalysts are extremely promising for future use in the mass production of direct methanol fuel cells. Xiaobo Yang: Investigation, Writing - original draft. Xili Tong: Project administration, Writing - review & editing, Supervision. Xingchen Liu: Investigation. Kaixi Li: Supervision. Nianjun Yang: Writing - review & editing, Supervision.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This work is financially supported by the Joint Funds of the National Natural Science Foundation of China (U1710112) and the State Key Laboratory of Coal Conversion (2020BWL001).Supplementary data to this article can be found online at https://doi.org/10.1016/j.elecom.2021.106917.The following are the Supplementary data to this article: Supplementary data 1

The performance of a direct methanol fuel cell (DMFC) is strongly dependent on the catalytic anode. A high-performance anode is expected to offer enhanced intrinsic activity and/or a large electrochemical surface area. Herein, a series of Ag-core/Pd-shell (Ag@Pd x , x = 1,3,5) catalysts are synthesized in which the thickness of the Pd shell is varied. Both tensional strain and electron transfer between the Ag core and the Pd shell are found to affect the intrinsic activity of these Ag@Pd x catalysts. Of these, the Ag@Pd3 catalyst exhibits the best performance for the methanol oxidation reaction (MOR), showing 4.1 times higher mass activity and 2.6 times higher specific activity than a Pd/C catalyst. Furthermore, density functional theory calculations show that this high MOR performance stems from a stronger adsorption of CH3OH and OH on the Pd active sites. This catalyst is thus a promising candidate for inclusion in a high-performance DMFC.

Increasing concerns about energy consumption, exhaustion of fossil fuel resources and global warming have enhanced attention in making of bio-based chemicals/fuels through biorefineries [1–3]. Lignocellulosic biomass is the prompting abundant carbon source in making value added chemicals and fuels. Cellulose and hemicellulose are the key composites in lignocellulosic biomass [4–7]. Hexoses such as glucose and fructose are the major products obtained from cellulosic biomass. HMF, a product derived from hexoses by dehydration, considered as a platform chemical from bio-renewables [8,9]. The high functionality of HMF allows to convert into diverse biofuel molecules like ethyl levulinate (EL), 5-ethoxymethylfurfural (EMF), 2,5-dimethylfuran (DMF), 2,5-dimethyltetrahydrofuran (DMTHF) and value-added chemicals such as levulinic acid (LA), 2,5- diformylfuran (DFF), 2,5-furandicarboxylic acid, terephthalic acid, and caprolactone, etc. [10–15]. Among various chemicals, DMF is more attractive in particular among other HMF derivatives because of its elevated boiling point, better energy density and high octane number which are greater than the current gasoline alternatives like ethanol and butanol. DMF is insoluble in water and is easy to store. These properties make DMF as an effective gasoline alternative and can be considered as an alternative transportation biofuel. DMF can also be used as a diene in Diels‐Alder reactions [16–18].Noble metal catalysts such as Pt and Pd are widely used in HMF hydrodeoxygenation (HDO) due to their high reactivity in hydrogenation reactions. Different kinds of carbon supported Ru based catalysts are used for HMF hydrodeoxygenation to DMF with good yields [19–22]. Ru based bimetallic catalysts are also showed high yield of DMF with excellent reusability of the catalysts [23–25]. Due to the high reactivity of Ru, these catalysts lead to furan ring over hydrogenation and formation of ring opened products. Pt supported on mesoporous carbon with nitrogen rich and Pt on reduced graphin oxide catalysts are tested for HDO of HMF. These catalytic systems showed moderate DMF yields [26,27]. Further few research groups worked on Ru and Co based bimetallic catalysts to improve DMF yield [28,29]. Even though these catalysts showed good DMF yield, they found the Co metal leaching due to the weak interaction with the supports. Yadav et al., reported Pd on K-10 supported Cs-modified heteropoly acid catalyst with comparable activity [30]. Pd-Au and Pd-Zn based bimetallic catalysts also used for HMF hydrogenation with enhanced activity [31,32]. These noble metal catalysts showed good activity; however, their high cost, non-selective nature due to high reactivity and limited stability limited their usage for the conversion of HMF to DMF. In this context, non noble metals like Co, Ni and Cu catalysts have attracted great attention because of their low cost, less hazardous nature and high selectivity compared to noble metals. Co and Cu bimetallic catalyst are also active for this reaction only at prolonged reaction times [33–35]. Also, there are many bifunctional Ni based catalysts reported for the HMF hydrogenation with considerable activity. Many of these catalysts showed high selectivity towards DMTHF over DMF and also observed side products like 2-methyl furan, THF and ring opened products [36–38]. Comparative to Ni based catalysts Cu catalysts showed better selectivity towards DMF as Cu shows high reactivity to C=O bond over C=C bond. Dumesic et al., reported copper chromite catalyst for HMF hydrodeoxygenation. In that study they converted sugars and obtained DMF with 67% selectivity. However, the high copper loading and toxicity of chromium had a negative impact [39]. Barta et al., reported the activity of different commercially available copper catalysts such as, Cu, CuO, CuO-Fe2O3, CuZnFe2O4 and CuZn. Among the catalysts, CuZnO showed better activity at 220 °C with 30 bar H2 pressure in 6 h of reaction time. This catalyst showed the activity drop up to 17% in its reusability [40]. Zhu et al., demonstrated CuZn catalyst with good activity towards DMF at 220 °C with 15 bar H2 in 5 h. Most of the studies focused on the Cu based mixed oxide catalysts with high content of Cu up to 54%, which causes the unstable nature of the catalyst during recyclability [41]. Rupert and group also reported CuZnO catalyst with self-tuned properties towards BHMF and DMF. Moreover, most of these bulk Cu based catalysts are not stable during recycling experiments [42]. Esteves et al. demonstrated the use different metal oxide (Al2O3, Nb2O5, Al2O3-Nb2O3) supported Cu catalysts for HDO of HMF [43]. There is a need to develop non-noble metal like Cu based active catalysts which can efficiently catalyse 5-HMF hydrodeoxygenation at moderate reaction conditions within reasonable reaction time. The activity of Cu based catalysts mainly depends on Cu dispersion which can be improved if porous metal oxides with high surface area are used as support [44]. It is thought to prepare Cu based catalyst on mesoporous materials with high surface area to achieve high activity with selectivity during HMF conversion to DMF.Here in this paper, copper supported SBA-15 catalysts are prepared and studied for the hydrodeoxygenation of HMF to selectively yield DMF. The catalysts are characterized with various techniques to derive their surface-structure characteristics. All these characteristics are utilized in understanding the catalysts for their selective HDO activity of 5-HMF.A set of Cu supported on SBA-15 catalysts were designed and prepared by impregnation method. Initially, mesoporous silica SBA-15 was prepared based on the reported literature using a silica source tetraethyl orthosilicate (TEOS) [45,46]. In the preparation, P123 is used as structure directing material. An aqueous solution with the composition of TEOS:P123:2M HCl:H2O = 4.25:2:60:15 (weight ratio) was stirred using magnetic bead at 40 °C for 24 h. Then this mixture was transferred to Teflon bottle for hydrothermal treatment for 24 h at 100 °C. After that the resultant gel solution was washed with distilled water for several times up to pH of the solution becomes neutral during filtration. The filtrate was oven dried for overnight at 80 °C. The calcination of dried solid was carried out at 550 °C in presence of air flow for 8 h. The resultant mesoporous silica (SBA-15) was used as support. Cu supported on SBA-15 catalysts with various loadings (5, 10, 15, 20 wt%) of Cu were prepared by wet impregnation method by using copper nitrate as precursor. The required quantity of the precursor dissolved in water was added to the SBA-15 and the mixture was dried with infrequent stirring on a hot plate. The solid mass was dried at 100 °C in an air oven for 12 h. Later calcined in air at 450 °C for 5 h. These samples were denoted as X%Cu/SBA-15, where X represents the Cu weight percentage.HMF hydrogenation was carried out in a 100 mL Parr autoclave reactor. Initially, the catalyst (0.15 g) was reduced at 400 °C in presence of H2 flow (35 mL/min) for 2 h. After reduction the sample was cooled to room temperature in presence N2 flow. The autoclave reactor was charged with the catalyst (0.15 g), HMF (2 mmol, 0.252 g) in THF (20 mL) and the reactor was flushed with H2 for three times. The reaction was performed at 180 °C under 2.0 MPa H2 pressure with a stirring rate of 300 rpm/min for stipulated reaction time. The reactor was cooled after completion of the reaction to room temperature and it was filtered for product analysis.Products were confirmed by GCMS (Shimadzu, GCMS-QP2010S) and conversion and yields were determined by using Shimadzu 2010 gas chromatography equipped with capillary column. Flame ionization detector was used to analyze the products by separating them on Inno wax capillary column (diameter: 0.25 mm, length: 30 m). Conversion of HMF and yields were estimated based on the equations shown below. H M F C o n v e r s i o n ( % ) = Mi - M f Mi × 100 Y i e l d ( % ) = Pi Mi × 100 Mi refers the initial moles of HMF, Mf refers the final moles of HMF and Pi stands for the moles of product i formed.The catalysts were characterized by different techniques like powder X-ray diffraction (XRD), BET surface area, temperature-programmed reduction (TPR), temperature-programmed desorption of NH3 (TPD-NH3) and transmission electron microscopy (TEM) and scanning electron microscopy (SEM). The details of these experimental procedures are reported in our earlier publications [47–49].The dissociative N2O chemisorption method was carried on the instrument (BELCAT-II, BEL Japan) used for TPR studies to determine copper metal surface area, particle size and dispersion. The sample (100 mg) was placed in the sample tube and pre-treated in Ar gas (50 mL/min) flow for 30 min at 100 °C. Firstly catalyst was pre-reduced by increasing the temperature to 400 °C with a ramp of 10 °C/min under 5% H2/Ar flow. This step is termed as TPR1. Then the sample was cooled in Ar flow to 50 °C and exposed sequentially to 10% N2O/He gas for 1 h, to re-oxidize metallic Cu to Cu2O by N2O dissociative chemisorption. The sample the purged in Ar flow at 50 °C for 15 min, again temperature programmed reduction (TPR2) was carried to reduce the oxidized Cu2O species to copper metal. The copper metallic area, particle size and dispersion were estimated by the following equation Dispersion D % = 2 × H 2 c o n s u m p t i o n i n T P R 2 H 2 c o n s u m p t i o n i n T P R 1 = × 100 Specific C u s u r f a c e a r e a S 2 × H 2 c o n s u m p t i o n i n T P R 2 × N H 2 c o n s u m p t i o n i n T P R 1 × M Cu × 1.4 × 10 9 = m 2 g - 1 Where N = Avogadro constant, M Cu = atomic mass (3.45 g mol−1) and the number of superficial Cu atoms per unit surface area as 1.47 × 1019 atoms/m2, and the density of copper as 8.92 g/cm3.The surface properties of the catalysts were drawn from N2-adsorption and desorption isotherms and the profiles are shown in Fig. 1 A. The isotherms of the catalysts and support representing type-IV with H2 like hysteresis loop related to the mesoporous nature of the catalysts [50]. These profiles show that the mesoporous structure of SBA-15 was retained in the catalysts even after the dispersion of CuO. The increase in N2 uptake was observed at high P/P0 could be initiated by capillary condensation of multi-layer adsorption. The textural properties of the catalysts are shown in Table 1 . The immobilization of Cu species significantly altered the surface properties of SBA-15. The surface area of bulk SBA-15 is 676 m2/g and the surface area values are decreased with increase in copper content on SBA-15. The inner pore blockage of SBA-15 with deposition of CuO causes decrease in the surface areas which is in good agreement with pore size distribution measurements. Pore size distribution curves were employed by BJH (Barrett-Joyner-Halenda) method considering by adsorption branches of isotherms, all the curves confirmed the pores are in mesoporous region as showed in Fig. 2(B). The presence of Cu particles with in the channels of SBA-15 block the pores of the support. Due to this pore blockage, there is a partial strain in the pores that leads to a marginal increase in pore diameter increase in Cu loading [51].Low-angle XRD patterns of Cu/SBA-15 samples along with support SBA-15 are shown in Fig. 2(A) . All the catalysts exhibited an intense peak at 2θ = 0.9° designated to (100) plane and two less intense reflections at 1.6, 1.8° corresponding to (110) and (200) planes in the 2D hexagonal pore arrangement [52]. A marginal shift in (100) plane 2θ value with the existence of CuO on SBA-15. This indicates the presence of intact ordered mesoporous structure with minor distortion after deposition of CuO. Fig. 2(B) shows the wide angle XRD patterns of pure SBA-15 and copper loaded SBA-15 catalysts. All the composite samples displayed a broad intense peak around 23.5°, assigned to the amorphous nature of SiO2. The XRD patterns of the samples showed the presence of intense reflections at 2θ = 32.4°, 35.7°, 38.9°, 48.6°, 53.7°, 58.1°,61.5°, 66.1°, 67.9, 72.3°, which are attributed to the monoclinic CuO (JCPDS No. 65-2309). These results indicating the presence of crystalline CuO over SBA-15 support. Powder XRD patterns of reduced catalysts are inserted in Fig. 2(B). The most intense peaks at 2θ = 44.3°, 50.4°, and 74.1° corresponding to (111), (200), and (220) planes of metallic copper are observed. No other peaks were noticed in the patterns, indicating that Cu existed as Cu0 species after the reduction.H2-TPR analysis was performed to know the reducibility of CuO particles that are in contact with support SBA-15 and the resulted profiles are shown in Fig. 3 . The TPR profiles of Cu loaded SBA-15 catalysts showed two reduction peaks, one is in the range of 280–300 °C and another one is at 390 °C. According to the literature the finely dispersed copper oxide species are easily reducible at lower temperatures comparatively the bulk species reduce at higher temperatures [53,54]. If the interaction between support and metal is weak then it would reduce at low temperature region and the strong interactions of metal species with support will be reduced at higher temperatures. When the Cu content enhanced from 5 to 20 wt%, a shift in the main reduction peak towards high temperature was noticed and it became relatively broad. This indicates that, with increasing the Cu loading on SBA-15 resulted to form strong interaction with support SBA-15. The catalysts with high Cu content (15 and 20 wt%) exhibited a small reduction peak at low temperature (220 °C) might be related to the reduction of bulk CuO species present in the samples. The N2O chemisorption technique used to estimate copper dispersion, metal surface area and particle size and the results are presented in Table 1. The N2O chemisorption results suggesting that the dispersion of Cu particles on SBA-15 was decreased with the increase in Cu loading. This could happen due to the formation of Cu metal aggregated species, which resulted in the decrease of metal area and increase of particle size. N2O chemisorption results also supporting the TPR results.Temperature programmed desorption of NH3 performed to calculate the total acidity and acidic strength of the catalysts. The acidity profiles of pure SBA-15 and Cu supported on SBA-15 are shown in Fig. 4 . In general, the desorbed peak of NH3 was classified into three temperatures corresponding to the acidic sites as weak (<200 °C), medium (200–400 °C) and strong (>400 °C). The Cu/SBA-15 catalysts showed all these three types of acidic sites. The increase in copper loading tend to increase area under NH3 desorbed peak. This indicates that the moderate acidic sites associated with Cu are increased with increase in its content in the catalysts. The acidity of the catalysts is mainly associated with the presence of ionic copper species [55]. The total acidity values calculated by considering the area under the desorbed peaks and the results are shown in Table 1. The results indicate that with the increase in copper loading led to the improvement in the total acidity.To visually investigate the distribution of the particles of copper on SBA-15 support, TEM was carried out. TEM micrographs of bare SBA-15 and 15%Cu/SBA-15 are shown in Fig. 5 . The images show the hexagonal array of uniform channels, which justify the highly ordered mesostructured materials. Copper particles are uniformly distributed on the surface of the support. The average particle size of 15%Cu/SBA-15 was 6.3 nm. TEM images further confirms that structural ordering is maintained even after the Cu-incorporation in SBA-15 matrices.SEM images of bulk SBA-15 support and prepared Cu/SBA-15 samples are shown in Fig. 5(D and E). The bare SBA-15 support showed a fine rod-like mesoporous structures. Hence this result confirms that SBA-15 has 2D hexagonal arrays of mesoporous structure. The morphology shows that the mesoporosity is retained even after impregnation with copper and catalysts exhibited homogeneously dispersed CuO particles. The SEM images of Cu/SBA-15 clearly indicate that the copper oxide is present in a highly dispersed state.Selective HDO of HMF was performed with copper loaded SBA-15 catalysts. The activity results are shown in Fig. 6 . The 5% Cu/SBA-15 catalyst showed 100% of HMF conversion with 65% of DMF yield and 33% yield of methyl furfural (MFA). With increasing the copper loading, HMF is totally converted and DMF yield increased for catalyst with 15% Cu loading and at the same time MFA yield was decreased. Further increase in copper loading to 20%, the yield of DMF was decreased to 74% from 90% and dimethyl tetrahydrofuran (DMTHF) yield was increased to 13% and also the ring opening products like 2, 5-hexenediol and alcohol are formed. Among the catalysts 15% Cu/SBA-15 showed promising catalytic activity towards DMF.The aforementioned results suggesting that the 15% Cu/SBA-15 catalyst was resulted in maximum of 90% DMF yield in 8 h of reaction time. The activity of the catalysts can be explained in the light of the characteristics of these catalysts derived from different analysis methods. The N2O chemisorption results suggesting that the dispersion and metal surface area of Cu particles on SBA-15 were decreased and particle size was increased with the increase in Cu loading. Metal particle size is increased with the Cu content on SBA-15, might be the reason for the marginal decrease in metal dispersion and surface area. The dispersion of copper species on support increases the number of active metal sites which are responsible for high hydrogenation activity of catalysts. Although the lower loading (5 and 10%) catalysts showed high metal surface area and high metal dispersion, these catalysts displayed low yields. This is due to the insufficient number of metal sites with low Cu content. The 15% Cu/SBA-15 catalyst with sufficient number of active metal sites and enough total acidity, showed high yield of DMF. Further increase in loading of Cu on SBA-15 resulted in the decline of DMF yield. The high amount of Cu and high acidity led to the formation of over hydrogenated product DMTHF and other ring opened products as by products. Similar kind of results were observed over Cu/SiO2 catalysts for HMF hydrogenation to DHMF [56]. They used mild reaction conditions of 100 °C, 4 h reaction time and 15 bar H2 Pressure for the preparation of DHMF at high loading of Cu (50%) on SiO2. MFA and DMF yields were increased because of high copper loading and acidity, which promoted the further hydrogenation of DHMF. Zhu et. al., developed the Cu/Zn catalysts derived from minerals. The Cu-ZnO catalyst with molar ratio 2 showed a high yield of 92% DMF. This is attributed to well dispersion of Cu metal sites over support ZnO surface, which allows the high Cu metal concentration and suitable acidity [41]. Ruppert et al., reported Cu/ZnO catalysts for DHMF and DMF synthesis from HMF over 10% Cu/ZnO catalyst. The selectivity was related to the presence of acid sites at the catalyst surface [42]. In the present work, comparatively lower loading of Cu (15%) on SBA-15 and targeted for the DMF synthesis from HMF. Availability of well dispersed copper metal with sufficient number of acidic sites and total acidity might be the reason for the selective formation of DMF over DHMF. The acidity of the catalysts is responsible for the carbonyl group activation [57]. Hydrogenolysis is more preferable over hydrogenation in case of the high acidic catalysts. TPD of NH3 results suggesting that the 15%Cu/SBA-15 have enough acidity to activate the carbonyl group of 5-HMF. Additionally, the literature results suggesting that the low temperatures reduction peaks in TPR indicate the CuO with small particles size or oligomeric clusters with a relatively weak interaction with support. The high temperature reduction peak is usually indicating a relatively stronger interaction of CuO with the support [58]. From the TPR observations, it is supporting that CuO species in a strong interaction with silica facilitates the hydrogenolysis reaction more. Moderate metal surface area with small particle size and enough acidity makes 15%Cu/SBA-15 catalyst more active than other catalysts for selective formation of DMF by the HDO of HMF. The 15%Cu/SBA-15 catalyst was more active among the all catalysts and further used for optimization of reaction conditions.In order to verify the effect of catalyst loading on HDO of HMF to DMF, the experiments were conducted with different amount of catalyst. Catalyst loading was calculated based on the ratio of catalyst weight to total reaction mixture weight (only liquid mass). Catalyst weight percentage was varied in between 0.27 and 1.09 wt% and the results are shown in Fig. 7 . When the catalyst amount is 0.27 wt%, the conversion of HMF was 70% with 37% yield of DMF with 30% of intermediate MFA. Upon increasing the catalyst amount conversion of HMF and DMF yields were increased up to the catalyst loading of 0.82 wt%. Further increase in catalyst loading to 1.09 wt%, DMF yield was decreased to 78% and DMTHF a ring hydrogenated product formation was increased to 10%. The maximum DMF yield was observed with catalyst weight of 0.82 wt%. At low catalyst loading the availability of number of active metal sites are low, which were not sufficient to convert maximum HMF to DMF and as a result, the maximum amount of intermediate MFA was remained in the reaction mixture. When high amount of catalyst was used, the number of active metallic sites and acidity were high, resulted in the formation of ring hydrogenated product DMTHF and also ring opened products.The effect of reaction time was carried to understand the reaction path way. The output of reaction time effect is shown in Fig. 8 . The reaction was performed in the time intervals of 2–10 h. At the initial time of 2 h, the conversion of HMF was reached to 69% with 33% of MFA as major product. With increasing the reaction time to 4 h, the conversion of HMF increased to 76% with 46% yield of DMF along with 28% of MFA. The maximum HMF conversion (100%) and DMF yield (90%) was observed at 8 h of reaction time. Further increase in time led to decrease in DMF yield (74%) and formation of DMTHF with a yield of 18% was observed. Some of the ring open products like 2, 5-hexanedione and 2, 5-hexanediols were observed at prolonged reaction time (10 h). These results suggest that the HDO of HMF over Cu/SBA-15 catalyst going through the formation of MFA as an intermediate before the formation of desired DMF. Further increase in time led to the conversion of DMF to the over hydrogenated product DMTHF i.e, furan ring hydrogenated product and ring opened products like 2, 5-hexanedione and diols.The influence of hydrogen pressure on HMF conversion and DMF yield over the catalyst was also investigated and the results are shown in Fig. 9 . When the pressure was 1 MPa, the conversion and selectivity are 88% and 55% respectively. Upon increasing pressure to 1.5 MPa the conversion of HMF is 100%, the selectivity of DMF (81%) is not maximum and the presence of intermediates were noticed. Further increasing pressure to 2 MPa, the DMF selectivity reached maximum of 100%. Further enhancement in pressure, HMF conversion remained constant and DMF selectivity decreased to 71% due to the formation of over hydrogenated product DMTHF with 22% selectivity.Effect of reaction temperature on hydrodeoxygenation of HMF is one of the crucial parameters in optimizing reaction parameters. The temperature effect was studied in the range of 140–200 °C at a fixed H2 pressure of 2.0 MPa and the results are presented in Fig. 10 . The HMF conversion (65%) and DMF yield (34%) were low at 140 °C. Further increase in temperature to 160 °C led to the maximum conversion of HMF (100%) with 70% of DMF yield and 23% yield of MFA. The reaction at 180 °C was given the maximum yield of 90% DMF with 5% of DMTHF and no MFA was observed at this temperature. Further elevation in the temperature to 200 °C, a drop in DMF yield (79%) was observed with increased DMTHF yield to 16%. From the above results it is observed that lower reaction temperatures favor the hydrogenolysis of O–H bond in HMF over Cu/SBA-15 catalyst which results in the formation of MFA as a major product. However, elevated temperatures encourage the C=O bond reduction along with the hydrogenolysis of O–H bond, which favor the conversion of intermediate MFA to DMF and also ring hydrogenated and ring opened products. The optimum reaction temperature was set as 180 °C based on these experiments.Based on the experiments carried at optimized reaction conditions at different time intervals, a plausible reaction mechanism was proposed for hydrodeoxygenation of HMF to DMF over Cu/SBA-15 catalyst (Scheme 1 ). During the hydrodeoxygenation reaction, hydrogen molecules dissociate on the Cu metallic sites. The alcohol group is removed as H2O on the acidic sites and forms methyl furfural. The formation of this intermediate is observed during the reaction. The carbonyl group of furfural moiety further hydrogenated to alcohol with the dissociated hydrogen present on Cu metal sites. Then the alcohol group is converted to methyl by the elimination water as mentioned above to form DMF.The present catalyst 15Cu/SBA-15 activity was compared with previously reported catalysts and the results are displayed in Table 2 . Srivastava et al. reported the hydrogenation of HMF over Cu–Co supported on Al2O3 and achieved 68.4% selectivity of DMF with 99.9% conversion of HMF in 8 h reaction time at a temperature of 200 °C with 3.0 MPa H2 pressure [59]. J. Wang and co-workers prepared cobalt-based N-doped carbon catalyst and obtained 83.1% conversion of HMF with 83.1% selectivity of DMF [12]. Cobalt-based catalysts exhibited 83.3% selectivity of DMF with 100% conversion of HMF at 170 °C in 12 h in 1,4-dioxane solvent [60]. Laura M. Esteves group reported copper supported on niobium-alumina mixed oxide and achieved 84.5% yield of DMF but in second run the yield of DMF was drastically decreased to 49.5% [43]. Jiang Li et al. reported iron-based catalyst which are tested at a high temperature of 240 °C for 12 h long duration and achieved only 75.3% DMF yield [61]. The present Cu/SBA-15 catalyst showed 100% HMF conversion with 90% of DMF yield within 8 h of reaction time with 2.0 MPa H2 pressure at 180 °C. This catalyst showed better yields at comparatively less H2 pressure and temperature than reported catalysts.The repeated use of catalyst was one of the key advantages of solid catalysts. Fig. 11 indicates the reusability of Cu/SBA-15 catalyst. After completion of the reaction, the catalyst was extracted from reaction mixture using centrifugation. Thus, recovered catalyst was washed with THF and dried at 80 °C before to use for next run. The Cu/SBA-15 catalyst was reused directly for the next run. The reused catalyst exhibited almost same activity after each cycle. The catalyst after being used for 5 cycles was characterized by XRD to know the structural stability and the results are shown in Fig. 12 . There was no change in its diffractogram of the used catalyst.Hydrodeoxygenation of HMF towards DMF was achieved over Cu supported on mesoporous SBA-15 catalysts. The 15%Cu/SBA-15 catalyst was active among the prepared catalysts and showed complete conversion of HMF with a maximum of 90% DMF yield. The presence of well dispersed Cu species over the support having strong interaction with SBA-15 and its acidity commutatively makes 15% Cu/SBA-15 as superior catalyst. The catalyst was selective for O–H bond cleavage and C=O bond hydrogenation, compared to the hydrogenation of the C=C bond of the furan ring. The reusability of the catalyst was examined and the catalyst was showed constant activity. The catalyst activity was also depended on the reaction conditions.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.Author D. Dhana Lakshmi acknowledge to UGC-New Delhi, India for financial support in the form of Senior Research Fellowship. We thank Director, CSIR-IICT for permitting to publish our results with a IICT communication number: IICT/Pubs./2021/183.

Selective catalytic hydrodeoxygenation (HDO) of 5-hydroxymethylfurfural (HMF) to prepare 2, 5-dimethylfuran (DMF) was studied as this product is a good biofuel. A sequence of copper dispersed on SBA-15 catalysts are designed and tested their activity for HMF hydrodeoxygenation reaction. The physico-chemical characteristics of the catalysts are gained from powder XRD, TEM, N2-adsorption desorption, NH3-TPD, H2-TPR and N2O-chemisorptions studies. Characterization results indicate the fine dispersion of Cu metal on SBA-15 with high surface area and appropriate acidic sites. The catalyst with 15%Cu on SBA-15 showed high activity towards DMF with 90% yield. The optimized reaction conditions were 180 °C of reaction temperature, 20 bar H2 pressure, and a reaction time of 8 h to achieve maximum yield. The catalyst is recyclable and exhibits consistent activity.

Hydrogen peroxide (H2O2) is one of the most important chemical compounds [1]. This environmentally friendly compound is used as an oxidizer in green chemistry due to its only by-product (H2O) harmless nature. Moreover, hydrogen peroxide has many uses in the pulp/paper industry, water purification, rocket fuel, and chemical synthesis owing to its high amount of active oxygen (47.1%) [2,3]. At present, most hydrogen peroxide is produced by the autoxidation (AO) process, involving continuous hydrogenation and oxidation of anthraquinone [4]. This procedure continuously produces hydrogen peroxide with a high concentration (45–70 wt%) and no contact between H2 and O2 under mild reaction conditions [3]. However, it is plagued with some problems such as high cost, use of toxic solvents (i.e., octanol, naphthalene), and the difficulties attributed to the separation and transportation of the product [4]. Regarding these problems and the increasing demand for H2O2 in the international market, direct synthesis of H2O2 from hydrogen and oxygen has attracted a great deal of attention as a promising substitute for the AO method, thanks to its remarkable advantages such as the smaller amount of environmental pollution and lower cost [5].The main reaction and side reactions in the direct synthesis of hydrogen peroxide are displayed in Scheme 1 . Since these reactions are characterized by negative Gibbs free energies, low selectivity of H2O2 is one of the main challenges of this process [2,6]. Therefore, many studies have been dedicated to find an efficient catalyst with high selectivity for the generation of H2O2, and many supported metallic catalysts have been tested, including Pt [7,8], Au [9–12], and Pd [13–21]. These studies have shown that the production of H2O2 was effectively performed in the presence of supported Pd catalysts. However, Pd is active not only for the production of H2O2 but also accelerates H2O generation [22,23].Consequently, various methods have been investigated to increase the selectivity of these catalysts toward H2O2. For instance, the addition of mineral acids and halides such as bromide can improve the selectivity owing to their adsorption on high energetic sites that generate H2O [24]. However, the presence of halides and acids in the reaction can lead to metal leaching, corrosion of the instruments, and deactivation of the catalyst at higher concentrations. One of the other effective methods to improve the selectivity towards H2O2 is adding of a second metal to the supported Pd catalysts. In recent years, this approach has been applied by many researchers and the effects of the other metals on the activity of the supported Pd catalysts have been extensively studied [25]. For instance, Menegazzo et al. produced H2O2 with 61% selectivity in the presence of PdAu@ZrO2 under mild conditions [26]. Freakley et al. used PdSn/TiO2 and PdSn/SiO2 catalysts and achieved a high H2O2 selectivity of up to 95% at 2 °C [27]. Gu et al. synthesized H2O2 with 70.9% selectivity using PdAg/C catalyst at 2 °C, which was higher than the obtained value over Pd/C catalyst [28]. Wang et al. prepared alumina-supported PdZn catalysts and studied the impact of Zn on the catalytic performance of these catalysts [29]. According to their work, H2O2 was produced with productivity's 25431 mol kgPd −1 h−1 and 8533 mol kgPd −1 h−1 in the presence of PdZn and Pd catalysts, respectively. Maity et al. obtained H2O2 with 95% selectivity using PdNi catalysts and attributed the improved catalytic activity to the presence of Ni [30].Besides the beneficial effects of the presence of the other metals in the structure of supported Pd catalysts on their catalytic activity to produce H2O2 with high selectivity, the nature of the support is also very effective due to its influence on the electronic structure of the metals. Accordingly, various supports have been used for this process, such as SiO2, zeolites, TiO2, and active carbon [31–35].We herein used mesoporous silica (KIT-6) as an effectual support to prepare the novel bimetallic catalysts, owing to its three-dimensional structure and large interconnected pores, facilitating the transport and diffusion of reactants/products. Then, a series of CoPd/KIT-6 catalysts with different Co:Pd molar ratios were synthesized by an impregnation method at different calcination temperatures. These catalysts were used for the direct production of H2O2, and the effect of calcination temperature as an efficient factor on their catalytic activity was investigated. The best calcination temperature was selected based on the highest H2O2 selectivity, and a series of CoPd/KIT-6 catalysts with various Co:Pd molar ratios were calcined at the desired temperature. Afterward, the optimum reaction conditions for direct synthesis of H2O2 were perused in the presence of the catalyst with the best Co:Pd molar ratio. Catalyst reusability was investigated as well. During these investigations, we give much of our attention to improve H2O2 selectivity, focusing largely on the electronic interactions of Pd in the structure of CoPd/KIT-6 catalysts.The synthesis of the CoPd/KIT-6 catalysts was performed in two steps. Firstly, mesoporous KIT-6 was prepared using the hydrothermal procedure presented by Kishor et al. [36]. Briefly, a mixture of 4.0 g P123 (Sigma–Aldrich), 144 mL distilled water, and 7.9 g HCl solution (Merck, 35%) was stirred at 35 °C. Then, 4.0 g 1-butanol (Merck, 99.9%) was added to the former homogeneous solution. After stirring for 1 h, 8.6 g of TEOS (Dae-Jung) was added to the solution and stirred for 24 h at 35 °C. The mixture was then transferred to an oven for 24 h at 100 °C, and the obtained solid was filtered and washed by ethanol (Merck, 99.9%). Afterward, the synthesized KIT-6 was dried at 100 °C for 24 h and then calcined at 550 °C for 6 h. Secondly, the incorporation of Pd and Co into the KIT-6 structure was obtained by a wet-impregnation method. CoPd/KIT-6 catalysts with different Co:Pd molar ratios (0.5:1, 1:1, and 2:1) were synthesized using PdCl2 (Merck) and CoCl2·6H2O (Merck) as the Pd and Co sources, respectively. To prepare the solutions of Pd and Co salts, 0.033 g of PdCl2 (dissolved in HCl) and a defined amount of CoCl2 (0.01, 0.02, and 0.04 g for 0.5CoPd, CoPd and 2CoPd, respectively) were separately added to 15 mL distilled water and stirred for 30 min at 80 °C. Subsequently, 0.5 g of the synthesized KIT-6 was added to 20 mL distilled water and stirred for 30 min. Then, this solution was added to the aqueous solutions of Pd and Co, followed by intense stirring for 12 h at room temperature. Then, the mixture was placed in an oven at 100 °C for 24. After synthesizing the catalysts using the above procedure, the CoPd/KIT-6 catalysts were calcined at different temperatures, namely 550, 450, and 350 °C; the corresponding obtained catalysts were named CoPd/KIT-550, CoPd/KIT-450, and CoPd/KIT-350, respectively. Also, the 0.5CoPd/KIT-350, and 2CoPd/KIT-350 catalysts were obtained via calcination of the 0.5CoPd/KIT and 2CoPd/KIT catalysts at 350 °C.All the experiments were performed using a Teflon-coated stainless-steel autoclave (volume: 60 mL, maximum working pressure: 15 MPa) equipped with a magnetic stirrer and a pressure gauge. Generally, 15 mL of 0.03 mol L−1 H2SO4/methanol solution and 2 mg catalyst were transferred to the autoclave. After that, the autoclave was purged with 10% H2 (1 MPa) for three times, filled with 10% H2 (1 MPa), and subjected to Ar to dilute H2 at the total pressure of 1.8 MPa. After stirring (1300 r min−1) for 5 min, O2 was added, and the total pressure of the reactor was increased to 2 MPa. Then, the reaction was carried out for 30 min at 25 °C. At the end of the reaction, the catalyst particles were collected, and the amount of produced H2O2 was evaluated by titration with KMnO4 (standardized with oxalic acid), while the amount of H2 was determined using gas chromatography (Teifgostar Faraz, GC-2552) equipped with a thermal conductivity detector (TCD) and Molecular-Sieve Packed Columns (6 m × 2 mm × 2 mm). Using the results of these investigations, H2O2 selectivity, H2 conversion, and H2O2 production rate were calculated as follows: (1) H 2  conversion ( % ) = mmoles of reacted  H 2 mmoles of initial  H 2 (2) H 2 O 2  selectivity  ( % ) = mmoles of produced  H 2 O 2 mmoles of reacted  H 2 (3) H 2 O 2 production rate ( mmol g − cat − 1 h − 1 ) = mmoles of produced  H 2 O 2   catalyst weight  ( g ) × reaction time ( h ) Finally, the collected particles of CoPd/KIT-350 catalyst, showing the highest yield of H2O2, were recycled in two ways: 1) the catalyst was washed with water, dried at 120 °C and reused (this catalyst was called CoPd/KIT-350 D ). 2) The catalyst was washed with water, calcined at 350 °C and reused (this catalyst was called CoPd/KIT-350 C ).To investigate the hydrogenation and decomposition reactions of H2O2, a series of control experiments were performed. For this purpose, the initial concentration of H2O2 (Merck, 35%) was 1 wt% and the reaction was carried out as described in section 2.2.1. However, decomposition tests were fed with Ar, while the hydrogenation tests were performed in an atmosphere of Ar/H2 without the presence of O2.Brunauer–Emmett–Teller (BET) surface area was determined using nitrogen adsorption/desorption isotherms at −196 °C on a PHS-1020 (PHS CHINA) apparatus. Before measurement, the samples were outgassed at 120 °C for 4 h. X-ray powder diffraction (XRD) patterns were evaluated from 2θ = 0.66–80° using an Asenware XDM-300 diffractometer by a Ni-filtered Cu Kα radiation (λ = 1.5406 Å). Transmission Electron Microscopy (TEM) analysis was performed using a Philips EM 208S microscope. Field emission scanning electron microscopy (FESEM) analysis was conducted with a QNANTA FEG 450 instrument. Also, Energy Dispersive X-ray Analysis (EDX) and elemental mapping were carried out in connection with SEM analysis. Inductively coupled plasma (ICP) spectrometry was performed using an Optima 7300 TV spectrometer. X-ray photoelectron spectroscopy (XPS) was carried out on a Kratos Axis UltraDLD spectrometer using a monochromatic Al Kα source (20 mA, 15 kV). The Kratos charge neutralizer system was used on all specimens. Survey scan analyses were carried out with an analysis area of 300 × 700 microns and at a pass energy of 160 eV. High-resolution analyses were carried out with the same analysis area and at a pass energy of 20 eV. The binding energy scale has been calibrated by setting the main line of the carbon 1s spectrum (adventitious carbon) to 284.8 eV. Spectra were analyzed using CasaXPS software (version 2.3.17). Fitting of the Pd 3d energy region was achieved using a pair of peaks for each Pd chemical state, fixing the area ratio between the Pd 3d5/2 and the Pd 3d3/2 components to 3:2 and the energy separation to 5.26 eV. In the case of Pd2+ species, symmetric peaks were used (using the GL (50) line shape in CasaXPS), while asymmetric line shape was chosen for Pd0 species (using the LA (2.5,7,10) line shape in CasaXPS).N2 adsorption–desorption isotherms of KIT and CoPd/KIT catalysts are depicted in Fig. 1 . The curve of pure KIT displays type IV isotherm with H1 hysteresis loop, indicating its high-quality mesoporous structure with uniform pore size [37]. For the other samples, the shape of the isotherms and hysteresis loops are very similar to those observed for KIT. This observation demonstrates that the addition of the metals has not affected the mesoporous structure of original support. Also, the pore size distribution illustrated in Fig. 1h clearly displays that all the samples possess narrow mesopores (between 1 and 5 nm) with a size distribution centered at ∼ 3.5 nm. The textural properties of these samples are given in Table 1 . These results show that the BET surface area and pore volume of KIT decrease with increasing the metals loading, which indicates that the modification step has been done successfully, and the metals have been inserted inside the pores of KIT. Also, the results corresponding to CoPd/KIT-350, CoPd/KIT-450 and CoPd/KIT-550 demonstrate that the increase in calcination temperature leads to the increase in BET surface area probably due to sintering of metal nanoparticles.The low-angle XRD patterns of the samples are presented in Fig. 2 a. The XRD patterns of all the cases show an intense peak at around 2θ = 1° corresponding to the (211) reflection of an ordered 3-D cubic symmetry with Ia3d space group, and another peak at 2θ = 1.1° corresponding to (220) reflections, indicating the well-ordered pore arrangement of the samples [38]. No change in the position of these peaks in the XRD patterns of CoPd/KIT-350, CoPd/KIT-450, and CoPd-KIT-550 samples confirms that the change in calcination temperature, as well as addition of the metals, has not affected the mesoporous structure of KIT. The insignificant shift corresponding to (211) reflection in the XRD pattern of 2Co/Pd-KIT-350 is probably due to the incorporation of a large amount of Co into the KIT structure.The wide-angle XRD patterns of all the samples illustrated in Fig. 2 show the broad peak corresponding to amorphous silica. For CoPd/KIT-550 sample (Fig. 2f), the diffraction peaks at 2θ = 33° (101), 41° (110), 54° (112), 60° (103) and 71° (202) are related to the presence of tetragonal PdO (JCPDS 00-041-1107) and the peaks at 2θ = 36° (311) and 65° (440) are attributed to the cubic phase of Co (JCPDS 01-078-1969). Also, the absence of any peaks corresponding to the crystalline forms of the metals in the XRD patterns of 0.5CoPd/KIT-350, CoPd/KIT-350, 2CoPd/KIT-350, and CoPd/KIT-450 samples confirms the good dispersion of the metal clusters on the mesoporous KIT.The FESEM images of KIT, Pd/KIT-350, CoPd/KIT-350, CoPd/KIT-450 and CoPd/KIT-550 samples are shown in Fig. 3 . The FESEM image of KIT (Fig. 3a) shows that this material is composed of rock-like aggregated irregular particles. However, as shown in Fig. 3b–e, this surface morphology has changed after the addition of the metals. Also, the comparison of the SEM image of CoPd/KIT-350 and those of CoPd/KIT-450 and CoPd/KIT-550 indicates that the surface morphology has been strongly influenced by the calcination temperature. Besides, the results of EDX and elemental mapping analysis of the Pd/KIT-350 and CoPd/KIT-350 samples shown in Fig. 4 confirm the existence of Pd and Co elements and their uniform distribution on the KIT surface.The topology and structural details of KIT and CoPd/KIT samples acquired by TEM analysis are shown in Fig. 5 . Fig. 5a and b reveals an ordered mesoporous structure with a bi-continuous network of channels for KIT sample, confirming the successful synthesis of this material [38]. The size distributions of the CoPd particles, obtained by TEM analysis for each of the samples, are reported in histograms in Fig. 6 . The data show that the particle size has decreased by an increase in Co amount, while this value has increased by an increase in calcination temperature.Surface compositions, electronic interactions between Pd and Co, and the chemical state of Pd species on the surface of the prepared catalysts were investigated using XPS analysis. The specimens were prepared by pressing few milligrams of the samples onto high purity indium foils (Sigma–Aldrich). Low-resolution wide scans are reported in Fig. S1 in the Supporting Information (SI) file, together with the high-resolution spectra obtained on O 1s, C 1s, and Si 2p regions (Fig. S2). The high-resolution spectra collected on the energy windows typical for Pd 3d and Co 2p peaks are depicted in Fig. 7 (for samples Pd/KIT-350, Co/KIT-350, CoPd/KIT-350, 2CoPd/KIT-350, and 0.5CoPd/KIT-350) and Fig. 8 (for samples CoPd/KIT-450, CoPd/KIT-550, CoPd/KIT-350 C and CoPd/KIT-350 D ); the results are summarized in and Table 2 . As shown in Fig. S1, the intensity of Pd 3d and Co 2p peaks is intensity extremely low, and long acquisition times were needed to obtain the spectra reported in Figs. 7 and 8. Here, the Pd 3d XPS spectra of the different bimetallic catalysts show two signals in the 334–344 eV range, related to the Pd 3d5/2 and Pd 3d3/2 components. After deconvolution, two contributions are observed for Pd. The peaks at (336.7 ± 0.2) eV and (342.0 ± 0.2) eV can be assigned to Pd2+ species, while the asymmetric peaks centered at (334.9 ± 0.2) eV and (340.2 ± 0.2) eV are due to the presence of Pd0 species [39]. The Co 2p XPS spectra of all the catalysts suggest the presence of Co+2 species, as each spectrum is characterized by peaks at 780, 782, 785 and 790 eV [40].As stated above, the presence of two palladium species was revealed in all catalysts. It has to be noted that X-ray induced reduction of high valence Pd compounds has been reported, and cannot be entirely excluded in the present case, also in consideration of the long acquisition time needed [41]. However, Pd2+/Pd0 ratios presented in Table 2 suggest that the oxidation state of Pd atoms might have been affected by the amount of Co in the samples. Indeed, when the Co loading increases from 0.6 to 2.0 wt%, the concentration of Pd2+ increases as well, probably due to the electron transfer from Pd to Co [42,43].The results corresponding to CoPd/KIT-350, CoPd/KIT-450, and CoPd/KIT-550 samples show that the different calcination temperatures affect the relative concentration of Pd2+ and Pd0. As shown in Table 2, when the calcination temperature rises from 350 to 450 °C, the Pd2+/Pd0 ratio decreases from 2.65 to 2.59. However, a further increase in the calcination temperature to 550 °C leads to the highest Pd2+/Pd0 ratio value (17.69). Table 2 also shows the results of XPS analysis for CoPd/KIT-350 C and CoPd/KIT-350 D samples. These results demonstrate that the concentration of Pd2+ in both samples is higher than that of Pd0, with a higher Pd2+/Pd0 ratio for CoPd/KIT-350 C . This observation suggests that calcination of the catalyst after the reaction can accelerate the oxidation of Pd.The results of the catalytic activity of the catalysts in the direct synthesis of H2O2 from H2 and O2 are listed in Table 3 . The results corresponding to Co/KIT-350 and Pd/KIT-350 catalysts show that Co is inactive in the direct synthesis of H2O2, while the reaction is performed in the presence of Pd and results in the production of H2O2 with 26% selectivity. Moreover, the results of the catalytic activity of CoPd/KIT catalysts reveal that the addition of Co enhances the catalytic activity of the Pd catalyst. As observed, with an increase in the molar ratio of Co:Pd from 0.5:1 to 1:1, the catalytic activity increases and reaches the highest level of H2O2 selectivity of 50%, H2 conversion of 51% and H2O2 productivity of 520 mmol catalyst g−1 h−1. On the other hand, the results of the catalytic activity of CoPd/KIT-350, CoPd/KIT-450, and CoPd/KIT-550 catalysts show a trend of decreasing the H2O2 productivity along with the increase in calcination temperature. This is probably due to the decrease in the number of catalytic active sites or to the decrease in the concentration of the metals on the surface of the catalyst. Table 3 also shows the results corresponding to the catalytic activity of CoPd/KIT-350 C and CoPd/KIT-350 D . It is obvious that the catalytic activity of these catalysts has reduced compared to that of CoPd/KIT-350, most likely due to slight leaching of the metals from the support during the reaction. However, a comparison between the obtained results for CoPd/KIT-350 D and CoPd/KIT-350 C catalysts shows that calcination of the catalyst after the reaction results in an increase in the stability of the catalyst.Also, to investigate the effect of calcination temperature on the catalytic activity of the catalysts upon re-use, the catalysts were dried at 120 °C after the reaction and reused in the reaction of H2 with O2. The results of this investigation are summarized in Table 4 . The catalytic activity of all the catalysts decreases after the reaction due to the leaching of the metals as discussed above. However, this leaching decreases with an increase in the calcination temperature. Also, the results corresponding to the CoPd/KIT-350 catalyst which was calcined at 350 °C after the reaction reveals that calcinating the catalyst can positively affect the catalytic activity of the catalyst upon re-use.Decomposition and hydrogenation of H2O2 are the main reason for the loss of selectivity in the direct synthesis of H2O2 [44]. Therefore, to investigate the catalytic activity of the catalysts in these reactions, the decomposition and hydrogenation experiments were performed in H2O2/methanol solutions under Ar and H2/Ar atmospheres, respectively. The results of these experiments are illustrated in Fig. 9 . These data show that the addition of Co to Pd/KIT-350 decreases the rate of the decomposition and hydrogenation reactions in the direct synthesis of H2O2. On the other hand, as shown in Fig. 9, different calcination temperatures have significantly affected the activity of the catalysts for decomposition and hydrogenation of H2O2 and the activity of the catalysts has decreased at higher calcination temperatures. According to the results presented in Table 3 and Fig. 9, it can be concluded that the addition of Co to the Pd catalyst and different calcination temperatures change the catalytic activity of the catalysts not only for H2O2 production but also for H2O2 decomposition and hydrogenation.For the direct synthesis of H2O2, it is well known that the interaction of molecular O2 and atomic H leads to the formation of H2O2, while O2 dissociation results in the production of H2O [45]. Therefore, inhibiting the O–O bond scission both in the initial mixture and in the product is a key factor in the selective formation of H2O2. Previous studies, including theoretical calculations and experimental works, have evidenced that the addition of a second metal to the Pd catalyst lowers the catalytic activity toward the breaking of the O–O bond by a change in the electronic interactions of Pd, creation of inactive sites for product decomposition, blocking Pd's active sites and increasing the Pd monomer sites [15,46,47].However, our results show that the introduction of Co into the Pd catalyst has not affected the electronic interactions of Pd significantly (Table 2, entries 1, 3, 4, and 5), while the H2O2 selectivity has increased from 26 to 49% (Table 3, entries 3, 4, 5, and 6). This observation confirms that the selectivity increment is more associated with the other factors mentioned above.Alternatively, a closer look at Table 3 (entries 5, 7, and 8) shows that calcination temperature can change the H2 conversion and H2O2 selectivity, as well. With regard to the effect of calcination temperature on the catalyst features, this change can be attributed to some factors such as a change in the electronic interactions of Pd (Table 2, entries 4, 6, and 7), an increase in the specific surface area (Table 1, entries 3, 5, and 6) and the particles size (Fig. 6) of the catalyst and phase formation in the structure of the catalyst (Fig. 2e and f). Therefore, to investigate the contribution of electronic interactions of Pd, the CoPd/KIT-550 catalyst, characterized by the highest Pd2+ content (Pd2+/Pd0 ratio = 17.69), was exposed to H2 for 18 h at 250 °C and used in the reaction of H2 with O2 under optimized reaction conditions. At the end of the reaction, the conversion increased from 35% to 47% due to the decrease in the amount of Pd2+. This observation implies that the change in the electronic interactions of Pd caused by an increase in calcination temperature has a considerable effect on this reaction.Although our results confirmed the effects of the Co addition and calcination temperature on the electronic interactions of Pd in the selective formation of H2O2, it should be noted that the presence of H2 and O2 in the reaction can also change the electron density of Pd. The comparison of XPS analysis for CoPd/KIT-350 and CoPd/KIT-350 D (Table 2) clearly shows that the presence of H2 and O2 in the reaction has led to a change in the concentrations of Pd0 and Pd2+ and as a result, the H2O2 selectivity has been altered. To prove that the decrease in H2O2 selectivity results from the change in the electron density of Pd, CoPd/KIT-350 C was used in the reaction of H2 with O2 under optimized reaction conditions. The results of this experiment (Table 2, entry 8 and Table 3, entry 9) show that the H2O2 selectivity increases again when the electron density of Pd approximately returns to the initial state (Table 2, entry 4 and Table 3, entry 5).From the above results, we could conclude that the electronic interactions of Pd can be one of the important factors affecting the H2O2 selectivity. However, the active oxidation state of Pd, which is responsible for the selective generation of H2O2, is still unknown. Some groups have claimed that Pd0 plays a more critical role than Pd2+ toward the selective formation of H2O2, and several other groups have confirmed that Pd2+ is more active in this reaction [15,22,28,29,48–53]. However, during our investigation, we found a relation between selectivity and Pd2+ and between conversion and Pd0. The results of these investigations are as follows:The results presented in Tables 3 and 4 are summarized in Fig. 10 to visualize better the relations between the catalytic activity of the catalysts and the active oxidation state of Pd. As shown in Fig. 10c, the curves corresponding to H2O2 selectivity and Pd2+ content show a similar trend. This observation indicates that Pd2+ possesses the higher activity for H2O2 formation due to its particular electronic and geometrical structures which would lead to the weak interaction of Pd atoms with O2, OOH, H2, and H2O2 adsorbents. As a result, the dissociation of these adsorbents decreases, and the selectivity of H2O2 as the main product increases [22]. However, according to the mechanism of this reaction, decreasing the activity of the catalyst for breaking the H–H bond is undesirable due to the decrease in H2 conversion (Fig. 10b). Therefore, H2 conversion on Pd2+ is less active than that on Pd0. The relation between Pd0 content and H2 conversion is also shown in Fig. 10a. As observed, H2 conversion increases by increasing the content of Pd0.The above results were confirmed by the final results of this reaction in the presence of CoPd/KIT-350 D and CoPd/KIT-350 C catalysts. For CoPd/KIT-350 D catalyst, the content of Pd2+ decreased after the reaction, probably due to the presence of O2 and H2 at the high pressure in the reaction and as a result, the H2O2 selectivity decreased (Fig. 10c) and H2 conversion increased (Fig. 10b). However, for CoPd/KIT-350 C catalyst, the content of Pd2+ increased due to the calcinating the catalyst after the reaction. Consequently, H2O2 selectivity increased, and H2 conversion decreased in the presence of this catalyst (Fig. 10c and b).Since calcinating the catalyst after the reaction showed to be advantageous for the direct synthesis of H2O2 (Section 3.2.1), the CoPd/KIT-350 C was selected and used for the reusability test. For this purpose, the catalyst was calcined at 350 °C after each run and re-used following the procedure stated in Section 2.2.1. The results are depicted in Fig. 11 . As can be seen, at the end of the second use, a significant loss of activity is observed, most likely due to the leaching of Pd atoms from the catalyst. However, no significant change in catalyst activity is seen in the next three successive runs. These results suggest that CoPd/KIT-350 catalyst has good stability for the direct synthesis of H2O2.In this work, the catalytic activity of the novel CoPd bimetallic catalysts supported on mesoporous KIT-6 was studied for the direct synthesis of H2O2 from H2 and O2. These catalysts were prepared using various Co:Pd ratios at different calcination temperatures and characterized with various analyses. By combining the results of the characterizations and the obtained catalytic activity, we determined that Co addition and calcination temperature have affected the electronic interactions of Pd as well as morphology, particle size, and BET surface area of the catalysts. Considering these changes in the structure of the CoPd/KIT catalysts, the electronic behavior of Pd during the reaction was investigated, and the results demonstrated that Pd2+ was the active phase for selective formation of H2O2 and could provide the higher selectivity for H2O2. In contrast, Pd0 resulted in low activity for H2O2 selectivity and increased the H2 conversion. As a result, the existence of Pd2+ alone in the structure of the catalyst is not enough to improve the process of direct synthesis of hydrogen peroxide because the conversion of H2 is dependent on the existence of Pd0. Also, the reusability experiments showed that calcinating the catalyst after the reaction increases the stability of the catalyst. Therefore, the results of this work not only provide novel insights into the catalyst design but also show that CoPd/KIT-350 is an efficient bimetallic catalyst for the selective generation of H2O2.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.We are grateful for the financial support (Research Council Grant) provided by Isfahan University of Technology (Iran).The following is the Supplementary data to this article: Multimedia component 1 Multimedia component 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.gee.2021.03.014.

A series of CoPd/KIT-6 bimetallic catalysts with various Co:Pd molar ratios at different calcination temperatures were prepared and used for the direct synthesis of H2O2 from H2 and O2. These catalysts were characterized by nitrogen adsorption–desorption, low and wide-angle X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), transmission electron microscopy (TEM), scanning electron microscopy (SEM), elemental mapping and energy-dispersive X-ray (EDX) methods. It was found that the particle size, electronic interactions, morphology, and textural properties of these catalysts as well as their catalytic activity in the reaction of H2 with O2 were affected by Co addition and different calcination temperatures. Also, the results showed that while the H2O2 selectivity depends on Pd2+ species, the H2 conversion is related to Pd0 active sites. Among these catalysts, CoPd/KIT-6 calcined at 350 °C (CoPd/KIT-350 catalyst) showed the best catalytic activity with 50% of H2O2 selectivity and 51% conversion of H2.

Now-a-days fuel cells are contributed as highly efficient and eco-friendly energy sources [1,2]. It converts the chemical energy into electrical energy by anodic fuel (ethanol/methanol) oxidation and cathodic oxygen reduction with very little environmental pollution [3]. Still, now it is challenging to use the same catalyst for both anode and cathode, which enhance the electro catalytic activity and durability in fuel cell [4, 5]. Generally platinum (Pt) and platinum related catalyst are mostly used for both anodic oxidation and cathodic reduction reaction [6–8] but due to limitation of Pt supply, high price, sluggish kinetics in cathode reaction (ORR), it is very important to use cost-effective some other non-Pt catalyst for large scale commercial application in fuel cell [9, 10]. In recent years, most of the researcher has been tried to decrease the use of expensive Pt based catalyst and increase the inexpensive metal or metal free catalyst for fuel cell [11–14]. Palladium nanoparticles (PdNPs) can be used instead of platinum because they are same group metals in the periodic table, cheaper than platinum and more abundant in the earth [15–18]. The catalytic properties of PdNPs can be improved by dispersing the PdNPs on carbon based support materials [19, 20]. It has been reported that the metal nanoparticles supported on carbon nanomaterial shows an improved catalytic activity [1, 21].Carbon nanotubes (CNTs) are one of the most promising carbon nano-materials for contributing in modern science [22, 23]. It possesses many fascinating physical properties such as, excellent mechanical strength, porous structure, electrical and thermal conductivities [22,23]. Due to its unique properties, CNT-based nanocomposites have been applied for a wide range of applications, such as energy conversion devices, batteries, supercapacitors, sensors and fuel cells [12, 24–27]. Moreover, Functionalization of CNT with some other hetero atoms (N,B, S and P) is an effective way to tune their intrinsic properties.Among them, nitrogen is the most effective dopants because of their similar small atomic size to that of the carbon atom and can be easily inserted in carbon nanotube frame [28–30]. Doping of CNT with nitrogen atoms can improves the physico-chemical properties due to conjugation between the lone-pair electrons of nitrogen and the π system of CNT [12]. It may improve the electro catalytic activity by providing the nucleation sites of metal nanoparticle and changing the electronic structure of the catalyst. Moreover, the incorporation of nitrogen atom on to CNT support may enhance the chemical binding energy between CNT and metal nanoparticles which is also favorable to increase the durability of the catalyst [30–32].Herein, we have synthesized nitrogen-functionalized carbon nanotube supported palladium nanoparticles (NCNT-Pd) via simple chemical reduction method at room temperature without using any surfactant (scheme 1 ). Transmission electron microscopy (TEM), X-ray photoelectron spectroscopy (XPS), X-ray diffraction (XRD) and Energy dispersion spectroscopy (EDS) analyses are performed to investigate the structural and morphological characteristics of the as synthesized catalyst, such as size and dispersion of the particles, crystallinity, and percentages of element. Then the NCNT-Pd was applied for oxygen reduction reaction in alkaline media by using cyclic voltammetry and hydrodynamic voltammetry. Furthermore, the catalytic performances of NCNT-Pd also investigated for electro-oxidation of ethanol in alkaline media by using cyclic voltammetry (CV). From the result, it is found that the NCNT-Pd catalyst not only showed higher catalytic activity for ORR but also showed excellent EOR in alkaline media. For comparison, we also synthesize CNT-Pd and NCNT by the same procedure.The pristine CNT was prepared by Carbon Nano Tech Co., Ltd. (Pohang, South Korea). Potassium tetrachloropalladate (99%) and hydrazine mono hydrate (65%) were purchased from sigma Aldrich. Methanol and ethanol (99.9%) were obtained from OCI Co., Ltd .All other reagents were analytical grade and used without further purification. Deionized (DI) water was used for preparing the electrolyte solution.The pristine carbon nanotubes (CNTs) were treated with a mixture of concentrated H2SO4/HNO3 (3:1, v/v) at 100 °C for 2 h to oxidize the surface of CNT [25,33]. The obtained product (CNT–COOH) was filtered and washed with deionized (DI) water and dried in vacuum oven at 50 °C for 6 h. Then 10 mg of CNT-COOH + 30 mL DI water was taken in a 100 mL round bottle flask and ultrasonicated for 1 hour. Then 100 μL hydrazine monohydrate (N2H4•H2O) followed by 30 mL ice cold sodium borohydride (0.074 mg/mL) was added under vigorous stirring. After that 15 mL (0.33 mg/mL) of potassium tetrachloropalladate (II) (K2PdCl4) was added drop wise and stir it for 20 h at room temperature. Finally, the black product was filtered and washed for several times with DI water. As prepared nitrogen-functionalized carbon nanotube supported palladium nanoparticles (NCNT-Pd) was then dried at 60 °C under vacuum condition for 24 h. For comparison, NCNT and CNT-Pd also prepared by the same method.The transmission electron microscopy (TEM) and energy dispersive X-ray spectroscopy (EDS) were takenon TECNAI model FI-20 microscope at 200 kV by using a copper (Cu) grid. X-ray diffraction (XRD) patterns of the catalyst were carried out with Rigaku micro-area high-resolution X-ray diffractometer using Cu, Kα radiation λ=1.5405Ao. The size and d -Spacing of Pd NPs were estimated from TEM images and XRD patterns by using image J and origin 9.1 software. X-ray photoelectron spectroscopy (XPS) were investigated by using a Multi-Lab 2000 with a monochromatic 14.9 keV Al Kα X-ray source. All electrochemical applications were investigated via cyclic voltammetry and hydrodynamic voltammetry with a CHI 700C (USA) potentiostat. The three electrode half-cell system assembled with a glassy carbon working electrode (GCE area: 0.0707 cm2), a Pt wire auxiliary and an Ag/AgCl reference electrode (Bioanalytical Systems Inc. 3 M NaCl) respectively. For hydrodynamic voltammetry measurements, a rotating ring disk electrode (disk area: 0.196 cm2; platinum ring area: 0.041 cm2) was used as an RRDE electrode. Electrochemical impedance spectroscopy (EIS) was investigated by using Versa STAT3 (Princeton Applied Research, USA) in three electrode cell system. All electrochemical experiments were conducted at 25 °C in 0.1 M KOH solution which is saturated by Ar and O2 for 30 min.XPS analysis was carried out to investigate the elemental composition and electronic structure of the as prepared NCNT-Pd catalyst. The survey spectrum of NCNT-Pd, NCNT and CNT (Fig. 1 (a)) reveals the presence of carbon (C1s, 284.08 eV), palladium (Pd 3d, 335.08 eV), nitrogen (N1s, 401.07 eV) and oxygen (O1s, 534.07 eV) which have similarities with the previous literatures [16,25] Fig. 1.(b) shows the XPS spectrum of C1s for CNT and NCNT-Pd. It is clearly seen that, four different peaks are present in C1s spectrum which is corresponding to the sp2 carbon (284.78) eV and oxygenated carbon (CO; 285.07 eV C = O; 286.09 eV; OC = O eV; 287.28 eV) [16, 25, 28, 34]. The relative intensity of sp2 carbon is higher than the oxygenated carbon that means percentages of sp2 carbon is higher than the oxygenated carbon which plays an important role for ORR [35]. Moreover the peak shifting of C1s of NCNT-Pd to the higher binding energy compared to CNT indicates that strong interaction of metals in graphitic system of CNT [36]. The XPS spectra of N1s is deconvoluted in one main peak which is located at 400.08 eV [28, 37]. The N1s peak of NCNT-Pd (Fig. 1(c)) suggested that carbon nanotube successfully functionalized by nitrogen and the percentages of N content is 2.15 wt%. The high resolution of Pd 3d spectrums are shown in Fig. 1(d). There are two well-defined Pd peaks are corresponding to the Pd 3d5/2 (335.1 eV) and Pd 3d3/2 (340.5 eV) which has similarities with the previous literature values [16, 38]. The Pd3d5/2 signals give three peaks at 335.1, 336.2, and 337.2 eV, which can be ascribed to Pd0, PdO, and PdO2 states,respectively [38, 39]. The content of Pd in NCNT-Pd catalyst is 7.24 wt%.The TEM images of NCNT-Pd are shown in Fig. 2 (a-c). From the TEM images, it can be clearly seen that the Pd nanoparticles are uniformly decorated on the most of the outer surface of NCNTs. The average diameter of the decorated Pd NPs on NCNT surface is about 7.22 nm which increase the surface area of the catalyst. The interplanar d -spacings for the lattice fringes of Pd were 0.230 nm and 0.195 nm, which is corresponding to the (111) and (200) lattice planes of the face-centered cubic (fcc) Pd structure (Fig. 2(c)) [15, 16]. The selected-area electron diffraction (SAED) pattern (Fig. 2(d)) was taken from single Pd NPs which is given five different rings corresponding to the five different crystal plane on the Pd NPs. The observed five different ring represents the (111), (200), (220), (311) and (222) planes of the fcc Pd, indicating the polycrystalline structure [40, 41]. EDS mapping is conducted to analyses the elemental distribution of NCNT-Pd. As shown in Fig. 2(e), only C, O, N and Pd are present in NCNT-Pd without any other species.XRD patterns of the NCNT-Pd are shown in Fig. 2(f). The XRD peak at 2θ=25.4° corresponds to the C(002) peak of NCNT [12, 42]. In addition there are five diffraction peaks at 2θ of 40°, 46°, 68°, 82° and 87° which can be ascribed to (111), (200), (220), (311) and (222) crystal plane of face centered cubic (fcc) Pd, respectively [16, 43] and these results have similarities with the SAED pattern. The average particle size of the prepared NCNT-Pd nanoparticles (d) was calculated by using the Scherrer Eq. (1) [43]. The diffraction peaks of Pd (111) at 2θ of 40° were used to calculate the size of particles (d). (1) d = k λ b C o s θ Where, k is a constant (0.9), λ is the wavelength of X-ray (0.15405 nm), b is the full width at half-maximum (FWHM) of the (111) diffraction peak (in radian) and θ is the angle of the maximum peak position. The average size of the Pd nanoparticles was estimated from the XRD results was 7.46 nm for NCNT-Pd which is similar to the TEM analysis.The electrochemical impedance spectroscopic (EIS) experiment are conducted to evaluate the electrical conductivity and charge transfer behavior of the electrodes. The EIS experiments were carried out by inserting the electrode in a solution of 0.1 M KCl containing 1.0 mM K3Fe(CN)6 and K4Fe(CN)6 (1:1) at a frequency ranges from 105 to 10−2 Hz. The charge-transfer kinetics can be estimated from the intercept region and shape of the impedance spectrum. The Nyquist plots obtained from EIS of bare GCE, NCNT, CNT-Pd and NCNT-Pd catalyst modified electrodes are shown in Fig. 3 (a). In the Nyquist plots, the intercept on the real axis at high-frequency region represents the electrolyte resistance (Rs) and the diameter of the semicircle portion at low-frequency region represents the electron-transfer impedance between electrolyte and electrode interface (Rct) [44, 45] . The Rs value of NCNT-Pd (SBH), CNT-Pd, NCNT and bare GCE are 110, 115, 130 and 137 Ώ respectively and the Rct values for them are 115Ω, 124Ω, 1.25 kΩ and 7.5 kΩ for NCNT-Pd, CNT-Pd, NCNT and bare GCE respectively. From the EIS it is found that NCNT-Pd shows lowest Rs and Rct value which is an indication of higher conductivity low charge transfer resistance of the NCNT-Pd electrode [44, 45]. The results indicate that the NCNT-Pd modified electrode was highly conductive with excellent catalytic activity at the interface than the other electrodesThe electro-catalytic activities of NCNT-Pd and other three catalysts (CNT-Pd, NCNT and 20% commercial Pt/C) toward ORR were first investigated by cyclic voltammograms (CV) in 0.1 M KOH electrolyte (saturated by O2 gas) at a scan rate 50 mVs−1 (Fig. 3(b)). For the NCNT-Pd, CNT-Pd and 20% Pt/C samples, peak I corresponds to hydrogen desorption and/or oxidation in this region [46]. A well-defined oxygen reduction peak started at −0.087 V (inset figure) is observed for O2-saturated electrolyte indicating that NCNT-Pd has an efficient catalytic activity for ORR in alkaline media. From the forward scan, peak II, which emerges at a potential of −0.4 V, correspond to the formation of Pd oxides and partially overlaps with the oxygen reduction peak. Peak III, centered at −0.33 V, can be attributed to the reduction of Pd oxide during the reduction sweep [16, 47]. Following the reduction sweep, peak IV (−0.8 V to −1.0 V) corresponds to the adsorption of hydrogen [46].There are some possible reasons for excellent ORR activity of NCNT-Pd. First, the functionalized N atoms are electron rich which breaks the electro-neutrality of sp2 carbon of CNT and produces charge sites which enhance the ORR activity by increasing oxygen adsorption [35]. Second, the hetero N atoms enhance the nucleation and uniformly dispersion of PdNPs on NCNT surface [16]. Third, the homogeneously disperse Pd NPs are increased the surface area of NCNT-Pd catalyst which is also favorable for oxygen reduction [1, 16].To further investigate the ORR kinetics of NCNT-Pd, CNT-Pd and NCNT modified electrode, we have conducted rotating disk electrode (RDE) measurement in 0.1 M KOH electrolyte Fig. 4.(a-c) represents the RDE voltammograms of NCNT-Pd, CNT-Pd and NCNT modified electrode at various rotation rate. It can be seen that catalytic current density increases with increasing rotation rate because of oxygen diffuse to catalyst modified electrode and reduced to directly OH by maximum current transfer [48]. Fig. 4(a′-c′) shows the Koutecky–Levich plots (J  −   1  vs. ω−1/2 ) at various electrode potential (data are taken from Fig. 4(a-c). From the slope of Koutecky–Levich plots (K-L plots), we can calculate the electron transfer number by using Koutecky-Levich Eq. (2) & (3) [48]] (2) 1 J = 1 J k + 1 j L = 1 J k + 1 B ω 1 2 (3) B = 0.62 nFA D o 2 3 V − 1 6 C 0 Where, J is the measured current density, Jk and JL are the kinetic and diffusion limiting current density, ω is the angular rotation rate of the electrode, n is the electron transferred number, F is the Faraday constant(F = 96,485 C mol−1), A is the geometric electrode area (cm2). K is the rate constant of the reaction, Co is the bulk concentration of O2 (1.2 × 10−6 mol cm−3), DO is the diffusion coefficient of O2 in the 0.1 M KOH solution (1.73 × 10−5 cm2 s − 1) and V is the kinetic viscosity of the electrolyte (1 × 10−2 cm2 s − 1) [12, 16, 48].The linearity and parallelism of the K-L plots indicate first-order reaction kinetics and similar electron transfer numbers for ORR at various potentials [49, 50]. The average electron transfer number which is calculated from the slope of Koutecky–Levich plots is ∼3.94 at −0.6 to −1.2 V, indicating NCNT-Pd catalyst proceed ORR process via 4e- transfer pathway as like as 20% commercial Pt/C catalyst.To quantify the production of peroxide species (HO2 −) during the ORR process, rotating ring disk electrode (RRDE) measurement are performed in O2-saturated 0.1 M KOH electrolyte [51] Fig. 5.(a) represents the comparison of RRDE measurements (at 2500 rpm) of NCNT-Pd, CNT-Pd, NCNT and 20% commercial Pt/C. It is found that NCNT-Pd shows higher disk current density (Fig. 5(a-b)) and lower ring current density than the others two catalysts (CNT-Pd and NCNT) and also 20% commercial Pt/C. In RRDE, O2 defuse to disk electrode and reduced as well as generated peroxide species which is detected by ring electrode. It is well known that the generation of HO2 − species during ORR process is not desirable because it decreases the efficiency of the catalyst and also can lead to the deterioration of it [52]. Our NCNT-Pd catalyst shows efficient catalytic activity with higher disk current and lower ring current that means the negligible amount of peroxide species produces during the ORR process. The peroxide species formation (% HO2 −) and transfer electron number (n) are calculated by the following equations (4 &5) [52, 53]: (4) % of HO 2 − = 200 × i r / N i d + i r / N (5) n = 4 × i d i d + ( i r / N ) Where, Id is disk current, Ir is ring current and N is current collection efficiency which was determined to be 0.37 from the reduction of K3Fe[CN]6.The calculated peroxide species (HO2 −) are below ∼10% for NCNT-Pd and over the potential range of −0.4 to −1.2 V (Fig. 5(d)). Also the calculated electron transfer number is ∼ 3.90 over the same potential range (Fig. 5(c)). This result has similarities with 20% commercial Pt/C electrode (Fig. 5(c-d)), that means NCNT-Pd is a promising cathode catalyst for the alkaline fuel cell. The summary of the ORR performance on NCNT-Pd, CNT-Pd, NCNT and 20% commercial Pt/C electrodes (2500 rpm) were enlisted in table 1 . The values NCNT-Pd are not only similar to that 20% commercial Pt/C, but also that of most reported Pt-based catalysts as presented in table 2 .Methanol crossover effect is one of the major concern to justify the fuel cell catalyst. chronoamperometry is used to show the methanol tolerance effect of NCNT-Pd and 20% commercial Pt/C in O2-saturated 0.1 M KOH at a constant voltage of −0.3 V (Fig. 6 (a)). As shown in Fig. 6(a), after injection of 0.1 M methanol a sharp current response observed for Pt/C electrode which is an indication of methanol oxidation reaction and CO poisoning effect. In contrast, NCNT-Pd shows a tiny current response on methanol injection, demonstrating a better methanol tolerance than the 20% commercial Pt/C.Furthermore, we also conducted stability test for NCNT-Pd and 20% commercial Pt/C by using chronoamperometric measurements in O2-saturated 0.1 M KOH at a constant potential of −0.3 V. As shown in Fig. 6(b), NCNT-Pd exhibits 84% retention current after 10 000 s. In contrast, 20% commercial Pt/C exhibits only 40% retention current after the same time. From the above result, it is clear that NCNT-Pd shows excellent methanol tolerance and better durability towards ORR.The cyclic stability of NCNT-Pd was also investigated in O2-saturated 0.1 M KOH electrolyte by cycling the catalyst between −0.2 V and 1.2 V (Fig. 6(c)). After 500 continuous cycles, the NCNT-Pd modified electrode showed a very little decrease in half wave potential (∼15 mV) along with very small decline in limiting current density indicating that the NCNT-Pd catalyst also has great advantages in terms of cyclic stabilityNCNT-Pd catalyst is also tested for ethanol oxidation reaction (EOR). The electrocatalytic performances for EOR are investigated by CV measurements Fig. 7.(a) represents the CV curves of NCNT-Pd, CNT-Pd and commercial Pd/C in Ar-saturated1M KOH electrolyte in the absence of ethanol at scan rate 50mVs−1.A s shown in Fig. 7(a), the cathodic peak at around −0.4 V is ascribed to the reduction of PdO [16, 47]. The electrochemically active surface areas (ECSA) can be calculated from the reduction peak of PdO by using the equation ECSA= Q/(0.405 × mPd), where Q is the integral columbic charge (mC) of the reduction area of PdO, 0.405 is the charge required for the reduction of PdO on catalyst and mPd is the Pd loading (g) on the electrode [54]. From the Fig. 7(a), it is clear that NCNT-Pd shows higher PdO peak than other two catalyst that means NCNT-Pd has more active surface area. The calculated ECSA values of NCNT-Pd, CNT-Pd and Pd/C are 74.7 m2gpd −1, 52.1 m2gpd −1 and 38.2 m2gpd −1 respectively. The NCNT-Pd shows the largest surface area which is 1.43 and 1.85 times higher than the CNT-Pd, and commercial Pd/C suggesting doping N in CNT plays an important role for more nucleation of Pd NPs on NCNT.The electrocatalytic performances for ethanol oxidation are investigated by CV in Ar-saturated 1 M KOH +0.1 M ethanol solution at 50 mVs−1 scan rate (Fig. 7(b)) Fig. 7.(b) represents the comparison of ethanol oxidation on NCNT-Pd, CNT-Pd, and commercial Pd/C electrode, respectively. As seen from Fig. 7(b), there are two well defined anodic oxidation peaks are observed for the forward and backward scan towards ethanol oxidation reaction. The oxidation peak in the forward scan is attributed to the oxidation of ethanol and the backward scan peak is produced by oxidation of remaining carbonaceous species which are not fully oxidized in the forward scan [55] [56]. The anodic oxidation peak current density of EOR on NCNT-Pd is 10.3 mA cm−2 which is 2.54 and 5.2 times higher current density than the CNT-Pd and commercial Pd/C. Moreover, the onset potential (in the forward scan) of NCNT-Pd for EOR is around −0.72 V which is shifted to more negative side than the CNT-Pd (−64 mV) and commercial Pd/C (−60 mV) indicating faster reaction kinetics for ethanol oxidation on the NCNT-Pd with lower over-potential. These results suggest that NCNT-Pd catalyst shows higher catalytic activity towards EOR than CNT-Pd and commercial Pd/C.With regard to the catalytic mechanism of EOR by Pd-based catalysts on ethanol in alkaline media can be represented in Eqs. (6)-(10) [47, 57] (6) M + O H − ↔ M − O H ads + e − (7) M+ CH3CH2OH ↔ M-(CH3CH2OH)ad (8) M-(CH3CH2OH)ads + 3OH−→ M-(CH3CO)ads + 3H2O + 3e− (9) M-(CH3CO)ads + M-OHads → M-CH3COOH + M (10) M-CH3COOH + OH− → M + CH3COO−+ H2O Where, M is the Pd based catalyst. The main intermediate species are CH3CO and OH produces during the reaction and OH facilitates the desorption of CH3CO releasing acetate as the main product [47] Furthermore, the stability of the catalysts is studied by the chronoamperometric method in 0.1 M CH3CH2OH + 1 M KOH solution at a constant potential −0.4 V Fig. 8.(a) represents the stability of different catalysts for ethanol oxidation reaction and it is clear that the current density rapidly decreases with the time and after 100 s they decay slowly and attain steady state current density. The initial and steady state current density of NCNT-Pd is higher than the CNT-Pd and commercial Pd/C, indicating the excellent tolerance of intermediate carbonaceous species and higher catalytic activity of NCNT-Pd towards EOR.To evaluate the cyclic stability of NCNT-Pd for EOR, cyclic voltammograms have been cvonducted for ethanol electro-oxidation reaction in 1 M KOH containing 0.1 M C2H5OH electrolyte by cycling between −0.4 V and 1.0 V at scan rate 50 mVs−1 (Fig. 8(b)). After 500 continuous cycles, the NCNT-Pd modified electrode showed around similer onset potential and a very small decrease in current density indicating that the NCNT-Pd catalyst also shows superior cyclic stability in EOR.The electrocatalytic parameters measured from the above voltammograms for EOR have been tabulated (table 3 .). From the table 3, it can be seen that electroactive surface area of NCNT-Pd higher than that of CNT-Pd and Pd/C which is favorable for the high portion of ethanol absorption and enhances EOR. Also, NCNT-Pd shows more negative onset potential and higher current density than CNT-Pd and commercial Pd/C that means NCNT-Pd can be also used as an efficient catalyst for ethanol oxidation. A comparison of EOR at different material modified electrodes can be observed in table 4 . It is clear that the NCNT-Pd also shows higher catalytic activity toward EOR than other materials in terms of onset potential, ECSA and current density.The above ORR and EOR results have demonstrated that NCNT-Pd shows higher catalytic performance toward ORR and EOR. Some key factors are responsible for this performance, first, hetero atom electron rich which plays an important role for Pd NPs nucleation and results in more surface area and enhance ORR and EOR [16, 32]. Second, the lone pair of N atom breaks the elctroneutrality of sp2 carbon and keeping enough pi electron on CNT network for conduction [16]. Third, due to strong interaction NCNT and Pd, NCNT-Pd shows excellent electrocatalytic activity for both ORR and EOR.Nitrogen functionalized NCNT supported Pd NPs were synthesized by simple chemical reduction method at room temperature. The elemental and morphological properties have analyzed by XPS, XRD, TEM, and EDX. From the TEM images, it is revealed that small size Pd nanoparticles are successfully synthesized on NCNT. The as prepared NCNT-Pd is applied for ORR and EOR in alkaline media. The electrochemical studies for ORR reveal that, NCNT-Pd shows more positive onset potential and higher current density than the CNT-Pd and NCNT. The RRDE analysis demonstrated that ORR proceeds by ∼4-electron transfer pathway with the negligible amount of peroxide species as like as commercial 20% Pt/C. Moreover, NCNT-Pd shows good catalytic performance toward EOR with higher peak current density and remarkable negative shift of onset potential. From the above discussion, it is clear that NCNT-Pd shows higher catalytic activity both ORR and EOR in alkaline media. So, the as synthesized NCNT-Pd can be used as an advance catalyst for both anode and cathode part in alkaline fuel cell.Graphical abstract.docxThe authors declare no competing interests.This research has supported by the National Research Foundation of Korea (NRF) and funded by the Ministry of Education, Science and Technology (NRF-2021R1F1A1047229).

In this report, we have synthesized nitrogen-functionalized carbon nanotube (NCNT) supported palladium nanoparticles (NCNT-Pd) by using facile chemical reduction method at room temperature. This material has been used as electrocatalyst for electrochemical oxygen reduction reaction (ORR) and ethanol oxidation reaction (EOR). In the case of ORR, NCNT-Pd shows more positive onset potential, higher current density, as well as superior methanol tolerance and long term durability compared with the 20% commercial Pt/C catalyst. Moreover, NCNT-Pd exhibits enhanced catalytic activity for EOR with more negative onset potential and higher current density. The current density of NCNT-Pd for EOR is 2.54 and 5.2 times higher than that of CNT-Pd and 10% Pd/C, respectively. Due to nitrogen functionalization and synergetic effect between NCNT and Pd, NCNT-Pd exhibits superior catalytic performance in ethanol oxidation and oxygen reduction reactions.

Deactivation factorUnit conversion factor ( mol C H 4 min g N i ∕ mmol C H 4 ∕ s ∕ m c a t 2 )Specific heat capacity of the macroparticle (J/kg p a r t i c l e /K)Specific heat capacity of the grain (J/kg s u p p o r t /K)WeiszPrater criterion (Equation 37)Effective diffusivity in the macroparticle ( m 2 ∕ s )Effective diffusivity in the product layer around the grains ( m ∕ s )External heat transfer coefficient ( W ∕ m i n t e r f a c e 2 ∕ K )Heat conductivity of the macroparticle (W/m/K)Heat conductivity of the grain (W/m/K)External mass transfer coefficient ( m g a s 3 ∕ m i n t e r f a c e 2 ∕ s )Kinetic rate constants (unit dependent on equation)Kinetic rate constants (unit dependent on equation)Equilibrium constant (atm)Molar concentration (mol/ m g a s 3 )Molar concentration outside the macroparticle (mol/ m g a s 3 )Initial molar concentration inside the macroparticle (mol/ m g a s 3 )Dimensionless concentration (Equation 32)Molar mass of methane ( kg C H 4 ∕ mol C H 4 )Thiele modulus (Equation 35)Number of grains in radial position rNumber of the grain layers in the macroparticleTotal number of the grainsPartial pressure of methane (atm)Partial pressure of hydrogen (atm)Initial reaction rate ( mmol C H 4 ∕ g N i ∕ min )Actual reaction rate ( mmol C H 4 ∕ g N i ∕ min )radial position in the macroparticle (m)Radial position in the grainPosition of the grain in the macroparticle (m)Dimensionless radial positionRadius of the macroparticle (m)Radius of the grain (m)Radius of the core of the grain (m)Average rate of reaction in given time and radial position ( mol C H 4 ∕ m p a r t i c l e 3 ∕ s )time (s)Dimensionless timeTemperature (K)Temperature outside the macroparticle (K)Initial temperature inside the macroparticle (K)Heat of reaction (kJ/ mol C H 4 )Porosity of macroparticle ( m p o r e 3 ∕ m m a c r o p a r t i c l e 3 )Density of the macroparticle ( kg ∕ m p a r t i c l e 3 )Density of the grain ( kg ∕ m s u p p o r t 3 )Density of produced carbon layer, including porosity ( kg ∕ m c a r b o n 3 )Hydrogen can be produced through different processes from different feedstocks, such as steam methane reforming, water splitting, and thermocatalytic decomposition of methane. Steam methane reforming coupled with CO2 capture and storage (CCS) technologies are the most known and investigated methods in the field of low carbon footprint hydrogen production. However, the separation of produced CO2 and handling and storage leads to costs for gas separation and storage management. Water splitting is an energy and capital intensive process and increases the final price of hydrogen. By contrast, methane decomposition to functional carbon materials and hydrogen has advantages to the alternative processes such as the elimination of additional purification/separation units and production of valuable carbon nanomaterials (tubes or fibers) instead of CO2. The potential applications of carbon nanomaterials in semiconductors, additives to building materials, energy storage, and catalytic materials due to their unique physicochemical properties such as high conductivity, high tenacity and mechanical strength, high specific surface area and semiconductor properties make TCD more economically and environmentally appealing [1–6].Methane, in absence of oxidizing agents or a catalyst (including inert heterogeneities), decomposes naturally to hydrogen and amorphous carbon at high temperatures, >1300 °C (reaction (1)) [7]. The addition of a catalyst facilitates the decomposition of methane in two ways. First, the activation energy and therefore the required temperature for conversion decreases (between 500 °C to 950 °C dependent on the active material of the catalyst); Second, solid carbon can be produced in specific nano-structured shapes, depending on the support and active materials of catalyst and operating conditions. Nickel, iron, copper, and carbon are the most common materials used as the active sites of the catalysts. Many studies have been performed on the catalyst preparation, different support and active materials, thermodynamics and kinetics of the thermocatalytic decomposition of methane [1,2,8–10]. In general, nickel has the highest activity and rate of methane decomposition. Nickel, compared to the others, is active in a lower temperature range and is also more likely to deactivate. However, the addition of a second (or even a third) metal such as iron or copper to the nickel, has shown improved stability at higher temperatures and less deactivation [11–14]. (1) CH4(g) ⟶ C(s) + H 2 (g) Δ H ( 298 K ) = + 74 . 52 kJ∕mol Despite the high potential of TCD for producing carbon nanomaterials and CO2-free hydrogen, it is greatly restricted for industrial applications due to inadequate productivity, uncertainties of process performance and operational challenges coming from carbon formation [2]. Therefore, in addition to studies on catalysts, the TCD reactor and process has to be developed, designed and controlled thoroughly to become feasible at industrial scale. For a rational reactor and process design, modeling and experimental studies can provide the required understanding and basic data for this. This understanding facilitates identification of optimal process conditions for maximum carbon nanomaterial production. Modeling of the catalyst performance as function of equivalent process time is critical for understanding and predicting the product and catalyst evolution in the reactor. This performance can be expressed with the ratio of the mass of produced carbon to the mass of fresh catalyst used, called carbon yield (Eq. (2)) and the change in catalyst particle size and density. (2) c a r b o n y i e l d = m a s s o f p r o d u c e d c a r b o n (g) m a s s o f o r i g i n a l c a t a l y s t (g) In the literature, including the work of Ashik some studies have been reported on modeling at the molecular scale [2,15,16], which helps scientists in catalyst evaluation and to develop a microscopic level of understanding of the complete chemical transformation. Although these models are helpful in the understanding of reaction mechanisms, a model that properly describes the behavior of a catalyst at the macro level has not been reported to the best of our knowledge. In particular, the formation of a functional carbon layer onto the catalyst phase leads to particle growth [2]. The particle size and thus growth is a crucial parameter in designing a reactor. In the present work, for the first time, the multi-grain model (MGM) based on the analogy with the growth of polyolefin particles is developed to describe the macroscopic behavior of growing particles in TCD of methane. The model couples different phenomena involved inside the catalyst particle (which is called macroparticle in this study) such as heat and mass transfer and chemical reaction. A short review of kinetic studies in the literature is provided in Section 2. In Section 3, the model description is presented. The model validation and its reactor predictions assessed in Sections 4 and 5 respectively.In TCD, the activity of the catalyst and kinetic rate of reaction decrease over time due to deactivation. The actual rate of reaction at time t > 0 is described using two parameters: the initial reaction rate and a time-dependent deactivation factor.The initial reaction rate and the reaction mechanism have been studied extensively [2]. Douven et al. and Yadav et al. proposed reaction rate equations for carbon nano-tubes (CNT) production by TCD of the methane which is only dependent on methane concentration [3,17], (see Eq. (3) below). Yadav found out that multi-walled CNT is produced with a different kinetic rate than single-walled CNT, both only depend on the concentration of methane [18], (see Eq. (4)). Ashik et al. 2017 proposed Eq. (5) for the initial reaction rate which also does not depend on the hydrogen concentration [19]. (3) r 0 [ k m o l C H 4 ∕ k g c a t ∕ s ] = K 1 P C H 4 [ a t m ] 1 + K 1 P C H 4 [ a t m ] K 2 2 (4) r 0 [ k m o l C H 4 ∕ k g c a t ∕ s ] = K 1 P C H 4 [ a t m ] 1 + K 1 P C H 4 [ a t m ] K 2 (5) r 0 [ m m o l C H 4 ∕ g c a t ∕ m i n ] = k p P C H 4 1 . 4 [ a t m ] The majority of studies have revealed that the hydrogen concentration has a negative effect on the initial reaction rate due to thermodynamic factors equilibrium and occurrence of the reverse reaction [5,7,20–23]. The kinetic models presented in Eqs. (3)–(5) must therefore be regarded as a simplified form of the actual kinetics. The equations Eqs. (6)–(8) that involve the effect of the hydrogen concentration have a very similar form, but differ mostly in the expression in the denominator. Amin et al. [7] and Snoeck et al. [21] derived Eq. (6) while Borghei et al. [22] suggested different powers for H2 and CH4, Eq. (7), which may be due to the use of a different catalyst and the specific operating conditions used in their studies [7,21,22]. Saraswat et al. [5] have reported a more extended form of the reaction rate, Eq. (8), which includes effects of hydrogen and methane partial pressures in Langmuir–Hinshelwood type of expressions [5]. (6) r 0 [ m m o l C H 4 ∕ g c a t ∕ m i n ] = k ( P C H 4 [ a t m ] − P H 2 2 [ a t m ] ∕ K p ) 1 + K H 2 P H 2 1 . 5 [ a t m ] + K C H 4 P C H 4 [ a t m ] 2 (7) r 0 [ m o l C H 4 ∕ g c a t ∕ h r ] = k ( P C H 4 [ a t m ] − P H 2 2 [ a t m ] ∕ K p ) 1 + K H 2 P H 2 0 . 5 [ a t m ] + K H 2 ∗ P H 2 1 . 5 [ a t m ] 2 (8) r 0 [ m o l C H 4 ∕ g c a t ∕ s ] = k 1 P C H 4 [ a t m ] − k 2 P H 2 2 [ a t m ] 1 + k 3 P C H 4 [ a t m ] + k 4 P H 2 0 . 5 [ a t m ] + k 5 P H 2 [ a t m ] + k 6 P H 2 1 . 5 [ a t m ] 2 According to the literature [5,7,23] The most likely mechanism of the reaction is based on molecular adsorption of methane as the first step, followed by a series of dehydrogenation reactions that are taking place one by one until it ends with separate adsorbed carbon and hydrogen atoms. Detachment of the first H from C H 4 is known to be the slowest and the rate determining step. Every two adsorbed hydrogen atoms form a single H 2 molecule, which is released from the catalytic surface. The carbon atom, however, can diffuse into the nickel catalyst and either forms nanomaterials or encapsulates the active site.The ratio of reaction rate at time t to the initial reaction rate (Eq. (9)) is called the deactivation factor. The deactivation factor expresses the stability of the catalyst over time, which is a crucial factor in order to obtain a high carbon yield. (9) a = r ( t ) r 0 Several different empirical or semi-empirical equations for the deactivation factor a have been defined. Borghei [22] proposed Eq. (10) for the deactivation factor, where b, c and d are constants and k d is defined by an Arrhenius type of temperature dependency. Douven [3] reported that deactivation is reversible and probably due to the formation of amorphous carbon and encapsulation of active sites. Douven used a sigmoid Eq. (11) as the deactivation factor, where parameter b is assumed to have only temperature dependency however parameters t 0 and c decrease slightly as the methane partial pressure increases. (10) a = 1 1 + ( d − 1 ) k d P C H 4 c [ a t m ] P H 2 b [ a t m ] t [ m i n ] 1 ∕ ( d − 1 ) (11) a = d − b tanh t [ s ] − t 0 c Eq. (12) is proposed by Amin et al. [7] and is based on the proposed mechanism in 2.1, mass balance of species on the surface of the catalyst and the assumption that all the reaction steps except one are in equilibrium. So, the concentrations of intermediate species are negligible. All parameters ( K d , K d , C , k d , C H 4 and k d , H 2 ) are temperature dependent following the Arrhenius equation and are determined by fitting the expression to experimental data. (12) a = 1 1 − 0 . 5 k d k d , C + k d , C H 4 P C H 4 + k d , H 2 P H 2 0 . 83 t − 0 . 8 For our model of the TCD reactor and the particle growth over time, the following physical phenomena have to be taken into account: 1. The transport of species into and out of the particle, being diffusion of methane into the catalyst pores and diffusion of hydrogen out of the catalyst pores. 2. The heat transfer into the macroparticle to provide the heat for the strongly endothermic reaction. This includes the transfer from the bulk gas to the external surface of the macroparticle, conduction within the macro macroparticle as well as conduction within the grains. 3. The decomposition of methane on the active sites of the macroparticles into hydrogen and solid carbon. 4. The accumulation of solid carbon onto the catalyst, with consequential growth of the catalyst and deactivation of the catalyst. The transport of species into and out of the particle, being diffusion of methane into the catalyst pores and diffusion of hydrogen out of the catalyst pores.The heat transfer into the macroparticle to provide the heat for the strongly endothermic reaction. This includes the transfer from the bulk gas to the external surface of the macroparticle, conduction within the macro macroparticle as well as conduction within the grains.The decomposition of methane on the active sites of the macroparticles into hydrogen and solid carbon.The accumulation of solid carbon onto the catalyst, with consequential growth of the catalyst and deactivation of the catalyst.These phenomena are very similar to the olefin polymerization, which also experiences particle growth of the macroparticle due to solid product formation. For the solids formation and particle growth different modeling approaches have been developed [24,25]. For our TCD process the most appropriate model, capable of modeling particle growth and convenient for further development to include fragmentation [26], is the Multi-grain model (MGM).MGM is based on two assumptions. Firstly, the macroparticle is spherical with only profiles in the radial direction. So, it is a 1D model in the radial direction. Secondly, the macroparticle is composed of layers of identical non-porous grains (microparticles) with active sites on their surface [25–29]. The growth and evolution of the macroparticle are due to the accumulation of produced carbon on the surface of grains. Fig. 1 shows the schematic of the both concepts of macroparticle and internal grain layers before and after the reaction takes place. Fig. 1(a) shows the schematic of a fresh porous macroparticle which consists of layers of non-porous grains that are illustrated as black spheres. Fig. 1(b) shows a circular sector of the same macroparticle after entering the reactor. The gray shell around each grain is the produced carbon on the grain. Methane and hydrogen diffusion through the pores of the macroparticle and through the layer of accumulated carbon surrounding the catalyst fragments is modeled at two different scales. At the scale of the macroparticle the process is modeled by the diffusion–reaction equation in spherical co-ordinates: (13) ∂ M ( r , t ) ∂ t = 1 r 2 ∂ ∂ r D e r 2 ∂ M ( r , t ) ∂ r − R ( r , t ) Where M is the molar concentration of a component, r is the radial position in macroparticle and D e is the effective diffusivity of the considered component. R ( r , t ) is the average rate of reaction at a given radial position: (14) R ( r , t ) = ( 1 − ε ) ∑ g = 1 N g , r 4 π R g 0 2 . r ( t ) 4 3 π r 3 − ( r − d r ) 3 Where ε is the porosity of catalyst, N g , r is the number of grains at radial position r and R g 0 is the radius of the core of the grains. Eq. (13) can be solved with the following boundary and initial conditions: (15) ∂ M ( 0 , t ) ∂ r = 0 (16) D e ∂ M ( R , t ) ∂ r = k o u t ( M − M b ) (17) M ( r , 0 ) = M 0 k o u t is the external mass transfer coefficient of the macroparticle, M b is the external concentration and M 0 is the initial concentration in the particle.The concentration at the grain scale is also modeled by the diffusion equation in spherical coordinates, Eq. (18). Considering the assumption that the core of the grains are non-porous, so there is no hydrogen or methane inside the core of the grains. Therefore, the boundary conditions, Eqs. (19) and (20) are defined at the surface of the core and outer surface of the grain particle. (18) ∂ M ( r g , t ) ∂ t = 1 r g 2 ∂ ∂ r g D p r g 2 ∂ M ( r g , t ) ∂ r g (19) D p ∂ M ( R g 0 , t ) ∂ r g = r ( t ) . C F (20) M ( R g , t ) = M ( r g , r , t ) Where r g is the radial position in the grain, D p is the diffusivity in the product layer around the grains, R g is the radius of whole-grain and r g , r is the position of grain in the macroparticle (Fig. 1) and C F is the unit conversion factor. The initial condition is the same for the macroparticle (Eq. (17)). Eqs. (13)–(17) are solved only once during each time step for the macro macroparticle, while Eq. (18) and its initial and boundary conditions are solved for all the internal grain layers at each time step.The heat transfer mechanism at both scales is based on conduction. The temperature profile in the macro-particle is calculated by Eq. (21): (21) ∂ T ( r , t ) ∂ t = 1 r 2 ∂ ∂ r k h ρ C P r 2 ∂ T ( r , t ) ∂ r − Δ H ρ C P R ( r , t ) Where k h , ρ and C P are the heat conductivity, the density and the specific heat capacity of the macroparticle respectively, Δ H is the heat of reaction and R ( r , t ) is obtained from Eq. (14). Eq. (21) can be solved with the following initial and boundary conditions (Eq. (22) to (24)): (22) ∂ T ( 0 , t ) ∂ r = 0 (23) k h ∂ T ( R , t ) ∂ r = h ( T − T b ) (24) T ( r , 0 ) = T 0 h is the external convective heat transfer coefficient outside the macroparticle which can be estimated from the Gunn correlation, T 0 and T b are respectively the initial temperature and the temperature outside the macroparticle.The radial temperature profile in the grains can be obtained from the heat conductivity equation in spherical co-ordinates for the whole domain of the core of the grain and the layer of carbon product. However, the heat conductivity, density and specific heat capacity ( k g , ρ g r and C P g ) have different values in the core of the grains and in the product layer. (25) ∂ T ( r g , t ) ∂ t = 1 r g 2 ∂ ∂ r g k g ρ g r C P g r g 2 ∂ T ( r g , t ) ∂ r g (26) k g ∂ T ( R g 0 , t ) ∂ r g = r ( t ) . C F . Δ H (27) T ( R g , t ) = T ( r g , r , t ) One can notice the analogy between mass and heat transfer at both scales of the macroparticle and the grains.In this study, Eq. (6) and (12) and corresponding parameter values are provided in Table 1 and are used in the model as the initial reaction rate and deactivation factor of reaction [7]. Carbon formation increases the radius of the grains and consequently the size of the macroparticles. The growth rate of the radius of the grains is calculated by Eq. (28) and the growth rate of the macroparticle equals the summation of the growth rate of all grain layers, Eq. (29). (28) ∂ R g ∂ t = M W C H 4 R g 0 2 r ( t ) ρ c a r b o n R g 2 (29) ∂ R ∂ t = 2 ∑ 1 N g ∂ R g ∂ t The performance and the results of the model are validated by comparing its results with analytical solutions and results obtained from an independent PDE solver. Two limiting cases are used to verify the implementation of the model. In the simplified case 1, it is assumed that there is only one layer of micro grains, the mass transfer limitation is low, and the reaction is first order without deactivation. In this case, the mass of carbon produced is calculated by Eq. (30): (30) c a r b o n p r o d u c e d (kg) = ( 4 π R g 0 2 . N . t . k . M C H 4 . M M C H 4 ) Where R g 0 is the radius of the core of micro grains, N is the number of micro grains in the only available layer in the macroparticle, t is the time passed since the start of the reaction, k is the kinetic coefficient of the first-order reaction per surface area of grain core ( mol ∕ m 2 ) and M M C H 4 is the molar mass of methane. Fig. 2 illustrates the high accuracy of MGM in case 1, by showing that the MGM results matches Eq. (30). In case 2, again it is assumed that the reaction is first order in methane and independent of the hydrogen concentration without any deactivation ( r = k . P C H 4 ). The second assumption is that the reaction takes place uniformly in the macroparticle. Finally, it is assumed that the physical properties of the macroparticle do not change with time as the reaction proceeds. In these conditions the methane concentration profile inside the particle can be calculated at any time by solving Eq. (31) and its associated initial and boundary conditions. (31) ∂ M ( r , t ) ∂ t = 1 r 2 ∂ ∂ r D e r 2 ∂ M ( r , t ) ∂ r − k . M ( r , t ) M ( r , t = 0 ) = M 0 ∂ M ( r = 0 , t ) ∂ r = 0 M ( r = R , t ) = M b The set of equations can be rewritten in dimensionless from via definition of the following dimensionless quantities: (32) M ̂ = M M b (33) r ̂ = r R (34) t ̂ = D R 2 t (35) M T = R k D e M T is a Thiele modulus and represents the ratio of reaction rate to diffusion rate. By using these dimensionless quantities in Eqs. (31), they change to: (36) ∂ M ̂ ( r ̂ , t ̂ ) ∂ t ̂ = 1 r ̂ 2 ∂ ∂ r ̂ r ̂ 2 ∂ M ̂ ( r , t ) ∂ r ̂ − M T 2 . M ̂ ( r ̂ , t ̂ ) M ̂ ( r ̂ , t ̂ = 0 ) = M ̂ 0 ∂ M ̂ ( r ̂ = 0 , t ̂ ) ∂ r ̂ = 0 M ̂ ( r ̂ = 1 , t ̂ ) = 1 Eq. (36) has been solved with an independent PDE-solver (Matlab pdepe solver). Fig. 3 compares the results of MGM with the Matlab solver. As is clearly illustrated, the results of the MGM are in almost perfect agreement with the Matlab solver over the time, from start of the reaction till the steady state condition has been reached (which is about one minute in this case). This confirms that the model works correctly and there are no errors in the calculation methods. The results of MGM simulations are evaluated by means of a sensitivity analysis and the assessing importance of different parameters and operating conditions in TCD. The temperature range used in our models is limited to the operating conditions that are used to derive the kinetic coefficients of Eqs. (6) and (12) (temperature: 500–650 ° C and maximum hydrogen fraction: 10%) to ensure of the validity of results [7]. Table 2 provides the most important parameters used in the model for the following cases, unless otherwise stated or the importance of the parameter is evaluated. To assess the impact of internal mass transfer limitations, the diffusivity was altered to one thousand times higher and lower values than the (base case) values stated in Table 2. The results are summarized in Fig. 4 and reveal that lowering the internal mass transfer rate does not affect the carbon yield. Hence, for the conditions of Table 2 the effect of diffusion limitation is negligible. On the other hand, if the mass transfer limitation increases to one thousand times higher, the carbon yield decreases by about 35%.The importance of internal diffusional resistance is also confirmed by the Weisz–Prater criterion (Eq. (37)) that estimates the importance of the diffusion on the reaction rates in heterogeneous catalytic reactions [30]. In the normal case, C W P = 9.02 × 10 −4 ≪ 1 which means that internal mass transfer does not influence the production rate of carbon. However, in the case with one thousand times lower effective diffusion coefficient, C W P = 0.902 which implies that for such low diffusion coefficients, internal mass transfer limitation is not negligible anymore compared to the reaction rate. (37) C W P = r 0 ρ R 2 D e M b The same procedure was applied to assess the role of the internal heat transfer limitation, by changing the thermal conductivity of the solid material composing the macroparticle. The results are presented in Fig. 5. Since the carbon yield is not affected by lowering the heat transfer limitation, for the conditions of Table 2 the heat transport resistance is negligible in comparison with other factors. However, in the case with 1000 times higher heat transfer limitation, the carbon yield is decreased dramatically. In this case, the temperature in the macroparticle increases relatively slowly, and as a result the reaction rate and therefore the slope of the curve increases gradually.In addition, there can prevail external heat and mass transfer limitations in the thin film around the macroparticle. However, it was observed that even with the highest external heat and mass transfer limitation, meaning a macroparticle in a stagnant gas phase ( N u = 2 and S h = 2 ) and with larger particle sizes ( 1000 μ m), the production rate of carbon is not reduced. These observations regarding mass transfer importance and their effect on the TCD process are in agreement with literature findings derived from both experiments and the Weisz–Prater criterion [5,22]. Fig. 6 illustrates in logarithmic scale how much the carbon yield changes if the initial reaction rate changes by a factor 1000. Reduction of the reaction rate leads to decrease by a factor 1000 in carbon yield. This finding is another confirmation of the fact that the reaction is the rate-determining step compared to the mass and heat transfer limitations (Section 5.1). On the other hand, if the reaction is one thousand times faster, the carbon yield increases around 450 times. This means that in this case mass and heat transfer limitations become also important which again is in agreement with Section 5.1. As can be seen in Fig. 7 adding inert gas (which means lowering the methane fraction) decreases the carbon yield. On the other hand, for a given fraction of methane, increasing the fraction of hydrogen leads to lower initial reaction rate and higher durability of the catalyst against deactivation. These effects are presented in Fig. 8. These two effects together lead to higher carbon yield, however, in comparison a pure methane feed yields a higher amount of carbon in a shorter amount of time. Temperature has two opposing effects in the TCD process. On one hand, higher temperature leads to a higher initial reaction rate and therefore a higher carbon production rate. On the other hand, increasing the temperature results in faster deactivation of the catalyst and lowers the final carbon yield. These two phenomena are illustrated in Fig. 9. At high temperatures deactivation proceeds more suddenly instead of gradual deactivation at lower temperatures. Thus, curves of 600 ° C and 650 ° C have a very short flat part, rather than longer flat tail.Increasing the temperature between 500 ° C to 650 ° C leads to a lower carbon yield due to the increased deactivation rate. However, it should be noted that this slightly lower amount of carbon is produced in a significantly shorter period of time. Therefore, in the examined conditions and with the used kinetic model, the optimum operating conditions will depend on economic considerations. It should be noted that using a different catalyst (and as a result, different kinetic models) may change this optimum condition.The number of micro grain layers in the macroparticle is a model parameter that is not straightforward to measure or estimate, as previous parameters were. As Fig. 10 shows, this number has a significant impact on the carbon yield. The effect of the number of grain layers is not linear and becomes stronger with an increase in the number of layers. Although physically the number of micro grain layers can be translated to the specific surface area of the macroparticle, the internal structure of an actual catalyst particle is more complex than the structure defined by many layers of identical spheres. Therefore, the number of grain layers will be used as the tuning parameter of the model against validated data. A Multi-Grain Model has been developed to model the heat and mass transfer inside macroparticles coupled with the decomposition reaction of methane. The reaction rate model and deactivation factor from Amin [7] are used, however, the model is suitable for the use of other kinetic models which can be easily accommodated.The effect of operating conditions and model parameters has been investigated by sensitivity analyses and it was found that the heat and mass transfer rates do not limit the carbon production rate and consequently the reaction is the rate-limiting step of the process. This fact is in agreement with experimental findings respected in the literature. However, if a catalyst is made with one thousand times higher ratio of kinetics rate to the mass and heat transfer rates (either by increasing the reaction rate or decreasing the mass and heat transfer rates), the heat and mass transfer limitations will affect the final carbon yield.It was found that, the presence of hydrogen causes a decrease in the reaction rate, however a higher carbon yield is achieved due to delayed deactivation of the catalyst. Moreover, increasing the operating temperature leads to a faster initial reaction rate and faster catalyst deactivation and hence an optimal, process dependent, temperature exists.In the future, it would be interesting to conduct experimental tests to tune, validate and further develop the model. The findings of the current article can be used in CFD models and enable researchers and industry to model, design and predict the behavior of multiple particles in the fixed or fluidized bed reactors employing TCD.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This work is part of the Advanced Research Center for Chemical Building Blocks, ARC CBBC, which is co-founded and co-financed by the Netherlands Organisation for Scientific Research (NWO) and the Netherlands Ministry of Economic Affairs .

ThermoCatalytic Decomposition of methane (TCD) is studied as a method to convert natural gas into hydrogen and functional carbon. In these processes the carbon typically formed on top of a catalyst phase leading to particle growth. Therefore, the development of a particle growth model is necessary to understand the limitations of thermocatalytic decomposition of methane and to assess optimal parameters and process conditions. In this paper, a particle growth model is presented to describe the growth of functional carbon on the catalyst particle. This coupled model requires kinetic equations and information on deactivation rates which have been studied from literature. The morphology of the particle changes due to carbon formation, which leads to eventual deactivation. Therefore, these kinetic expressions are coupled to a particle growth model based on the analogy with the growth of particles in polyolefin production. To combine the effects of particle growth, kinetics, and internal heat and mass transfer, the Multi-Grain Model (MGM) was used. Results confirm that with the currently available catalysts the carbon yield is not affected by heat and mass transfer limitations, however, with the availability of more active catalysts these limitations will become important. Temperature, however, has a significant role in that it regulates the kinetic rate and thus growth rate, which in turn influences the catalyst deactivation. The optimum temperature for the production of nano-carbon, within a reasonable process time, therefore sensitively depends on the choice of catalyst.

In the past century, petroleum has made important contributions to the rapid development of economy and society. As a result of the gradual depletion and non-renewability of petroleum, the use of non-petroleum resources instead of petroleum is of great significance. This alternative route can be generally achieved through syngas (the mixture of H2 and CO) chemistry because syngas cannot only be derived from non-petroleum carbon resources such as natural gas, coal, biomass, and CO2 by gasification, water-gas-shift reactions, or reverse-water-gas-shift reactions (Scheme 1 ) but also directly or indirectly synthesize a wide variety of fuels and basic chemicals. 1–3 At present, the industrial technology for syngas production is relatively mature, but there are still many challenges in highly selectively and stably converting syngas into target products. 4–6 The primary transportation fuel gasoline, which contains hydrocarbons with 5–11 carbons (C5–11) and is almost entirely derived from petroleum now, can also be produced from non-petroleum syngas. 7 As illustrated in Scheme 1, syngas can be initially converted into methanol via methanol synthesis (MS) reaction and then transformed into gasoline via a methanol-to-gasoline (MTG) reaction. Over 30 years ago, Mobil MTG technology was industrialized in New Zealand. 8 To reduce investment and save energy consumption, directly synthesizing gasoline from syngas (STG) without separating intermediate products has been receiving extensive attention. 4 It is well known that gasoline can be obtained from Fischer-Tropsch (F-T) synthesis (mode I). However, because of the limitation of the Anderson-Schulz-Flory distribution, the selectivity of C5–11 is less than 50%. 9 Moreover, the carbon chain-growth mechanism determines that n-paraffins with low-octane value are predominant. 10 Zeolites (or molecular sieves) have uniform pore structures and acidity, which can not only limit the distribution of hydrocarbons according to the size of the micropores but also catalyze the isomerization of n-paraffins. 1 , 2 Therefore, they are often used to mix with F-T catalysts (mode II) to upgrade F-T hydrocarbons. However, it is not easy for zeolite to effectively convert the inert short-chain hydrocarbons, such as methane and ethane, or the heavy hydrocarbons larger than its pores, resulting in a low C5–11 selectivity. 6 Enhancing diffusion of heavy hydrocarbons by increasing the mesoporosity of zeolites or improving the proximity of acid sites and metal by well designing core-shell structures can help alleviate this problem. For example, Tsubaki and colleagues obtained approximately 74% C5–11 at 34% CO conversion on mesoporous Y zeolite combined with Co, 11 while Khodakov and colleagues acquired approximately 61% C5–12 at 37% CO conversion over core-shell ZSM-5/Ru/ZSM-5. 10 Unlike the mode II catalyst, where hydrocarbons are initially produced on the F-T catalyst, the mode III catalyst, which is made by a mixture of MS and zeolite catalysts, yields hydrocarbons at acid sites of zeolite micropores. As a result, heavy hydrocarbons are difficult to generate due to the limitation of micropores. Conventional CuZnAl (CZA) MS catalyst is not suitable for preparing the mode III catalyst because its optimal reaction temperature (473–543 K) cannot effectively start the reaction catalyzed by zeolite. 12 , 13 Therefore, high-temperature MS catalysts have usually been selected. For example, Bao and colleagues recently reported a 76.7% gasoline selectivity with 20.3% CO conversion at 633 K over ZnMnO x -SAPO-11 composite catalyst. 14 Compared with the mode II catalyst, the mode III catalyst tends to achieve high C5–11 selectivity, but because of the low activity of the MS catalyst at high temperature, the conversion efficiency of the latter is significantly lower than that of the former. Taking into account the inconsistency of the reaction temperature and lifetime of the metal (oxides) and zeolite catalysts, placing them in two independent reactors in tandem (dual-bed) is expected to achieve better performance. 15–18 Just like the mode II catalyst, the mode IV catalyst, which contains an F-T catalyst and zeolite separately in a dual-bed reactor, can also adjust the distribution of F-T products. Ding and colleagues recently reported that CO conversion and C5–11 selectivity could both achieve 67% over the dual-bed catalyst Fe3O4@MnO2/H-ZSM-5. 19 Besides the four types of catalysts mentioned above, the dual-bed mode V catalyst, which consists of a syngas-to-DME (STD) catalyst and DME-to-gasoline (DTG) zeolite catalysts in separate beds, can also transform syngas to gasoline. The STD process is very good at converting syngas, which helps to improve the efficiency. 20 In 1987, the Haldor Topsøe TIGAS process, based on the mode V catalyst, was successfully demonstrated in Houston, Texas. 21 , 22 In addition, the Karlsruhe Institute of Technology has developed a similar technology called the bioliq process. 23 Although some technologies for this dual-bed mode V process have existed, still few studies obtain high C5–11 selectivity at high syngas conversion or the mechanism of the zeolite characteristics on selectivity and stability.Here, we report an 80.6% C5–11 selectivity without CO2 at 86.3% CO conversion over a dual-bed catalyst (CZA + Al2O3)/N-ZSM-5(97) that includes a STD catalyst CZA + Al2O3 in the upper bed and a DTG catalyst N-ZSM-5(97) in the lower bed (Figure S1). A low amount of acid and the nano-sized structure of the ZSM-5 zeolite are beneficial to C5–11 selectivity and stability, respectively. The deactivation mechanism is also explored and discussed.The performance of syngas conversion on various catalysts was compared at 573 K and 3.0 MPa with identical syngas feed. As shown in Figure 1 A, CZA + Al2O3 is a typical STD catalyst that gives a 90.4% DME selectivity at 64.5% CO conversion. When this STD catalyst is mixed with N-ZSM-5(97) (nano-sized ZSM-5 with Si/Al = 97), the CO conversion and CO2 selectivity are significantly improved, and a considerable amount of the DME is converted to hydrocarbons. The selectivity of C1–2 light hydrocarbons is as high as 20.0%, whereas the selectivity of C5–11 liquid hydrocarbons without CO2 is only 26.1%. Increasing the proximity by three-component grinding and pressing leads to further decrease the selectivity of C5–11 hydrocarbons (Figure S2). It is interesting to find that the reaction result of the dual-bed catalyst (CZA + Al2O3)/N-ZSM-5(97), which is configured by CZA + Al2O3 on the upper bed and N-ZSM-5(97) on the lower bed, is quite different from the mixed catalyst (CZA + Al2O3+N-ZSM-5(97)). The DME and MeOH are completely converted, the C5–11 selectivity reaches up to 75.9%, and the C1–2 selectivity is as low as 4.6% over (CZA + Al2O3)/N-ZSM-5(97). The effect of the structure and Si/Al ratio of ZSM-5 catalysts in the lower bed on the performance of the STG reaction was explored. The conversion of CO over the four dual-bed catalysts in Figure 1B is close because the upper-bed STD catalysts and the reaction conditions are the same. However, their C5–11 selectivity and stability are quite different. As shown in Figures 1B and S3A–S3D, the dual-bed catalyst with the nano-sized N-ZSM-5(97) or N-ZSM-5(21) has much better stability than the micro-sized M-ZSM-5(116) or M-ZSM-5(18), respectively. Furthermore, the dual-bed catalyst with N-ZSM-5(97) or M-ZSM-5(116) with a high Si/Al ratio exhibits considerably higher C5–11 selectivity than that with N-ZSM-5(21) or M-ZSM-5(18) with a lower Si/Al ratio, respectively. This suggests that the nano-sized structure, which generally means good diffusion ability, 24 is beneficial to extend the lifetime; meanwhile, the high Si/Al ratio, which usually represents a low amount of acid, is conducive to inhibit the formation of light hydrocarbons. It should be mentioned that the above results were obtained by a high-throughput reactor, and the reaction temperatures (573 K) of the upper and lower beds were the same. In fact, such a high temperature is not suitable for the STD reaction (Figure S4). To gain better results, we studied in detail the STG reaction over the dual-bed (CZA + Al2O3)/N-ZSM-5(97) catalyst at different reaction temperatures for the two beds.The STG reaction on (CZA + Al2O3)/N-ZSM-5(97) was investigated at T (upper bed) = 533 K, T (lower bed) = 593 K, P = 3.0 MPa, H2/CO = 2, and gas hourly space velocity (GHSV) = 1,500 mL g−1 h−1. As shown in Figure 2 A, the selectivity of C5–11, C3–11, aromatics, or CO2 was kept at approximately 79%, 34%, 98%, or 32%, respectively, with 87% CO conversion. The light C1–2 selectivity was less than 1.7%. The activity of this dual-bed catalyst did not decrease at all within 110 h on stream. The detailed distribution of gasoline-range C5–11 can be observed in Figure 2A. The selectivity of iso-paraffins reached approximately 40%, whereas the selectivity of olefins was less than 2%. Moreover, the iso/n-paraffin ratio reached 18, which is much higher than that in F–T products. 6 , 25–28 Figures S5 and S6 indicate that increasing GHSV or H2/CO ratio had little effect on C5–11 selectivity. Figure 2C shows that increasing the pressure significantly increased the CO conversion without affecting the C5–11 selectivity, which is extremely valuable for improving the efficiency of the STG process. Notably, at P = 4.0 MPa, H2/CO = 2, and GHSV = 3,000 mL g−1 h−1, the CO conversion was as high as 86.3%, while the selectivity of C5–11, C3–11, and CO2 reached 80.6%, 98.2%, and 30.6%, respectively. By calculation, the space time yield was up to 0.28 g C5–11 per hour per gram of dual-bed catalyst. The main research progress concerning the conversion of syngas to C5–11 liquid hydrocarbons (including aromatics) in the past 3 years is listed in Table S1. 10 , 11 , 14 , 19 , 29–36 It is apparent that, compared with F-T-based catalysts (modes I, II, or IV), this dual-bed catalyst (mode V) has significant advantages in suppressing the low-value light C1–2 and achieving a high C5–11 and iso/n-paraffin ratio. Besides, compared with MS-catalyst-based physically mixed catalysts (mode III), this mode V catalyst has remarkable advantages in achieving high CO conversion and regenerating the deactivated catalyst after long-term operation. The effect of the reaction temperature on the lower bed was studied. It can be seen from Figure 2D that increasing the temperature is harmful to the generation of C5–11.Note that the dual-bed catalyst (CZA + Al2O3)/N-ZSM-5(97) for all of them was configured by 1.5 g CZA + Al2O3 (upper bed) and 1.0 g N-ZSM-5(97) (lower bed).The X-ray diffraction (XRD) patterns in Figure 3 A indicate that the four ZSM-5 zeolites in this study possess a typical pure MFI structure. The XRD patterns in Figure S7 suggest that CZA and Al2O3 exhibit the structures of a conventional industrial CuZnAl MS catalyst and a γ-Al2O3 catalyst, respectively. 37 The NH3-TPD results in Figure 3B show that the acid amount follows the order M-ZSM-5(116) < N-ZSM-5(97) < N-ZSM-5(21) < M-ZSM-5(18), which is exactly the reverse order of the Si/Al ratio (Figures 3C–3F). The NH3-TPD result in Figure S8 proves that Al2O3 has an acid property, which should be derived from Lewis acid sites. 20 The XRF results listed in Table S2 show that the Cu/Zn/Al molar ratio is 4.8:1.8:1 for CZA. It can be observed from Figures 3C–3F that N-ZSM-5(97) and N-ZSM-5(21) are composed of approximately 200 and 50 nm particles, respectively, while M-ZSM-5(116) and M-ZSM-5(18) are both made up of 2–5 μm hexagonal crystals. Table S3 shows that N-ZSM-5(97) has the highest Brunauer-Emmett-Teller (BET) area and external area. Combined with the results in Figures 1B, S3, and 3C–3F, there is no doubt that reducing the crystal size of the lower-bed ZSM-5 zeolite will considerably prolong the lifetime of the overall dual-bed catalyst. In addition, it can be inferred from Figures 2B and 3B that, whether using nano- or micro-structured ZSM-5 zeolite, low acid content can facilitate the formation of C5–11 and depress the generation of light hydrocarbons. Generally, the growth of the carbon chains largely depends on the oligomerization of the initial light olefins. 38 However, the hydrogenation of light olefins, which can be catalyzed by the acid sites (especially Brønsted acids) of H-form ZSM-5 zeolite, 39 , 40 is disadvantageous. Compared with oligomerization, a higher acid content for ZSM-5 zeolite can be more beneficial to the hydrogenation, which results in a lower C5–11.The four ZSM-5 zeolite catalysts in the lower bed after reaction shown in Figure 1B were analyzed by the thermogravimetric method. As presented in Figure 4 A, their weight losses follow the order N-ZSM-5(97) < N-ZSM-5(21) < M-ZSM-5(116) < M-ZSM-5(18). This demonstrates that for ZSM-5 zeolites with approximate Si/Al ratios, nano-sized structures are more resistant to coke, while for ZSM-5 zeolites with similar structures, a low Si/Al ratio (or high acid content) is ready to cause carbon deposits. Also, the catalytic results of a lower-bed ZSM-5 catalyst with a particle size of approximately 500 nm also prove that decreasing the zeolite particle size is beneficial to reduce coke and prolong lifetime (Figures S9A–S9D). During syngas conversion reactions, after the zeolite catalyst is mixed with the metal catalyst, the coke formation rate will be generally significantly reduced, and the catalyst stability will be greatly improved. 41 , 42 As shown in Figure 4A, the weight loss of the spent N-ZSM-5(97) in the dual-bed catalyst is close to that in the mixed catalyst, which implies that the dual-bed catalyst (CZA + Al2O3)/N-ZSM-5(97) has the potential for long-term use. After reaction, the four ZSM-5 zeolite catalysts in the lower bed were dissolved by hydrofluoric acid, and then the retained organic species were extracted by dichloromethane and analyzed by gas chromatography-mass spectrometry. As presented in Figure 4C, the content of aromatics (species 1, 2, 5, 6, and 8) with no more than ten carbons (or no larger than tetramethyl-benzene) in the spent N-ZSM-5(97) is much less than that in the others. However, the selectivity of aromatics for (CZA + Al2O3)/N-ZSM-5(97) is substantially higher than that for the other three dual-bed catalysts in Figures S3A–S3D. This indicates that the products no larger than the size of micropores (0.53 × 0.56 nm) for MFI topology are easily diffused out of N-ZSM-5(97). It also can be seen that the naphthalene derivatives (species 14–16) are hardly generated for N-ZSM-5(97). These polycyclic aromatics are too large to block micropores and cause catalyst deactivation. 43 Besides methylbenzenes and polycyclic aromatics, a considerable amount of oxygenates, which include the derivatives of 2-cyclopenten-1-one (species 3–6) and phenol (species 13), can be observed in Figure 4B. The mass spectra of the typical oxygenates are shown in Figures 4C and S10. Our previous researches have proved that the 2-cyclopenten-1-one species, which can be produced via a series of C–C bond formation reactions, such as carbonylation, aldol, prins, and hydroacylation, are important intermediates for the formation of single-ring aromatics. 44–46 Compared with the spent N-ZSM-5(97), the spent N-ZSM-5(21) obviously contains more 2-cyclopenten-1-one species. However, the selectivity of aromatics over (CZA + Al2O3)/N-ZSM-5(97) is apparently higher than that over (CZA + Al2O3)/N-ZSM-5(21) (Figures S3A and S3B). This means that there are other ways to generate aromatics. Recently, Wei and colleagues found that phenol species formed by the aldol cycle in the syngas atmosphere can be transformed to aromatics. 47 The amount of phenol species in spent N-ZSM-5(97) is more than that in spent N-ZSM-5(21), which is positively related to the selectivity of the aforementioned aromatics. This suggests that phenol species are likely to act as intermediates for aromatics. In the spent N-ZSM-5(97) of the mixed catalyst (CZA + Al2O3 + N-ZSM-5(97)), 2-cyclopenten-1-one species can be hardly found, whereas phenol species can be detected. This mixed catalyst can actually produce few aromatics (6%), which also suggests that the speculation above is reasonable. Although the mechanism of synthesizing single-ring aromatics from these oxygenates, such as 2-cyclopenten-1-one and phenol species, has been well explored by the previous works of ourselves and others, 15 , 44 , 45 , 47 the pathway of generating polycyclic aromatics, which are the main factors leading to catalyst deactivation, has still not been revealed. The weight loss for the spent M-ZSM-5(18) is more than twice that for the spent N-ZSM-5(97) (Figure 4A); however, the amount of their soluble organics is relatively close (Figure 4B). This demonstrates that there is a large amount of insoluble heavy carbon deposits, which are generally polycyclic aromatics, 43 in the spent M-ZSM-5(18). Interestingly, some 2,3-dihydro-1H-inden-1-one oxygenates (species 11 and 12) can be definitely detected and identified in the spent M-ZSM-5(18) (Figures 4B, 4D, and S3D). We consider that they can be generated through carbonylation and condensation of phenol species and transformed to polycyclic aromatics by reactions such as isomerization and dehydration because this is similar to the mechanism for the formation and conversion of 2-cyclopenten-1-one species. 44–46 In summary, high and selective conversion of syngas to gasoline-range C5–11 liquid hydrocarbons can be simultaneously achieved over a dual-bed catalyst (CZA + Al2O3)/N-ZSM-5(97) that consists of a STD catalyst CZA + Al2O3 (a mixture of CuZnAl MS catalyst and acidic γ-Al2O3 catalyst) in the upper bed and a DTG catalyst N-ZSM-5(97) (nano-sized H-ZSM-5 zeolite with Si/Al ratio = 97) in the lower bed. The selectivity of C5–11 and C3–11 can reach 80.6% and 98.2%, respectively, along with 86.3% CO conversion at T (upper bed) = 533 K, T (lower bed) = 593 K, P = 4.0 MPa, H2/CO = 2, and GHSV = 3,000 mL g−1 h−1. This dual-bed catalyst exhibits an excellent stability during a 110-h test. The iso/n-paraffin ratio in the C5–11 is up to 18. By comparing four lower-bed ZSM-5 zeolite catalysts with various particle sizes and acid content, we found that the nano-sized structure is beneficial to reduce coke and prolong lifetime; meanwhile, the low acid content is advantageous to increase C5–11 selectivity. The 2,3-dihydro-1H-inden-1-one species can be definitely detected and identified in the spent lower-bed micro-sized M-ZSM-5(18). They are regarded as intermediates to generate polycyclic aromatics, which generally lead to catalyst deactivation. The dual-bed catalyst (CZA + Al2O3)/N-ZSM-5(97) suggests a promising application in producing gasoline from syngas.Full experimental procedures are provided in the supplemental information.Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Professor Zhongmin Liu ([email protected]).This study did not generate new unique reagents.The published article includes all datasets generated or analyzed during this study.We acknowledge financial support from the National Natural Science Foundation of China (grant nos. 21978285, 21991093, and 21991090) and the “Transformational Technologies for Clean Energy and Demonstration” Strategic Priority Research Program of the Chinese Academy of Sciences (grant no. XDA21030100). We thank Weichen Zhang for assistance in the experiments.Y.N. designed and performed the experiments, analyzed the data, and wrote the manuscript. K.W. discussed the results. W.Z. and Z.L. supervised the study, discussed the results, designed the experiments, and revised the manuscript.The authors declare no competing interests.Supplemental information can be found online at https://doi.org/10.1016/j.checat.2021.02.003. Document S1. Supplemental experimental procedures, Figures S1–S10, and Tables S1–S3 Document S2. Article plus supplemental information

Achieving high conversion of syngas to fuels and basic chemicals with excellent selectivity and stability remains a challenge. Here, we report an 80.6% selectivity of gasoline-range C5–11 hydrocarbons at 86.3% CO conversion over a dual-bed catalyst (CZA + Al2O3)/N-ZSM-5(97) that consists of an upper-bed syngas-to-dimethyl ether (DME) catalyst (CZA + Al2O3) and a lower-bed DME-to-gasoline catalyst (nano-sized N-ZSM-5(97) zeolite). This dual-bed catalyst exhibits an excellent stability in a 110-h reaction test. The iso/n-paraffin ratio in the C5–11 is up to 18. For the lower-bed zeolite catalyst, the nano-sized structure is beneficial to reduce coke and prolong lifetime; meanwhile, the low acid content is advantageous to increase C5–11 selectivity. The 2,3-dihydro-1H-inden-1-one species can be definitely detected and identified in the spent lower-bed micro-sized M-ZSM-5(18) catalyst. They are regarded as intermediates to generate polycyclic aromatics, which generally lead to catalyst deactivation. The dual-bed catalyst (CZA + Al2O3)/N-ZSM-5(97) suggests a promising application in producing gasoline from syngas.

With the current energy crisis and increasingly serious environmental issues, it is very inevitable to search for renewable and sustainable energies as alternatives to fossil fuels [1]. In the portfolio of renewable energies development, H2 energy from formic acid (FA) is generally regarded as one of the most promising paradigms. It has recently attracted tremendous research passion because FA with high H2 capacity of 4.3 wt% has high stability, low-toxicity, low-flammability, and biodegradability [2]. It is universally acknowledged that H2 evolution from FA involves a dehydrogenation pathway ( H C O O H → C O 2 + H 2 ) [3] and a dehydration reaction ( H C O O H → C O + H 2 O ) [4]. The produced CO in the later undesired reaction is highly poisonous to the catalysts. Thus, the main challenge in the FA decomposition for H2 production is to develop a highly efficient and selective catalyst to avoid the undesired CO generation. To date, supported Pd-based heterogeneous catalysts have been proven to be the most efficient for H2 evolution from FA system [3,5–8]. However, the sluggish kinetics of H2 evolution over Pd-based catalysts cannot satisfy the industrial applications [8]. From this point of view, the development of Pd-based catalysts with sufficient active sites and selectivity for H2 evolution from FA is of great importance to enter the hydrogen economy era.Supported Pd-based catalysts, including mono-, bi- and tri-metals, have been intensively investigated for FA decomposition in the literature [9–11]. The fact is widely regarded that the addition of exotic metals and the nature of the support matrix are critical to dictate the performance of catalyst [12,13]. On one hand, for the Pd-based catalysis FA decomposition reactions, the most common additives are Au (Pd–Au) and Ag (Pd–Ag) [13]. On the other hand, various supports have been demonstrated to be effective for the Pd-based catalysts, including carbon materials (activated carbon (AC), mesoporous carbon (MSC), reduced graphene oxidization (rGO), etc.), metal-organic frameworks (MOFs), metal oxides, mesoporous silica, and macroreticular basic resin [14–17]. However, the preparation procedure of above-mentioned Pd-based support materials is time-consuming and costly. Thus, it is of great interest to develop the supported Pb-based catalysts on cheap biomass resource matrix and to explore its catalytic performance for FA dehydrogenation.Herein, the present work proposes a facile and cost-effective approach to load PdAg bimetal nanoparticles on cellulose modified with polyetherimide (PEI) as an efficient catalyst for H2 generation from FA decomposition in a sodium formate-free aqueous system. The cellulose was obtained from Eucalyptus biomass through 70% FA fractionation. Interestingly, the resultant cellulose-derived PdAg bimetallic catalyst (PdAg-PEI-FAC) is a core-shell-like structure and shows a high catalytic performance with the turnover frequency (TOF) value of 2875 h−1. The as-obtained catalyst has excellent stability toward FA decomposition with no loss of catalytic activity after five recycles. The remarkable catalytic activities of this catalyst result from high dispersion Pd and synergistic effects between the PdAg bimetallic system. In order to give more insight into FA fractionation, the properties of isolated lignin and hemicelluose fractions from Eucalyptus biomass through 70% FA fractionation were also characterized in this present work as Supporting information. In addition, the probable catalytic mechanism and recyclability of PdAg-PEI-FAC for H2 evolution from FA solutions were also evaluated. . Eucalyptus wood was kindly gifted from Yingqiang Wei, a staff of Gaofeng Forest Farm, Guangxi province, China. The contents of cellulose, hemicellulose and lignin of Eucalyptu biomass were calculated to be approx. 42.2%, 12.6% and 31.8%, respectively, according to the methods detailed in our previous work [18]. Formic acid (FA, 88%) and polyetherimide (PEI, Mw = 70,000 g mol−1) were purchased from Aladdin Biochemical Technology Co., Ltd., Shanghai, China. Glucose standard (99.5%) was bought from Guangfu Fine Chemical Co. Ltd., Tianjin, China. 5-Hydroxymethyl furfural (5-HMF, 99%) and furfural (99%) standards were of chromatographic-grade and purchased from J & K Scientific GmbH, Pforzheim, Germany. All other reagents including PdCl3 (AR, Pd content ≥ 59%), AgNO3 (AR, ≥ 99.8%), CoNO3·6H2O (AR, ≥ 98.5%), AuCl3·HCl·4H2O (AR, Au conent ≥ 48.7%), NiNO3(AR, ≥ 99%), NaBH4 (AR, ≥ 99%) and AlCl3 (AR, ≥ 99.8%) were of analytic grade and bought from Sinopharm Chemical Reagent Co., Ltd, Beijing, China.The fractionation flow of Eucalyptus biomass through 70% FA is depicted in Scheme 1 . For this FA fractionation processing, the resultant cellulose was modified with PEI for preparation of the cellulose-derived PdAg bimetallic catalyst for H2 evolution, while the properties of lignin fraction and hemicellulose upgrade were also investigated.In brief, 44 g of the milled Eucalyptus biomass powder along with 445 mL 70% FA were loaded into a 2 L round-bottom glass flask, which was placed into an aluminum heating module with a magnetic digital stirring hotplate (Gongyi, Henan, China). The weight/volume ratio of biomass solid weight (g) to FA liquid volume (mL) was set at 1 : 10 (w/v). The solid/liquid slurry was stirred at 1300 rpm and 130 °C. After 3 h reaction, the reaction mixture was cooled to room temperature and filtered using 0.45 μm filter paper to separate the solid residue cellulose and filtrate. Approx. 18 g of the solid residue cellulose fraction was obtained and assigned as FAC (formic acid cellulose). 400 mL of the black liquor filtrate containing FA, soluble lignin and carbohydrate hydrolysates was collected in a glass flask. After evaporation in vacuum conditions of −0.1 MPa and 50 °C, about 340 mL of FA could be recovered from the filtrate for re-usage. Once the FA was recoved, lignin was precipitated at the bottom of flask. Subsequently, 300 mL distilled water was used to wash the precipitated lignin in duplicate. Through filtration, approx. 9 g of brown lignin pellets were achieved, and approx. 450 mL of the aqueous liquid consisting of hemicellulose and cellulose hydrolysates (including C5 and C6 monomeric and oligomers) were collected. These carbohydrate hydrolysates could be further converted into furans using AlCl3 as a Lewis catalyst.The synthesis procedure of polyetherimide (PEI) modified cellulose-derived Pd/Ag bimetallic catalyst (PdAg-PEI-FAC) is shown in Fig. 1 a in Section 3.2. First of all, FAC (200 mg) was modified with 40 mL of 1.0 wt% PEI with Mw = 70,000 g mol−1 to obtain PEI-FAC support. Then, the as-prepared PEI-FAC was homogeneously dispersed in 80 mL of deionic water, followed by the addition of 3.6 mL of Na2PdCl4 (20 mmol L−1) and 2 mL of AgNO3 (35 mmol L−1) solution with the Pd/Ag mass/weight ratio of 3.75/3.75. The mixture solution was stirred at room temperature and 500 rpm for 30 min. Subsequently, the sample was reduced with 0.213 mol NaBH4 for 1 h at the stirring rate of 500 rpm. After filtration, the black solid residue was washed with distilled water several times to remove the unreacted metal ions and redundant NaBH4. After lyophilization, PdAg-PEI-FAC catalyst was achieved and labeled as Pd3.75Ag3.75-PEI-FAC. When metal Ag in Pd3.75Ag3.75-PEI-FAC was substituted with other metal elements (M = Co, Ni, and Au) with the Pd/M mass/weight ratio of 3.75/3.75, three other bimetallic catalysts were obtained and labeled as Pd3.75Co3.75-PEI-FAC, Pd3.75Ni3.75-PEI-FAC and Pd3.75Au3.75-PEI-FAC. In addition, by changing the PdAg compositions (Pd/Ag mass/weight ratio = 1.5/1.5, 2.5/2.5, 5/5 and 3.75/0), four relative counterpart bimetallic catalysts of Pd1.5Ag1.5-PEI-FAC, Pd2.5Ag2.5-PEI-FAC, Pd5Ag5-PEI-FAC and Pd3.75-PEI-FAC were also prepared.In order to demonstrate that PEI-FAC is a suitable support material, several counterpart support matrices, such as reduced oxidative graphite (rGO), nitrogen-doped carbon (N@C), and carbon blank (C) were used to load PdAg bimetals as catalysts and labeled as Pd3.75Ag3.75-rGO, Pd3.75Ag3.75-N@C and Pd3.75Ag3.75-C, respectively. All those catalysts were compared to evaluate their catalytic performance of H2 generation from FA solution.The crystalline structure of PdAg-PEI-FAC was characterized by Rigaku Ultima IV X-ray diffractometer (XRD, Rigaku, Japan) with Ni-filtered CuK radiation operated at 40 kV and 40 mA in the air, with an environmental humidity of around 60%. Samples were tested under a diffraction angle, 2, in the range of 5° ∼30° with a step interval of 0.5° min−1.The elemental distribution and chemical state of catalyst was analyzed by XPS (ESCALAB 250, Thermo Fisher Scientific, USA). The data was acquired using Monochromated Al Kalph (150 W), a pass energy of 200 eV for survey, 30 eV for high resolution scans. The analyzed area was 500 × 500 μm.The metal Pd and Ag contents in PdAg-PEI-FAC samples were detected by PS-4 inductively coupled plasma atomic emission analysis (ICP-AES) spectrometer (Baird Co., USA). The data was analyzed by software PLASMAⅣ. The data accuracy was up to 10 ± 0.5%.The pore properties and surface area of the PEI-FAC and Pd3.75Ag3.75-PEI-FAC were tested by Brunauer Emmett-Teller (BET) analyses, which were performed on Autosorb IQ Quantachrome (USA) using N2. The samples were degassed at 250 °C for 12 h before measurement.The morphology and particle size of Pd3.75Ag3.75-PEI-FAC and Pd3.75Ag3.75-FAC were detected by high angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images on JEOL JEM-ARM200F (Japan) at 200 kV.To examine the catalytic performance of PdAg-PEI-FAC catalysts for H2 production from FA decomposition, the effects of four critical factors on H2 generation were investigated, including FA concentration (0.5 mol L−1, 1 mol L−1, 1.5 mol L−1, 2.0 mol L−1, 5.0 mol L−1 and 10 mol L−1), reaction temperature (30 °C, 40 °C, 50 °C and 60 °C), Pd/Ag composition (1.25/1.25, 2.75/2.75, 3.75/3.75 and 5.0/5.0 wt%) and support matrix (rGO, N@C, carbon, FAC-PEI). The total reaction system was fixed at 5 mL, and the loading of the catalyst was set at 100 mg. The quantity of total gas volume (Vgas, or V H 2 + C O 2 ) produced from the reaction was collected using a 500 mL gas burette by a drainage method.Apart from the total gas volume (Vgas, or V H 2 + C O 2 ), the catalytic activities of PdAg-PEI-FAC were also evaluated from the following two aspects, turnover frequency (TOF, h−1) and activated energy (E a , kJ mol−1).TOF was calculated through Pd content detected by ICP-AES according to Eq. (1): (1) T O F = T o t a l v o l u m e o f H 2 g e n e r a t i o n ( m L ) R e a c t i o n t i m e ( h ) × P d c o n t e n t i n t h e c a t a l y s t ( g ) Ea was calculated according to Arrhenius equation (Eq. (2)): (2) ln T O F = ln A − E α R T where, A means frequency factor, R is gas constant, R = 8.314 J K−1 mol−1, T is reaction temperature, K. The value of Ea is obtained from the slope of the fitting straight line.Composition analyses of CO2, H2 and CO were conducted on a SP-2000 (Beijing Beifen Ruili Analytical Instrument Co., Ltd.). The compositions of H2 and CO2 were determined by a GC spectrum using a thermal conductivity detector (TCD), while the CO content was detected by a GC spectrum using a flame ionization detector (FID)-Methanator.The statistical analyses were determined using origin 10.0 (OriginLab Co., USA), and the final data were expressed by average ± SD.In order to obtain the cellulose from the biomass to synthesize the cellulose-derived bimetallic catalyst, 70% aqueous FA was used for the efficient fractionation of Eucalyptus biomass (Fig. S1). The reasons to choose 70% FA as fractionation solvent were: On one hand, FA is a biomass-derived and sustainable renewable solvent [19] and has been intensively employed for nondestructive fractionation of biomass in the literature [20]. On the other hand, other articles reported having employed concentrated FA (≥ 80 wt%) [20], much lower FA concentration is rarely used for biomass fractionation. The lower FA concentration is, the less FA erodes the equipment. During FA fractionation processing, FA plays three major roles: solubilization effect dissolving lignin, catalyst function hydrolyzing hemicellulose, and reaction substrate formylation with hydroxyl groups on the surface of cellulose and lignin. Although increasing FA concentration and temperature could lead to high lignin dissolution and cellulose yield, it was prone to cellulose degradation and lignin repolymerization [20]. In this present work, approx. 73.18% of dissolved lignin is isolated using 70% FA as the fractionation solvent at 130 °C for 3 h, this value is similar to those reported by Zhou et al. [21] and Li et al. [22].The structural properties of the isolated cellulose were characterized by SEM and HPLC, the results are shown in Fig. 1. As seen from Fig. 1a–f, the FAC morphology of SEM image is much smoother (Fig. 1f), while the surface of the native cellulose from Eucalyptus is rough (Fig. 1d and e). Subsequently, the compositions of native cellulose and FAC were analyzed by HPLC (Fig. 1g), and results are shown in Fig. 1h. After FAC was acid hydrolyzed before HPLC determination, 2.49% of formic acid was detected (Fig. 1g and h), indicating that formylation has occurred on the surface of FAC during FA fractionation. This phenomenon was intensively demonstrated in the earlier reported papers [20–22]. The subtle change of formylation on the FAC surface is directly detrimental to cellulose enzymatic hydrolysis with only ∼62% of saccharification efficiency. Our results agree with the report in the literature [21]. As an alternative promising option, FAC can be employed to synthesize cellulose-derived PdAg bimetallic catalysts after decoration with polyetherimide (PEI). The as-prepared PdAg bimetallic catalyst is used to produce hydrogen from FA decomposition at room temperature.Apart from cellulose, the other two fractions of lignin and hemicellulose were also obtained from FA fractionation of Eucalyptus biomass (Fig. S1). As observed from the mass balance of the isolated principle fractions of cellulose, hemicellulose and lignin in Fig. S2, it is worthily noticeable that 90.31% lignin and nearly 95% hemicellulose are dissolved in 70% FA aqueous solution, leaving behind a high purity of 80.14% cellulose as solid residue, which has a yield as high as ∼82%. According to the data in the lab-scale level, 100 g of dry biomass can achieve 33 g of cellulose solid residue with the purity of 80.14%, 27 g of lignin with high purity of 90.31%, and 9.07 g of dried hemicellulose hydrolysate powder with xylose concentration of 65.62%.Importantly, more than 90% of FA can be recovered for recycleability. It has been demonstrated that the recycled FA shows the similar excellent fractionation capacity as the fresh FA in terms of isolated cellulose and lignin yields (Fig. S3).In order to fully understand the FA fractionation, the properties of isolated lignin from Eucalyptus biomass through 70% FA fractionation were also characterized FT-IR, GPC, and TGA. As seen from FT-IR profile in Fig. S4, albeit with formylation, lignin possesses hydroxyl groups very similar to the native lignin from sugarcane bagasse [23]. The typical strong absorbance peaks corresponding to aromatic skeleton vibrations at 1603 cm−1, 1509 cm−1, 1458 cm−1 and 826 cm−1 are observed in FT-IR for the resultant lignin. Table S1 shows FT-IR characteristic peaks of lignin isolated from FA fractionation of Eucalyptus biomass. To our delight, compared with the native lignin (Mp = 6217 g mol−1), GPC determination shows a decrease of mass weight of 2582 g mol−1 for the resultant lignin from FA fractionation (Fig. S5), indicating approx. 17% of lignin are depolymerized into mono- or oligo-aromatic compounds, which are not detected in this work. Interestingly, depolymerization and condensation phenomena of the resultant lignin is not observed using low FA concentration (70%) as fractionation solvent, which is quite different from the previous reports in the case of high FA (≥ 88%) fractionation of biomass [20,21]. TGA analysis shows that the depolymerization temperature related to inter-unit linkage of lignin backbone is 515.7 °C, and only approx 5.74% of mass residue was retained at 800 °C (Fig. S6). It indicates that the resultant lignin isolated from FA fractionation has high purity (≥ 90%), and is quite suitable for downstream upgrading towards carbon functional materials, such as lignin-derived catalysts [24].For FA fractionation of Eucalyptus, hemicellulose and part of cellulose are hydrolyzed to C5 and C6 sugars (30.2 g L−1), including 19.8 g L−1 xylose, 7.7 g L−1 glucose, 1.2 g L−1 xylose oligomers, and 1.8 g L−1 glucose oligomers (Fig. S7). These sugar mixtures can be converted into high value-added commercial building blocks, such as furural and 5-hydroxylmethylfurual (5-HMF) using AlCl3 as a Lewis catalyst [25]. Results in Fig. 9d show that 80% furural derived from C5 sugars and 60% 5-HMF from C6 sugars are simultaneously obtained at 110 °C and 2 h. The results are similar with those using pure xylose and glucose conversion into furan compounds in the literature [26,27].As aforementioned, the subtle change of formylation on the FAC surface is directly detrimental to cellulose enzymatic hydrolysis with approx. 62% of saccharification efficiency. In the present work, FAC is decorated with PEI and then covalent with bimetallic ions Pd2+ and Ag+ to synthesize cellulose-derived PdAg bimetallic catalyst for hydrogen production from aqueous FA in a sodium formate (SF)-free solution. Fig. 2 a shows the synthesis schematic flow of cellulose-derived PdAg bimetallic catalyst.X-ray photoelectron spectroscopy (XPS) is employed to examine the chemical valence states of Pd and Ag in the as-prepared catalysts. From the XPS profile of Pd in Pd3.75Ag3.75-PEI-FAC(Fig. 2b), Pd0 (Pd 3d5/2 335.4 eV, Pd 3d3/2 340.7 eV) and Pd2+ (Pd 3d5/2, 337.7 eV; Pd 3d3/2, 342.7 eV) [28]. The existence of oxidized Pd species can be ascribed to the oxidation of metallic Pd in the air during preparation. While for Pd3.7-PEI-FAC without Ag composition, Pd0 (Pd 3d5/2 335.7 eV, Pd 3d3/2 341.2 eV) is shifted to higher bind energy and Pd2+ (Pd 3d5/2, 337.7 eV; Pd 3d3/2, 342.7 eV) remains stable. However, XPS profile of Pd3.75Ag3.75-FAC without PEI decoration shows that only Pd0 (Pd 3d5/2 335.7 eV, Pd 3d3/2 341.2 eV) is observed and oxidized Pd species (Pd2+) are not detected. The probable reason is the fact that Pd and Ag ions are reduced into nanoparticles by NaBH4 before deposition on the suface of the FAC. These phenomena demonstrate the transfer of electrons and the interaction among Pd–Ag bimetals and Pd metal-support [29]. The function of PEI decoration can enhance the binding energy of metal-support and stabilize Ag nanoparticles with small size, which can improve the catalytic performance and durability of Pd3.75Ag3.75-PEI-FAC catalyst. The XPS signal of Ag 3d orbit in Pd3.75Ag3.75-PEI-FAC (Fig. 2c) gives the binding energies of the Ag 3d5/2 peak at 368.0 eV and Ag 3d3/2 peak at 374.0 eV. The 6.0 eV gap between the two states is a typical characteristic of metallic Ag [29]. It reveals that the Ag species in Pd3.75Ag3.75-PEI-FAC exist predominantly in the metallic form. This data is good consistent with the previous report in the literature [30].As seen from X-ray diffraction (XRD) patterns in Fig. 2d, no metal diffraction peaks are observed for PEI-FAC support, which preserves typical cellulose XRD feature of 2θ = 21.6° [31]. Interestingly, no diffraction peak of Pd (111) at 2θ = 40.2° (PDF#46-1043) is found in Pd3.75-PEI-FAC and Pd3.75Ag3.75-PEI-FAC, which may be attributed to atomic dispersed Pd in PEI-FAC support. However, the typical diffraction peak of Ag (111) at 2θ = 38.1° (PDF#04-0783) is observed in Pd3.75Ag3.75-PEI-FAC, confirming the presence of Ag metallic nanoparticle, which is also demonstrated by XPS measurement in Fig. 2c. Based on the above observations, it is reasonably speculated that the Pd3.75Ag3.75-PEI-FAC catalyst has core–shell structure with an Ag nanoparticle as the core in the in-layer and homo-dispersed Pd as the shell in the out-layer. Because the particle size of the catalyst has a significant effect on the catalytic performance, HAADF-STEM was employed to detect the nanoparticle diameter. From the upper TEM images of Pd3.75Ag3.75-PEI-FAC (Fig. 3 a), it can be clearly seen that larger nanoparticles of Ag with the average diameter approx. 8.7 nm (yellow circle), and smaller nanoparticles of Pd with the average diameter approx. 2.3 nm (red circle). The Pd nanoparticle is encased in the surface of Ag nanoparticle. However, from the below TEM images of PdAg-FAC (Fig. 3b), it can be seen that the AgPd nanoparticles were dispersed on the surface of the FAC support with the average diameters of 6.5 nm. Some Ag–Pd alloy was also observed, because the lattice space is 0.231 nm, which is between the (111) lattice spacing of face centered cubic Pd (0.22 nm) and Ag (0.24 nm) [32]. Large Pd nanoparticle in PdAg-FAC shows lower catalytic activity than small Pd nanoparticle in PdAg-PEI-FAC [32].BET surface area and pore diameter of the supported catalyst have a vital role in the catalytic dehydrogenation reaction. BET surface area and pore diameter of the support (PEI-FAC) and catalyst (Pd3.75Ag3.75-PEI-FAC) was tested by nitrogen adsorption desorption isotherms. BET profiles of the support and catalyst in Fig. 4 appear in their hysteresis loops, demonstrating the generation of mesopores. The content of Pd and Ag metal in Pd3.75Ag3.75-PEI-FAC was measured by ICP-OES, and the values are 0.8 wt% and 1.5 wt%, respectively. Pd and Ag contents in other PdAg catalysts with different compositions were shown in Table S2.After illustrating the structural properties of the Pd3.75Ag3.75-PEI-FAC catalyst, we are further evaluating its catalytic activities of hydrogen generation from aqueous FA solution in view of three aspects, the total gas volume (Vgas (H2 + CO2)), turnover frequency (TOF, h−1) and activated energy (Ea, kJ mol−1).The support material has been demonstrated to be one of the most significant factors for the FA dehydrogenation because the interaction between the support and metal would subtly modify the physical and chemical properties of the catalyst and then enhance their activities [8]. It can be seen in Fig. 5 a that Pd3.75Ag3.75-PEI-FAC (using PEI-FAC as support, 112.5 mL Vgas min−1) exhibits almost similar hydrogen generation rate as Pd3.75Ag3.75-rGO (using reduced oxidation graphene as support, 110 mL Vgas min−1), higher than Pd3.75Ag3.75-N@C (using N-doped carbon as support, 90 mL Vgas min−1). However, when black carbon is used as support matrix to load the Pd3.75Ag3.75, it shows no H2 generation activity, indicating N species plays a critical role in H2 evolution from FA. TOF values in Fig. 5b further confirm that Pd3.75Ag3.75-PEI-FAC (2185 h−1) shows the similar TOF value as Pd3.75Ag3.75-rGO (2206 h−1) and Pd3.75Ag3.75-N@C (2179 h−1). Different from rGO and N@C support matrix, PEI-FAC is cost-effective, sustainable and prepared easily from biomass fractionation. From the viewpoint of Vgas data, it has been demonstrated that the amide group (-NH) in PEI-FAC support matrix plays a positive function for FA dehydrogenation activity over Pd3.75Ag3.75-PEI-FAC. The phenomenon was also confirmed by Li and coworker [22].As observed in Fig. 5c and d, the effect of Pd/M compositions of bimetallic catalysts using PEI-FAC as support matrix on catalytic activities of catalyst is significant. After comparative examination of catalytic activities for different bimetallic compositions (Pd/Au, Pd/Ni, Pd/Co, and Pd/Ag) catalysts, Pd2.5Ag2.5-PEI-FAC is found to display the highest catalytic activity with TOF value of 1817 h−1, following with Pd2.5Au2.5-PEI-FAC > Pd2.5Ni2.5-PEI-FAC > Pd2.5Co2.5-PEI-FAC > Pd2.5-PEI-FAC, which may be attributed to electron transfer variation between Pd and adscititious metal elements [25]. The work function of Pd, Au, Ni, Co, and Ag are 5.12 eV, 5.1 eV, 4.6 eV, 5.0 eV and 4.26 eV, respectively [33]. Electrons tend to transfer from elements with lower work functions to those with higher ones. The larger the gap of the work function between Pd and the adscititious metal element is, the better the electron transfer will be. Therefore, the synergistic function between Pd and Ag plays a critical role resulting in the improvement of catalytic performance for Pd3.75Ag3.75-PEI-FAC, this phenomenon is consistent with the reportings of previous literature [23].Three major operational parameters affecting the H2 evolution from FA over Pd3.75Ag3.75-PEI-FAC were optimized, including FA concentration (0.5, 1, 1.5, 2 and 5 mol L−1), reaction temperature (30, 40, 50 and 60 °C) and PdAg mass/weigh ratio (1.25/1.25, 2.5/2.5, 3.75/3.75 and 5/5). The results are shown in Fig. 6 .As shown in Fig. 6a, FA concentration shows negative effects on dehydrogenation activity of catalysts. The higher the FA concentration is, the lower the Vgas generation will be. For instance, approx. 350 mL of Vgas is obtained within 30 min at 1.5 mol L−1 FA, while less than 125 mL of Vgas is released within 30 min at 10 mol L−1 FA. After 20 min of reaction, the average conversion rate of FA to H2 is nearly up to 7.8 mL min−1 mol−1, and the final concentration of FA is too low to detect by HPLC. As seen in Fig. 6b, the reaction rate of hydrogen generation greatly depends on reaction temperature, and 238 mL of Vgas can be readily released within 5 min at 60 °C, corresponding to almost full conversion of FA into H2 and CO2 without the detection of toxic CO by GC measurement. However, the reaction rate of hydrogen generation will be greatly reduced and only ∼80 mL of total Vgas is released in 5 min at 30 °C. Appropriate loading of Pd and Ag compositions in the catalyst is also important for hydrogen generation from FA system. The data in Fig. 6c reveal that Pd3.75Ag3.75-PEI-FAC delivers the highest catalytic activity among all tested PdAg-PEI-FAC catalysts with different mass/weight ratio. Further increases of the PdAg mass/weight ratio up to 5/5 did not result in increased H2 evolution from FA over Pd5Ag5-PEI-FAC is not increasing in comparison with Pd3.75Ag3.75-PEI-FAC.To further explore the catalytic active site and the roles of Ag and the –NH group in Pd3.75Ag3.75-PEI-FAC, the H2 generation from FA by several catalyst counterparts was investigated. The results are shown in Fig. 7 .As seen from Fig. 7a, Ag3.75-PEI-FAC shows no hydrogen generation activity at 60 °C when Ag is solely loaded on the PEI-FAC, indicating Ag is not the active site. On the other hand, when Pd is solely loaded on the PEI-FAC, Pd3.75-PEI-FAC exhibits hydrogen generation activity with total Vgas of 125 mL within 10 min. The two comparative experiments indicate that Pd acts as the active site instead of Ag for the hydrogen generation from the FA system. Apple-to-apple comparison of Pd3.75-PEI-FAC (Vgas of 125 mL in 10 min) and Pd3.75Ag3.75-PEI-FAC (Vgas of 250 mL in 10 min) figures out that adding Ag with appropriate content can improve the catalytic activity of the PdAg bimetallic catalyst due to the synergistic effect between Pd and Ag [13]. Furthermore, significant differences of catalytic activities between Pd3.75Ag3.75-PEI-FAC (Vgas of 250 mL in 10 min) and Pd3.75Ag3.75-FAC (Vgas of ≤ 50 mL in 40 min) demonstrate that the amino group (-NH) of PEI fosters both the C–H cleavage and H2 desorption step in the dehydrogenation of FA [24]. It is universally accepted that the major function of the –NH group is not only to act as a proton scavenger during the catalytic FA dehydrogenation to expedite the C–H cleavage, but also acts as a capping agent to avoid Ag nanoparticles enhancement during PdAg-PEI-FAC syntheses. To our delight, as seen from Fig. 7b, the addition of sodium formate (SF) has a neglective effect on the catalytic activity of Pd3.75Ag3.75-PEI-FAC. It indicates that Pd3.75Ag3.75-PEI-FAC shows excellent hydrogen generation from FA system in SF-free conditions, which is very cost-effective and environmental friendly for H2 production in practical applications.On the basis of the above XRD and HAADF-STEM data, Pd3.75Ag3.75-PEI-FAC might have a core–shell structure with Ag nanoparticles as the core in the in-layer and homo-dispersed Pd as the shell in the out-layer. Highly homo-dispersed Pd and synergistic effects between Pd and Ag of catalysts can improve hydrogen evolution from FA (Figs. 5 and 7). In addition, an amide (–NH) group coated on cellulose surface to act as a proton scavenger can also efficiently enhance the catalytic performance of Pd3.75Ag3.75-PEI-FAC (Figs. 5 and 7). Therefore, a plausible catalytic mechanism pathway over Pd3.75Ag3.75-PEI-FAC was proposed in Fig. 8 .As seen from Fig. 8, three reaction steps will take place for H2 evolution from FA over Pd3.75Ag3.75-PEI-FAC. In step I, O–H bond cleavage provides a proton (H+), and then AgPd-formate (AgPd-[HCOO]-) intermediate is formed; For step II, C–H bond dissociation affords an AgPd hydride (AgPd-[H]-) via isomerization, and one mole of CO2 is released; In step III, H2 evolution occurs through the recombination of AgPd-[H]- with a H+, and the catalyst is regenerated for next reaction. Navlani-García and co-workers pointed out that the positive role of the bimetallic catalyst was to boost C–H bond dissociation (Step II), in which a reduction of 48% of the energy barrier calculated by density functional theory (DFT) for Pd catalyst was achieved by the bimetallic catalyst [5].The TOF value of Pd3.75Ag3.75-PEI-FAC is as high as 2875 h−1, which is superior to most of the previous PdAg bimetallic catalysts supported on other matrices, such as carbon (854 h−1) [34], N-rGO (171 h−1) [6], N-GCNT (413 h−1) [35], g-C3N4 (420 h−1) [36], and graphene (572 h−1) [37], for H2 generation from FA aqueous solution in the literature (Table S3). It indicates that cellulose isolated from biomass after modification with PEI is suitable for PdAg support as a bimetallic catalyst. However, some excellent PdAg heterogeneous catalysts showed much higher TOF values than Pd3.75Ag3.75-PEI-FAC in our work for H2 evolution from FA (Table S3), such as. Ag1Pd9@NPC with TOF of 3000 h−1 [32], Ag1Pd9-MnOx/carbonsphere with TOF of 3558 h−1 [35], PdAg@ZrO2/C with TOF of 9206 h−1 [32], PdAg@ZrO2/rGO with TOF of 4500 h−1 [15], PdAg/amine-MSC with TOF of 5638 h−1 [38], and Pd0.50Ag0.50/PDA-rGO with TOF of 6980 h−1 [35]. Different from Pd3.75Ag3.75-PEI-FAC, all those reported PdAg heterogeneous catalysts needed sodium formate as additive for H2 evolution. From the Arrhenius curve of ln(TOF) versus 1000/T, the apparent activation energy (Ea) of Pd3.75Ag3.75-PEI-FAC catalyst is calculated to be as low as 53.97 ± 2.31 kJ mol−1 (Fig. 9 a). To our delight, Pd3.75Ag3.75-PEI-FAC shows unique robustness and satisfactory re-usability, which is confirmed by the data in Fig. 9b. This catalyst retains excellent stability with no loss of catalytic activity after five recycles (Fig. 9b).In summary, FA fractionation of lignocellulose towards cellulose-derived catalysts has been successfully developed in this work. Owing to formylation, the resulting cellulose is not suitable for directly enzymatic hydrolysis for downstream production of C6 sugars and ethanol fermentation. However, it is a promising alternative option for the resulting cellulose to synthesize cellulose-derived PdAg bimetallic catalyst for hydrogen production from FA aqueous solution at room temperature. Among the as-prepared bimetallic catalysts in hand, Pd3.75Ag3.75-PEI-FAC shows the highest activity with approx. 350 mL of total hydrogen within 30 min from 1.5 mol L−1 FA aquous solution at 60 °C. The TOF value of Pd3.75Ag3.75-PEI-FAC reaches a high value of 2875 h−1, which greatly outperforms most of the previously reported Pd-based heterogenous catalysts for H2 generation from FA aqueous solution in the literature. Furthermore, the apparent activation energy (Ea) of Pd3.75Ag3.75-PEI-FAC is calculated to be 53.97 ± 2.31 kJ mol−1. As expected, Pd3.75Ag3.75-PEI-FAC possesses high selectivity, durability and stability over 5 cycles with no loss of catalytic activity. The findings in this work open a new window for formosolv fractionation of biomass towards cellulose-derived PdAg bimetallic catalyst for hydrogen evolution from FA decomposition at room temperature, taking considerable account into biomass valorization simultaneously.YL (Yun Liu) completed conceptualization and supervision of the project, wrote and revised the manuscript. YY (Yanyan Yu) prepared the draft manuscript. HX (Huanghui Xu) completed methodology and investigation. HY (Hongfei Yu) finished XPS, HAADF-STEM and BET experiments. LH (Lihong Hu) characterized lignin's structure via GPC, FT-IR, and TGA.The authors declare no conflict of interests.This study was financially funded by the National Natural Science Foundation of China (NSFC, 21476016; 21776009), and Fundamental Research Funds for the Central Universities. The authors also acknowledge the special project for the construction of innovative province in Hunan Province of China (2019NK2031-3) The authors acknowledge professor Wensheng Qin, from Lakehead University of Canada, for checking English writing of revised manuscript.The following is the Supplementary data to this article: Multimedia component 1 Multimedia component 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.gee.2020.08.006.

The present work, in which cellulose isolated from formic acid fractionation (FAC) is decorated with polyetherimide (PEI) to attain highly efficient cellulose-derived PdAgbimetallic catalyst (PdAg-PEI-FAC), has been investigated, and the catalyst properties are characterized by XRD, XPS, BET, ICP-AES and HAADF-STEM. The as-obtained Pd3.75Ag3.75-PEI-FAC exhibits excellent catalytic performance for H2 evolution from a sodium formate-free formic acid (FA) aqueous medium at ambient temperature and the turnover frequency (TOF) reaches a high value of 2875 h−1, which is superior to most of the previously reported Pd-based heterogeneous catalysts supported on a carbon matrix in the literature. The remarkable catalytic activities of PdAg-PEI-FAC result from high dispersion Pd and synergistic effects between the PdAg bimetallic system. Furthermore, the amide (-NH) group in PEI coated on cellulose acting as a proton scavenger efficiently improves the catalytic property of catalyst. In addition, the critical factors affecting H2 release, such as FA concentration, reaction temperature, PdAg compositions and support matrix type, are also evaluated. Based on the experimental results, the probable three-step mechanism of H2 evolution from FA over Pd3.75Ag3.75-PEI-FAC is proposed. In the end, the activation energy (Ea) of Pd3.75Ag3.75-PEI-FAC catalyst is calculated to 53.97 kJ mol-1, and this catalyst shows unique robustness and satisfactory re-usability with no loss of catalytic activity after five recycles. The findings in this work provide a novel routine from lignocellulose fractionation towards cellulose-derived catalyst for H2 evolution.

As one of the most important inorganic chemicals, ammonia is responsible for supporting approximately 27 % of global population [1,2]. The Haber-Bosch process for ammonia synthesis is regarded as one of the greatest inventions of the last century [1]. Since then, a great deal of effort has been made to develop new ammonia synthesis catalysts to allow the operation of the Haber-Bosch process at reduced temperature and pressure. Intensive investigations have been carried on Ru-based catalysts promoted through a variety of different support materials Ru/HT-C12A7:e− [3], Ru/ Ba-Ca(NH2)2 [4], Ru/BaTiO3-xHx [5], Ru/TiH2, Ru/BaTiO2.5H0.5 [6] and Ru/BaCeO3-xNyHz [7]. Atomically dispersed Co supported on N-doped hollow carbon spheres also exhibit excellent catalytic activity at 350 °C [8]. Novel ammonia synthesis methods including electrochemical [9,10] and photocatalytic synthesis [11] allow ammonia synthesis at ambient conditions. The promotional effects of applied electric fields to ammonia synthesis have been reported [12]. The use of single atom catalysts for electrochemical synthesis of ammonia has been investigated through density functional theory calculations [13]. Due to the high cost of Ru, large scale application is limited with around ten ammonia synthesis plants using Ru-based catalysts globally (in which some plants are combined with Fe-catalysts too), all remaining plants use cheap fused Fe catalysts. A large Haber-Bosch ammonia synthesis plant may need 300 tons of catalyst thus cost is extremely important. Metal hydrides, oxyhydride and oxynitride hydride have been investigated as efficient promoters/supports for Ru, Ni, Fe and Co-based catalysts [5,7,14,15]. In general, metal hydrides are sensitive to air and moisture which may limit their practical large scale applications [15,16]. In laboratory conditions, most of the electride or hydride-based catalysts are handled in a glove-box to avoid their reaction with H2O [4–7]. In conclusion, no matter whether the catalyst is Fe/Co/Ni or Ru-based, they still fall short in terms of cost, moisture and oxygenate tolerance, therefore further improvements in this key area is still required.In ammonia industry, whether using Fe or Ru-based catalysts, a heavy gas purification process is applied to purify the feed gases, H2 and N2, to avoid catalyst poisoning. Trace amounts of oxygenates (as low as 10 atomic oxygen) such as O2, H2O, CO, and CO2 will deactivate the Fe-based catalyst [16–22]. In a recent report, it has been demonstrated that even impurities below 1 ppm of oxygen lead to a significant loss in activity for a state-of-the-art multi-promoted iron-based industrial catalyst [21]. A high purity gas feed of over 99.99995 % (0.5 ppm impurity), is normally used in reported papers for Haber-Bosch processes [4–7]. Inevitable intensive gas purification of both H2 and N2 will lead to a relevant increase in capital investment on the facility as well as additional energy inputs, lowering overall efficiency. One of the strategies to improve the oxygenate tolerance of the Fe or Ru catalysts is to prevent the particle growth, under the ammonia synthesis conditions through strong metal support interaction (SMSI) [22,23]. It has been reported that the strong interaction between Ru and defects in CNTs can significantly improve the catalytic activity of Ru-based catalyst for ammonia synthesis [24].In conventional fused Fe-based industrial catalysts, there are strong interactions between iron and the oxygen vacancies in the oxide promoters, although this intrinsic oxygen vacancy concentration is limited [18,22]. We previously reported that Ni promoted by BaZr0.1Ce0.7Y0.2O3− δ and Fe promoted by Ce0.8Sm0.2O2- δ catalysts, display good ammonia synthesis activities due to the key role of extrinsic oxygen vacancies [22,25]. Anion vacancies, in particular nitrogen vacancy containing materials, provide the next step in this concept [26]. For the synthesis of ammonia through the Haber-Bosch process, the important role of nitrogen vacancies was also observed by Hosono and his co-workers in Fe, Co, Ru catalysts supported on Ba-CeO3-xNyHz and Ni supported on LaN, with it found that nitrogen vacancies on LaN can efficiently bind and activate N2 [7,27]. Using CeO2 as an example, on heating up to high temperatures, intrinsic oxygen vacancies will be generated through the reduction of Ce4+ ions and the loss of lattice oxygen atoms. (1) 2 C e C e × + O O × = 2 C e C e ' + V O ⋅ ⋅ + 1 2 O 2 ( g ) Here Kroger-Vink notations for defect chemistry is used in this article [28].Under the ammonia synthesis conditions, due to the presence of H2 at high temperature, some CeO2 will be reduced to oxygen deficient CeO2-δ thus forming more oxygen vacancies. (2) 2 C e C e × + O O × + H 2 ( g ) = 2 C e C e ' + V O ⋅ ⋅ + H 2 O ( g ) These oxygen vacancies will be a kind of nest, able to form strong interactions with the Fe or other transition metal atoms via SMSI, similar to the case for the Ag - CeO2 system [22,23,29]. Similarly, Goula et al. reported excellent catalytic activity and stability for Ni supported on Sm2O3, Pr2O3 and MgO promoted (doped) CeO2, attributed to the high concentration of oxygen vacancies [30].Oxygen vacancies introduced through thermal treatment or reduction are referred to as intrinsic oxygen vacancies. Their concentration is greatly related to temperature and oxygen partial pressure. If the promotion effects were solely related to intrinsic oxygen vacancies, it would be very limited. However, oxygen vacancies can be deliberately introduced into oxides through doping with another metal oxide with lower metal valence than the parent oxides theoretically existing in any temperature range, provided a stable solid solution is formed. These oxygen vacancies are referred to as extrinsic oxygen vacancies. The concentration of oxygen vacancies can be controlled through adjusting the doping level within the solid solution limit while more oxygen vacancies will be generated at higher doping levels. This technology is used for solid oxide fuel cells (SOFCs) and other electrochemical devices. For example, Sm2O3 can be used to dope CeO2, generating extrinsic oxygen vacancies, (3) 1 - z C e O 2 + z S m O 1.5 = 1 - z C e C e × + 2 - 0.5 z O O × + z S m C e ' + 0.5 z V O ∙ ∙ In order to maximize the concentration of oxygen vacancies, doping the oxygen with another more negatively charged anion, such as N3− ions is another option. It has been reported that when firing a porous CeO2 membrane in NH3 at 550 °C, N-doped CeO2 i.e. CeO2-xNy was formed, with x = 0.1 [31]. During the activation of the ammonia synthesis catalysts in mixed N2 and H2, it is inevitable that ammonia will be generated, further reacting with CeO2 or doped CeO2 to form N-doped CeO2, as shown in the reaction below. (4) C e O 2 + 2 x 3 N H 3 = C e C e × + 2 - x O O × + 2 x 3 N O ' + x 3 V O ∙ ∙ + x H 2 O When Sm-doped CeO2, Ce1-zSmzO2-δ, is exposed to NH3 at high temperature, cation Sm3+ and anion N3− co-doped CeO2, Ce1-zSmzO2-xNy, oxynitrides will be formed which have a higher concentration of oxygen/anion vacancies. Due to the second type of anions, N3− ions, the vacancies are more precisely described as anion vacancies, which could be either oxygen vacancies, V O ∙ ∙ or/and nitrogen vacancies, V N ∙ ∙ ∙ . Fig. 1 shows the diagram to maximize the anion vacancies in CeO2 through both cation and anion co-doping. From the charge balance, more anion vacancies will be generated when some O2- ions are further replaced by lower valence ions such as N3− ions. Co-doping of Sm3+ and N3− ions in CeO2 will maximise the generation of extrinsic anion vacancies. It is anticipated that a significantly greater promotion effects can be achieved by using materials with more tailorable extrinsic anion vacancies such as Sm-doped cerium oxynitrides, Ce1-zSmzO2-xNy. At high anion vacancy concentrations the interaction between the positively charged anion vacancies and the electron-rich metal particles will be stronger, adding to an already strong SMSI, the nested or anchored metal particles on the anion vacancies will inhibit sintering and growth, improving metal catalyst stability [22,32–34].Another poisoning mechanism of oxygenates is the competitive adsorption between oxygen and nitrogen on the catalyst surface, with oxygen occupying active sites limiting catalytic activity [17]. The presence of large amounts of extrinsic anion vacancies on the surface of Sm-doped cerium oxynitrides will provide more active sites (anion vacancies) available for the adsorption of oxygenates such as O2, and H2O, reducing poisoning severity on the metal catalyst while retaining high activity even at high concentrations of oxygenate impurities. CeO2-based oxides have been reported as excellent oxygen storage materials for efficient catalytical conversion of H2, CO, CO2 and hydrocarbons [[35–38].Nitrogen is larger than oxygen, no matter in atomic or ionic format. At a coordination number of 4 (CN = 4), the environment of anions in fluorite structure, the ionic size of O2− and N3- ions is 1.38 and 1.46 Å respectively [39]. Thus, the size of the nitrogen atom in N2 may match better with nitrogen vacancies than oxygen vacancies, introducing extra nitrogen vacancies may further improve the promotion effect for efficient synthesis of ammonia through the Haber-Bosch process. Therefore, in this study, Sm3+ and N3- co-doped CeO2, Sm-doped cerium oxynitrides (Ce1-zSmzO2-xNy) with z ≤ 0.5 were synthesized and their promotion effects on ammonia synthesis were investigated in detail. The Ce1-zSmzO2-xNy promoter was synthesized in a simple one-step combustion synthesis method carried out in air. It is found that both the stability and catalytic activity of iron based catalysts have been improved by deliberately introducing extrinsic anion (oxygen and nitrogen) vacancies into the Ce1-zSmzO2-xNy promoter/co-catalyst.CeO2-xNy was prepared from a mixture of Ce(NO3)3·6H2O (99.5 % Alfa Aesar) and urea with a mole ratio of 1 to 10 respectively [40]. 50 mL of water was then added to the mixture in a ceramic evaporating dish. The mixture was then treated at 120 °C for 24 h to form a gel like product which was combusted at 400 °C to for the desired CeO2-xNy powder. Part of this sample was then further calcined at 550 °C for 3 h in air to form the calcined CeO2-xNy powder. The resulting CeO2-xNy powder was then mechanically mixed with Fe2O3 and reduced using the same method previously reported [22].The solubility limit of Sm2O3 in Ce1-zSmzO2-δ is roughly at z = 0.5 [41]. The CeO2-xNy and Ce1-zSmzO2-xNy with z ≤ 0.5 were synthesised from cerium nitrate, samarium nitrate and urea through a simple combustion method [40]. Ce1-zSmzO2-xNy was prepared from a mixture of Ce(NO3)3·6H2O (99.5 % Alfa Aesar), Sm(NO3)3⋅6H2O (99.9 % Alfa Aesar), and urea with a ratio of 1 mole of total metal ions to 10 mole of urea. The rest of the synthesis process was the same as described above for CeO2-xNy and was repeated for z values of 0.1, 0.2, 0.3, 0.4, and 0.5. The rest is the same as for preparation of CeO2-xNy. The combustion method was used to prepare pure CeO2 as previously reported [22].Before materials characterisation, all samples were washed multiple times by water and ethanol to remove any residual urea or other hydrocarbons. The catalyst was characterised using X-ray Diffraction (XRD), X-ray photoelectron spectroscopy (XPS), Carbon, Hydrogen and Nitrogen (CHN), Raman spectroscopy, Scanning electron microscopy (SEM) and high resolution transmission electron microscopy (HRTEM). XRD analyses were carried out on a Panalytical X’Pert Pro Multi-Purpose Diffractometer (MPD), with Cu K α1 radiation, working at 45 kV and 40 mA. The nitrogen content was measured by a CHN analysis, performed on a FlashEA® 1112 Element Analyzer at MEDAC Ltd adhering to UKAS ISO17025 accreditation, with a standard deviation of ±0.3 wt%. The same CHN facility has previously been used to determine the nitrogen content in ammonia synthesis catalyst (Ni,M)2Mo3N (M = Cu or Fe) [42]. The SEM images were obtained with ZEISS SUPRA 55-VP operating at 10 kV. Elemental compositions were analysed with an energy-dispersive X-ray spectrometer (EDX) attached to the SEM. X-ray Photoelectron Spectroscopy (XPS) data were collected at the Warwick Photoemission Facility, University of Warwick. Samples were attached to electrically conductive carbon tape and mounted on a sample bar loaded in to a Kratos Axis Ultra DLD spectrometer (base pressure < 2 × 10−10 mbar). Samples were illuminated using a monochromated Al Kα X-ray source (hv = 1486.7 eV). The core level spectra were recorded using a pass energy of 20 eV (resolution approx. 0.4 eV), from an analysis area of 300 μm x 700 μm. The data were analysed in the CasaXPS package, using Shirley backgrounds and mixed Gaussian-Lorentzian (Voigt) lineshapes. High resolution transmission electron microscopy (HRTEM) observations of the samples were carried out on a JEOL2100 microscope, operated at 200 kV, equipped with an Oxford Instruments 80 mm2 SDD EDX detector. EDX spectra were collected by focusing the electron beam onto a certain area of the samples.To measure the catalytic activity, a fixed bed stainless steel reactor was used with the catalyst held in place in the centre by quartz granules and glass fibre. The total weight after activation / reduction was approximately 0.3 g for each catalyst tested. Full details on the testing parameters used along with the gas purification process can be found in our previous report [22]. The data points in Fig. 9 were obtained through the purification and known impurity injection process detailed in our previous work [22].Produced ammonia was collected in dilute sulfuric acid (0.01 M) with concentration measured using an ISE Thermo Scientific Orion Star A214 ammonia meter [25]. The rate of ammonia production (in mol g−1 h−1) was calculated according to the following equation. (5) r N H 3 = N H 4 + × V t × m Where [NH+ 4] is ammonia concentration in mol L−1, V is volume of 0.01 M H2SO4 in L, t is time in hours and m is catalyst mass in g.Carbon, Hydrogen and Nitrogen (CHN) analysis was carried out to precisely measure the overall nitrogen content in the synthesised oxynitrides. In our synthesis process of cerium oxynitride, urea was used as the nitrogen source. It is believed that, during the combustion process, ammonia is generated from urea to further react with CeO2 to form cerium oxynitride, CeO2-xNy. Assuming the charge for cerium is +4, from the charge balance, the formula of the cerium oxynitride can be written as C e O 2 - x N 2 x 3 V x 3 , where ‘V’ represents anion vacancies, oxygen or/and nitrogen vacancies. For Sm-doped cerium oxynitrides, the general formula can be written as C e 1 - z S m z O 2 - y + z 2 N y V 3 y + z 2 . As z is known from the starting materials, and since the weight percentage of nitrogen in the synthesised oxynitrides is measured, y can be then deduced. Therefore, the general formulae of these new oxynitrides can be determined based on the nitrogen content, which are listed in Table 1 . For samples with z = 0 and 0.1, the nitrogen content is y = 0.07 for anion sites for both samples (Table 1). However, the anion vacancy concentration for the Sm-doped sample, Ce0.9Sm0.1O1.84N0.07, is 4.5 %, which is slightly higher than that of the Sm-free sample, CeO1.89N0.07 (3.5 %). This is due to the doping of low valent Sm3+ cations in the Ce0.9Sm0.1O1.84N0.07 sample, thus more anion vacancies are generated (Fig. 1, eq. 3). According to the data in Table 1, in general, the nitrogen content and anion vacancies increase with increased Sm-doping level. The only exceptional situation is that, the nitrogen content in sample Ce0.5Sm0.5O1.51N0.16 is slightly lower than that for sample Ce0.6Sm0.4O1.52N0.19. The possible reason is, at z = 0.5, the cation doping level is already very high leading to a high concentration of anion vacancies. If more oxygen is replaced by nitrogen, more anion vacancies will be generated. However, there is a limit on the anion vacancy concentration in the oxynitride in order to maintain the crystal lattice. Under this circumstance the nitrogen content is slightly reduced for the sample Ce0.5Sm0.5O1.51N0.16. For representative sample Ce0.5Sm0.5O2-xNy, the compositions were also estimated using X-ray Photoelectron Spectroscopy (XPS). Fig. 2 shows the selected XPS data for pure Ce0.5Sm0.5O2-xNy promoter plus the Ce0.5Sm0.5O2-xNy promoter with Fe2O3 and Fe before and after the activity test. For the pure oxynitride, XPS revealed a composition of Ce0.5Sm0.5O1.45N0.12, a slightly lower nitrogen content than that derived from the CHN results. The ratio of Ce(IV) to Ce(III) was found to be 3.8, with a corresponding ratio of Sm(III) to Sm(II) of 2.6, the XPS spectrum of Ce 3d and Sm 3d are shown in Figure S1. Analysis of the N 1s spectrum acquired from the Ce0.5Sm0.5O2-xNy sample is shown in Fig. 2b and corresponds to a nitrogen content of 4.67 at% (Table 1). The N 1s spectrum is further shown in Figure S1c, showing large deviation in the exp count values. This indicates that the nature of the nitrogen bonds obtained from the XPS spectra is unreliable with the main conclusion being the presence of nitrogen. The C 1s spectrum is shown in Figure S1d showing no CN bonds confirming no residual urea present in the sample. The discrepancy in nitrogen content derived from XPS and CHN is due to the surface specificity of XPS. While CHN measures the overall nitrogen content, the sampling depth in XPS is limited to the outermost few nm of the material and thus a little deviation is common. The deduced overall unoccupied anion sites, i.e., the total anionic vacancies is the same (16.5 %), due to the detection of Ce(III) and Sm(II) in the XPS results, allowing us to remove the assumption that all cerium has a charge of +4 and all Sm has a charge of +3 (Table S1).The compositions of the Fe2O3 and Fe promoted catalysts are less reliable with Ce to Sm ratio varying drastically from the expected value. This is due to the overlap between Ce 3d and the Fe Auger emission as well as Ce 4d and Sm 4d, both of which make the determination of the amount of Ce less accurate. To resolve this, a sample with low Ce content was examined in order to obtain a reliable line shape for the Fe LM23M23 Auger emission. For the two promoted catalysts, Fe2O3 and Fe before and after the activity test, examination of the Sm 3d region showed no Sm(II) present, with all samarium being Sm(III). Although no Sm(II) was detected on the surface in these samples it should be noted that the weight percent of oxynitride in these samples is only 20 wt%, much lower than in the pure Ce0.5Sm0.5O2-xNy sample and therefore signal for Sm(II) could be below the detection limit. However, for the calculation of composition it was assumed that only Sm(III) was present. The ratio of Ce(IV) to Ce(III) was 0.87 and 3.7 in 85 wt% Fe2O3 – 15 wt % Ce0.5Sm0.5O2-xNy and Fe – 20 wt % Ce0.5Sm0.5O2-xNy respectively. This deviation from the pure oxynitride sample was expected to be due to the overlap of emission spectra as described above. For the mixed 85 wt% Fe2O3 – 15 wt % Ce0.5Sm0.5O2-xNy, the Fe spectra is shown in Fig. 2c. All Fe is in the form of Fe2O3. After the catalytic activity test, the Fe spectra for the 80 wt%Fe – 20 wt % Ce0.5Sm0.5O2-xNy catalyst is shown in Fig. 2e. 76 % of this region is made up of Fe2O3 which is expected due to the reoxidation of the small Fe catalyst particles upon removal of the catalyst from the reactor. 21 % of the region is Fe (II) with the remaining 3 % being zero valence Fe(0). However, after reduction and the catalytic activity test, no nitrogen is detected in the 80 wt%Fe – 20 wt % Ce0.5Sm0.5O2-xNy sample. One possibility is the nitrogen content is too low, beyond the measuring limit of XPS. Another possible reason is, the oxynitride sample is partially oxidized by the oxygenates present in the gas stream. However, the Sm-doped CeO2, Ce1-zSmzO2-δ, still exhibit good stability and activity for ammonia synthesis, which is related to the extrinsic oxygen vacancies, although the overall activity is slightly lower [22].X-ray diffraction (XRD) was used to determine the phase and structure of the synthesised oxynitrides. As shown in Fig. 3 a, the XRD patterns of pure and Sm-doped CeO2-xNy are similar to CeO2, indicating they have the same or a similar crystal structure to CeO2. Rietveld refinement of these oxynitrides were carried out by GSAS + EXPGUI using the fluorite structure for pure CeO2 as the parent phase (Figure S2) [43,44]. During the refinement, cubic CeO2 with a space group F m 3 ¯ m   ( 225 ) was used as the parent phase. It was assumed that Ce and Sm share the 4a (0,0,0) sites, O and N share the 8c (1/4,1/4,1/4) sites [45]. The oxygen and nitrogen occupancies were taken from the chemical composition of these oxynitrides measured by CHN analysis because CHN can provide the overall nitrogen content while XPS can only provide the information on the surface (Table 1). The real and calculated XRD patterns provide a good fit, indicating all these new materials are single phase. The lattice parameters, and cell volume thermal factors are listed in Table 2 . The lattice parameters and cell volumes of Ce1-zSmzO2-xNy with z = 0 to 0.5 are also shown in Fig. 3b. It is believed that oxygen and nitrogen share the same 8c sites, ordering of nitrogen and oxygen, as has been observed in some oxynitrides, but does not happen in Ce1-zSmzO2-xNy. It is then reasonably deduced that nitrogen is homogeneously distributed in the bulk, although the defect concentration including anion vacancies are normally higher on the surface of a particle. From XPS analysis of sample Ce0.5Sm0.5O2-xNy, un-occupied anion sites is 16.5 %, which is the same as deduced from CHN analysis (Table 1). The crystal structure of Ce0.5Sm0.5O2-xNy is also consistent with the observed d-spacing from high resolution transmission electron microscopy (HRTEM) (Fig. 4 a), indicating it is correct. To the best of our knowledge, Sm-doped CeO2-xNy is the first cation doped fluorite oxynitride.It is noticed that the lattice parameter change in Ce1-zSmzO2-xNy with z = 0 to 0.5 does not follow the Vegard's law, i.e., the lattice parameter change should be proportional to the change of z in Ce1-zSmzO2-xNy (Fig. 3b). As Ce1-zSmzO2-xNy is both a cation (Sm3+) and anion (N3−) co-doped solid state solution, which is more complicated. It may not necessarily follow Vegard's law, which normally applies to only cation doped materials. According to CHN analysis, based on the contents of nitrogen, it can be deduced that the formula for Ce1-zSmzO2-xNy with z = 0 and 0.1 is CeO1.89N0.07 and Ce0.9Sm0.1O1.84N0.07 respectively (Table 1). The ionic size for Ce4+ and Sm3+ ions at coordination number of 8 (CN = 8) is 0.97 and 1.079 Å respectively [39]. Doping of CeO2 by larger Sm3+ ions should lead to the lattice expansion. This has been previously observed in Sm-doped CeO2 [46]. However, the introduction of nitrogen in the lattice makes things more complicated. The lattice parameter of the synthesized CeO2-xNy is a = 5.4273(1) Å. This is slightly lower than the reported a = 5.5133(1) Å for pure CeO2 [47]. The ionic size of N3- ions is larger than O2- ions [39]. From this point of view, partially replacing O2- ions with larger N3- ions in CeO2 should lead to an increased lattice parameter. On the other hand, this anion doping also generates anion vacancies (charged voids), as shown in Eq. (4). This may result in lattice shrinking. The final lattice parameters of CeO2-xNy is the combined effects of lattice expansion due to larger N3- ions and lattice shrinking due to the formation of anion vacancies. Lattice shrinking was also observed in some perovskite oxynitrides where some lattice O2- ions are replaced by large N3- ions, which is attributed to the formation of higher valent cation ions [48]. XPS analyses indicate both Ce4+ and Ce3+ for element Ce, and Sm3+ and Sm2+ for element Sm, exist on the surface of pure Ce0.5Sm0.5O2-xNy (Table S1), thus no higher valent cations are formed in Ce0.5Sm0.5O2-xNy. A similar situation may happen on sample Ce0.9Sm0.1O2-xNy, which means that the lattice shrinking for sample Ce0.9Sm0.1O2-xNy is likely due to the extra anion vacancies due to the doping of more negative N3− ions in the lattice (Fig. 3b). However, from z = 0.1 to 0.5, the lattice parameter gradually increases indicating that the effect of larger Sm3+ ions on the lattice parameters becomes more significant than that from anion vacancies (Fig. 3b). These results indicate that the relationship between lattice parameters and doping level in both the cation and anion-co-doped CeO2 is very complicated, and may not necessarily follow the Vegard's law.To study the oxygen vacancies in the as-prepared oxynitrides, Raman spectra of these samples were collected (Fig. 3c). Pure CeO2, and raw CeO2-xNy show a sharp F2g peak at 465 cm−1, corresponding to the typical fluorite structure of CeO2 [40,49,50]. The peak at 570 cm−1 is attributed to oxygen vacancies [40,50]. No peak was observed at 570 cm−1 for pure CeO2 indicating that for pure CeO2, intrinsic oxygen vacancy concentration is not high enough to be detected by the Raman spectroscope. The peak at 570 cm−1 for raw CeO2-xNy and Ce0.9Sm0.1O2-xNy is very weak, indicating a low concentration of oxygen vacancies. This peak becomes stronger with increased Sm doping level, indicating a higher concentration of oxygen vacancies. Sample Ce0.5Sm0.5O2-xNy with the highest doping level gives the strongest peak at 570 cm−1 indicating the highest concentration of oxygen/nitrogen vacancies. These results are consistent with the deduced chemical formula from CHN analyses and the corresponding anion vacancy concertation (Table 1).SEM imaging was employed to further examine the surface of the catalyst promoter as well as the promoted iron catalyst before and after reduction, EDS layering allowed for a clear distinction between promoter and iron catalyst to be seen, giving a further picture of how the promoter is distributed within the catalyst. The SEM/EDS images of sample CeO2-xNy are shown in Figure S3. Figure S3a shows the image for CeO2-xNy with EDS layering, it is observed from this images that there is a secondary particle size distribution of approximately 1–6 μm. In Figures S3b, 3c and 3d the unreduced catalyst, catalyst after carrying out the activity test, and catalyst after carrying out the stability test are shown. The size distribution of the CeO2-xNy promoter does not change showing good stability throughout the reduction and reaction processes. Figure S4 shows the SEM/EDS images of sample Ce0.5Sm0.5O2-xNy along with the unreduced α-Fe2O3 - Ce0.5Sm0.5O2-xNy promoted catalyst and the reduced catalyst after both the activity and stability test. It can be seen from Figures S4a and 4b that Ce0.5Sm0.5O2-xNy has a similar structure to that of CeO2-xNy with a secondary particle size distribution of approximately 1–10 μm. A similar particle size distribution can be seen for Ce0.5Sm0.5O2-xNy before reduction as well as for both reduced catalysts after activity and stability tests showing good stability throughout this process, again similar to CeO2-xNy.HRTEM images of the mixed 85 wt% Fe2O3-15 wt% Ce0.5Sm0.5O2-xNy catalyst before and after catalytic activity measurements are shown in Figs. 4a & 4b, respectively. In Fig. 4a, the closest match to 0.294 nm is the (111) spacing of Ce0.5Sm0.5O2-xNy (0.3134 nm) and, for 0.482 nm it is the (003) spacing of α-Fe2O3 (0.4582 nm). This indicates the crystal structure determined by Rietvild refinement for sample Ce0.5Sm0.5O2-xNy is correct. In Fig. 4b, the closest match to 0.318 nm is Ce0.5Sm0.5O2-xNy (111) (0.3134 nm), for 0.271 nm it is Ce0.5Sm0.5O2-xNy (200) (0.271 nm), and for 0.44 nm it is α-Fe2O3 (100) (0.436 nm). This is consistent with the XRD pattern of the Fe- Ce0.5Sm0.5O2-xNy catalyst after the catalytic activity measurement (Figure S5b). From XPS analyses, only 2.6at% of iron is in metallic Fe form in the sample after the catalytic measurement with the rest reoxidized by air when removed from the reactor at room temperature. Therefore it is difficult to find metallic Fe particles during TEM observations (Table S1). In Fig. 4b it can be seen that small α-Fe2O3 is present on the surface of Ce0.5Sm0.5O2-xNy particles providing indirect evidence of anchoring of Fe in Ce0.5Sm0.5O2-xNy (or Ce0.5Sm0.5O2-δ after oxynitride is oxidised to oxide by oxygenates) through SMSI. Iron is expected to be in the form of metallic Fe under ammonia synthesis conditions. However, when removed from the reactor the small particle size of metallic iron will cause re-oxidation, which has been confirmed by XPS analyses (Fig. 2 and Table S1). Figure S6 shows the TEM images with EDX for the 85 wt% Fe2O3-15 wt% Ce0.5Sm0.5O2-xNy catalyst before and after catalytic measurements. Before the catalytic measurement, it is a mixture of Fe2O3 and Ce0.5Sm0.5O2-xNy resulting from the mechanical mixture (Figure S5b & S6a). Element nitrogen was not detected by EDX because the content is too low, beyond the measuring limit of EDX. Figure S6b shows the presence of small particle sized Fe2O3 on the Ce0.5Sm0.5O2-xNy surface while a large portion of Fe or FeOx is not in direct contact with Ce0.5Sm0.5O2-xNy because it contains only 20 wt% in the composite.From the analyses above, single phase doped oxynitrides, Ce1-zSmzO2-xNy with a large amount of extrinsic anion vacancies have been successfully synthesised and confirmed. They are expected to be excellent promoters/co-catalysts for ammonia synthesis catalysts, which are investigated in detail below.These oxynitrides were investigated as promoters with a Fe catalyst for the synthesis of ammonia. The experimental details are described in the experimental section. In our experiments it was found that, under 3 MPa, pure Fe catalyst using α- Fe2O3 as the precursor, pure CeO2, pure Sm2O3, CeO2-xNy and Ce0.5Sm0.5O2-xNy all show no activity towards ammonia synthesis on their own at temperatures up to 600 °C with feed gas purity of 99.996 %. When α-Fe2O3 was mixed with CeO2-xNy, ammonia was successfully generated (Fig. 5 a&5b). This indicates the catalytic activity is a synergetic process between the α-Fe catalyst and the oxynitride promoter/co-catalyst. The oxynitride may not be a simple promoter as pure Fe from reduction of α-Fe2O3 itself does not show any activity. The oxynitride in the Fe-oxynitride composite is more likely a co-catalyst, which needs further investigation.For a synergetic process, normally there is an optimised ratio between Fe and the oxynitride, which exhibits the best catalytic activity. To find an optimal ratio between the Fe catalyst and the oxynitride promoter, the weight of oxynitride promoter was varied from 14 wt% to 26 wt% in the total Fe-oxynitride composite catalysts. The catalytic activities of these Fe-based catalysts were initially measured in a fixed bed reactor using BOC Zero grade N2 and H2 as the feed gases without further purification. The impurity level of these gases has been listed in a previous report [22]. Unless specified, all the activities were obtained from Zero grade feed gases. Fig. 5a & b show the ammonia synthesis activity of CeO2-xNy promoted Fe catalysts at different weight percentages over a temperature range of 600 °C - 250 °C, at 1 MPa and 3 MPa respectively. Apart from 77 % Fe - 23 % CeO2-xNy and 74 % Fe - 26 % CeO2-xNy, both achieving their maximum activity at 500 °C and 1 MPa, all the other experiments suggested an optimum operation temperature of 450 °C. At 3 MPa the highest activity was 17.2 mmol g−1 h−1 with the optimum weight ratio of 80 % Fe – 20 % CeO2-xNy. At 1 MPa, the highest activity was again achieved for 80 % Fe – 20 % CeO2-xNy, 8.86 mmol g−1 h−1. The optimum mass ratio of Fe to CeO2-xNy is found to be 80:20, as could be reasonably expected according to the synergetic process between Fe and oxynitride promoter. It is expected that the greater the oxynitride content then the stronger SMSI effect will be, as described above [29]. More oxynitride means more anion vacancies thus higher activity. However, the content of Fe is also important as Fe is the actual catalyst or a co-catalyst. Therefore, if Fe is diluted too much then activity will be lower. A balance between these two effects can be achieved at an oxynitride weight percent of 20 %. The SMSI between Fe and the anion vacancies and the possible in situ Ce-H species formation on the CeO2-xNy surface possibly donating electrons to the nested/anchored α-Fe particle, facilitating the dissociation of N N bonds, thus improving the ammonia synthesis reaction [7]. The anion vacancies in CeO2-xNy may adsorb N2, which takes part in the ammonia synthesis process as proposed in nitride (Co3Mo3N, LaN), perovskite oxynitride hydride (BaCeO3−xNyHz) catalysts [7,27,51].The apparent activation energy (Ea) of the Fe-CeO2-xNy composite catalysts at a temperature below 450 °C is obtained from the slope when plotting the logarithm of the ammonia synthesis rate vs. 1000/T (Fig. 6 a & b) [5,6]. At a temperature above 450 °C, limited by the thermodynamic equilibrium and the greater thermal decomposition of ammonia, the activity of Fe-based catalyst normally starts to decrease at a temperature between 450 – 500 °C [4]. Due to limited data-points at low temperature, some of the obtained apparent activation energies may have a relatively large deviation. Considering the Ea for different compositions at both 1 MPa and 3 MPa, the 80 wt% Fe-20 wt% CeO2-xNy composite catalyst tends to have low apparent activation energy and high activity (Figs. 6a &b). The Ea for 80 wt% Fe-20 wt% CeO2-xNy composite catalyst is 66 ± 4.78 and 50 ± 8.22 kJ/mol at 1 MPa and 3 MPa respectively. This is comparable to the Ea of 70 kJ/mol for fused industrial Fe-catalyst (Haldor Topsoe KM1) at low pressure [5,52]. For ammonia synthesis catalysts using excellent promoters, a low apparent activity energy around 50 kJ/mol is normally observed [5,7,14,27,53]. As the 80 wt% Fe catalyst in the Fe-CeO2-xNy composites exhibits the optimum highest activity, the mass ratio between Fe and oxynitride is fixed to 80:20 in the composite catalysts in the following study.From previous reports and the analyses above, anion vacancies play a crucial role in the stability and catalytic activity of ammonia synthesis catalysts. To further increase anion vacancies, new Sm-doped cerium oxynitrides of Ce1-zSmzO2-xNy with z = 0.1 to 0.5 were synthesised. Catalytic activity of the Fe- Ce1-zSmzO2-xNy composite catalysts with mass ratio of Fe-catalyst to Ce1-zSmzO2-xNy of 80:20 at different temperatures and pressures, 1 MPa, 3 MPa were investigated respectively (Fig. 5c & d). At both 1 MPa and 3 MPa, the sample 80 wt%Fe-20 wt%Ce0.5Sm0.5O2-xNy where z = 0.5 achieved the highest activity. The introduction of samarium into the CeO2-xNy promoter/co-catalyst will create extrinsic anion vacancies, confirmed by Raman spectra (Fig. 3c). From CHN analyses, nitrogen content in Ce1-zSmzO2-xNy where x > 0.2 is significantly higher than in the other samples, indicating the introduction of appropriate amounts of Sm3+ ions in CeO2 also facilitate the incorporation of nitrogen into the lattice (Table 1). According to Eq. (4), the concentration of negatively charged nitrogen defects N O ' will also be higher in samples with high nitrogen content. These negatively charged nitrogen defects, similar to negatively charged H− ions, theoretically may donate electrons to nearby Fe particles, helping in the dissociation of strong N N bonds, thereby leading to higher activities. At 3 MPa, the optimum temperature at which the highest activity was achieved was 500 °C for Fe-Ce0.8Sm0.2O2-xNy, 400 °C for Fe-Ce0.5Sm0.5O2-xNy, and 450 °C for the other Ce1-zSmzO2-xNy promoted Fe-catalysts (Fig. 5c & d). The optimum temperature of Fe-Ce0.5Sm0.5O2-xNy is similar to the expensive Ru-based catalysts [16]. It is noted that sample Ce0.5Sm0.5O2-xNy has the highest concentration of anion vacancies, which could be related to the lower optimum operating temperature. At 1 MPa the difference in activities is much less, although the highest activity was still obtained for the Fe-Ce0.5Sm0.5O2-xNy catalyst. At 3 MPa the highest activity of 18.8 mmol g-1 h-1 at 400 °C was obtained from the 80 % Fe – 20 % Ce0.5Sm0.5O2-xNy catalyst. The weight hourly space velocity (WHSV) is 16000 mL g-1 h-1, which is less than half of those in most of the reported papers (Table S2). This low WHSV is a result of the high catalyst loading (300 mg vs 100 mg) and larger reactor (external diameter of ½ inch instead of 3/8 inch), compared to other research groups [5–7]. Considering the WHSV, at 400 °C, 1 MPa, the 80 wt%Fe-20 wt% Ce0.5Sm0.5O2-xNy composite catalyst exhibits an activity (5.6 mmol g-1 h-1 at 16000 mL g-1 h-1) comparable to the best industrial benchmark Wüstite fused Fe catalyst (13.9 mmol g-1 h-1 at 36000 mL g-1 h-1, 400 °C, 0.9 MPa). However, purity of feed gas in this study is only 99.996 %, much lower than the 99.9999 % and 99.99995 % used in previous reports (Table S2).In the Haber-Bosch process, conversion and ammonia yield are limited by thermodynamic equilibrium at high temperatures as the reaction is exothermic [54]. Therefore, synthesis of ammonia at reduced temperature will have higher conversion and reduced energy consumption. At 1 MPa and 350 °C, the activity of 80 % Fe – 20 % Ce0.5Sm0.5O2-xNy is 2.86 mmol g−1 h−1 at WHSV of 16000 ml g−1 h−1, comparable to Fe-LiH (11 mmol g−1 h−1 at WHSV of 60000 ml g−1 h−1 [14], Ni-LaN (5.2 mmol g−1 h−1 at WHSV of 36000 ml g-1 h−1, 0.9 MPa) [29] (Table S2). At 350 °C, the activity of our Fe-Ce0.5Sm0.5O2-xNy is among the highest for all reported non-Ru catalysts for the Haber-Bosch reaction despite lower feed gas purity (99.996 %) (Table S2). The loading of cheap Fe (80 wt%) in our composite catalysts is much higher than the 0.4 wt% and 1.2 wt% Fe in the BaTiO2.4H0.6 and BaCeO3-xNyHz supported catalysts making direct comparisons less meaningful. To some extent, they are different catalyst types as our oxynitride promoted Fe catalyst is closer to the industrial fused Fe catalysts which normally contain over 90 wt% Fe. It is estimated that the cost of our 80 wt%Fe-20 wt% Ce0.5Sm0.5O2-xNy catalyst would be much lower than the LiH, BaTiO2.4H0.6, BaCeO3-xNyHz or LaN supported catalysts as cheap iron makes up the majority of the composition in our catalysts. Those with only a few weight percentage of transition metal such as Ru, Fe, Co, Ni are typically supported catalysts. The Fe-oxynitrides with 80 wt% Fe should be classified as composite catalysts.The apparent activation energy for the 80 wt%Fe – 20 wt% Ce1-zSmzO2-xNy catalysts with z = 0.1 to 0.5 at 1 MPa and 3 MPa are also obtained through the Arrhenius plots using the activity data at a temperature below 450 °C (Fig. 6c & d). With increased z in the Sm-doped cerium oxynitrides, the Ea tends to decrease. This tendency is quite clear when the Ea is plotted against z in Ce1-zSmzO2-xNy (Fig. 6e). For z = 0.1, at 1 MPa, the Ea is 76 kJ/mol, which is comparable to the 70 kJ/mol at 1 MPa representative industrial KM1 catalysts [5,52]. However, when the pressure is increased to 3 MPa, Ea is also increased from 76 to 98 kJ/mol. This is common as Ea for industrial fused Fe catalyst increased to 180 kJ/mol at 10 MPa [5,55]. When z ≥ 0.2, either at 1 MPa or 3 MPa, the Ea for all the Fe-Ce1-zSmzO2-xNy composite catalysts is lower than the Ea of 70 kJ/mol for the industrial Fe catalyst. When z ≥ 0.3, the Ea is in the range of 45 kJ/mol and is less sensitive to pressure change from 1 to 3 MPa. The Ea for sample Fe- Ce1-zSmzO2-xNy is 43 ± 3.22 kJ mol−1 at 1 MPa, 44 ± 6.27 kJ mol−1 at 3 MPa, around 45 kJ mol−1 (Fig. 6e). This is comparable with the lowest Ea in reported papers for promoters/co-catalysts such as Ru/C12A7:e (49 kJ/mol at 0.1 MPa) [5,53], Fe/LiH (46.5 kJ/mol at 1 MPa) [5,14], Fe/BaCeO3-xNyHz (46 kJ/mol at 0.9 MPa) [7], Fe/BaTiO2.4H0.6 (63 kJ/mol at 4 MPa) [5] and Ni-LaN (57.5 kJ/mol at 0.9 MPa) [27]. The low apparent activation energy of our Fe-Ce1-zSmzO2-xNy composite catalyst with z ≥ 0.3 is clearly related to the high concentration of anion vacancies. The doping of large Sm3+ ions and introduction of nitrogen vacancies through partial replacement of lattice oxygen by the larger nitrogen ions, leads to a lattice expansion (Fig. 3b). Please note the lattice parameters for Ce1-zSmzO2-xNy with z ≥ 0.3 (a ≥ 5.4281(1) Å) are much larger than that for sample CeO2-xNy (a = 5.4273(1)Å) (Table 2, Fig. 3b). Larger lattice parameters means the void (free volume) for mobile anions vacancies is also bigger, making the possible adsorption of large N2 molecular easier if the nitrogen vacancies take part in the ammonia synthesis reaction, as is the case for perovskite oxynitride hydride BaCeO3-xNyHz and LaN [7,27]. A large lattice parameter will lead to high mobility of anions, which also benefits the reaction. This will be discussed later.In order to figure out the relationship between lattice volume and the ammonia synthesis activity, the activity of the 80 wt%Fe-20 wt% Ce1-zSmzO2-xNy composite catalysts at different temperatures are plotted against the z values in the Ce1-zSmzO2-xNy (Fig. 7 ). For both 1 MPa and 3 MPa, at different temperatures, the lowest ammonia synthesis rate was observed for the Fe - Ce0.9Sm0.1O2-xNy catalyst. This is consistent with the relatively small lattice parameters (cell volume) for Ce1-zSmzO2-xNy samples (Fig. 3b). Comparing the nitrogen content and anion vacancy concentration for samples CeO2-xNy and Ce0.9Sm0.1O2-xNy, they have the same level of nitrogen content (y = 0.07) (Table 1), while the anion concentration in sample Ce0.9Sm0.1O2-xNy is higher than that for CeO2-xNy due to the doping of low-valent element Sm. However, the activity of Fe-Ce0.9Sm0.1O2-xNy is lower, which seems correlated to the smaller lattice parameters (cell volume). From this point of view, the catalytic activity is correlated to both the concentration of anion vacancies and the size/volume of the crystal lattice. In solid state ionics, large lattice parameters will lead bigger cell volume to more ‘free volume’ (void not occupied by ions), making the migration of anions much easier, leading high ionic conductivity. It has been reported that partially replacing Sm3+ ions in Ce0.8Sm0.2O2-δ with larger Ca2+ ions, leads to increased O2− ionic conductivity because of the increased ‘free volume’, making the migration of O2− much easier [56]. It is expected that the same situation may happen on Sm3+ and N3- co-doped CeO2, which shares the same crystal structure as Ce0.8Sm0.2O2-δ (Table 1). This indicates that high mobility of the anions, particularly N3- ions in the oxynitrides, which may participate in the reaction for ammonia synthesis, is another important parameter to achieve high catalytic activity [7]. Theoretically high anion conductivity, particularly N3- ion conductivity, can extend the reaction zone of ammonia synthesis reaction as N3- ions can be formed at anywhere on the surface, then quickly diffuse through the oxynitride particles to an active site. These active sites are the contact points between α-Fe and Ce1-zSmzO2-xNy, and is where the reaction is completed (Fig. 10), leading to higher catalytic activity. Introduction of O2− ionic conduction in electrodes to facilitate the electrode reaction has been widely used in SOFCs. This key strategy may also be employed to the catalytic reaction for ammonia synthesis to improve the activity of composite catalysts if anions such as N3- ions also take part in the reaction.The anion vacancy promotion effect due to the doping of nitrogen in CeO2, CeO2 mixed with the Fe-catalyst at the same mass ratio (20 : 80) was also investigated, as shown in Fig. 8 a&b. In the measured temperature range, Fe-CeO2-xNy shows much higher activity than Fe-CeO2, indicating the vacancy-rich CeO2-xNy is a better promoter. At 3 MPa and 450 °C, the ammonia formation rate increases from 9.7 mmol g−1 h−1 to 17.2 mmol g−1 h−1 when Fe–CeO2 is replaced with Fe–CeO2-xNy, almost doubling the activity. This provides further evidence that the anion vacancies, particularly nitrogen vacancies, may take part in the reaction, leading to increased activity [7,22,26,27]. The slightly increased activity of the Fe-CeO2 catalyst at 600 °C and 3 MPa (Fig. 3b) could be related to the formation of intrinsic oxygen vacancies at high temperature with the presence of high pressure H2 (Eq. (2)). The maximum rate allowed at our reactor conditions according to thermodynamic equilibrium is shown in Fig. 8a & b [54]. It can be seen that for both 1 MPa and 3 MPa, the reactions approach the equilibrium conversion as temperature increases. For both pressures, the catalyst reaches their peak activity at values lower than the thermodynamic equilibrium rate.The apparent activation energies of the Fe catalysts promoted by all three promoters are lower than 70 kJ/mol for the representative industrial Fe catalyst (Fig. 8c & d) [5,52]. The activation energies for Fe-CeO2 and Fe-CeO2-xNy are around 60 kJ mol−1, slightly higher than that for Fe-Ce0.5Sm0.5O2-xNy (Fig. 6e). The presence of a large amount of anion vacancies in Ce0.5Sm0.5O2-xNy benefits the ammonia synthesis reaction, reducing the apparent activation energy (Table 1).The ammonia synthesis rate is a key parameter when people talk about the activity of a catalyst. The ammonia synthesis rate is related to the activity of the catalyst, the loading of the catalyst and gas flow rate (space velocity). The effect of space velocity on the ammonia synthesis rate and conversion of 80 wt%Fe-20 wt% Ce0.5Sm0.5O2-xNy at 400 °C, 3 MPa is shown in Figure S7. The rate of formed ammonia increased as flow rate of the feed gases and WHSV was increased. However, total conversion increases at reduced feed gas flow rates due to the longer residence time of the reactants in the catalyst bed. A nearly linear relationship between total flow rate and outlet conversion was observed indicating good mass transfer properties between that catalyst and reactants. This experiment indicates that the ammonia synthesis rate is linear to the space velocity in the region tested, the ammonia synthesis rate will be doubled if the space velocity is doubled and vice versa. Therefore, WHSV should be taken into account when comparing the ammonia synthesis rate from different sources [22]. In Table S2, the ammonia synthesis rates from different Fe-based catalysts plus the representative Cs-Ru/MgO are listed together alongside their respective WHSV.Stability is an important parameter for industrial ammonia synthesis catalysts. Among all the investigated promoters (CeO2, CeO2-xNy and Ce1-zSmzO2-xNy), Ce0.5Sm0.5O2-xNy exhibits the highest promotion effect (or co-catalyst) to the Fe catalyst for ammonia synthesis. Industrial ammonia reactors usually run continuously for long periods of up to years at a time. Long term stability tests of the new catalysts are carried out as it is vital for commercial applications. The stability of the 80 wt% Fe – 20 wt% CeO2-xNy catalyst was measured for nearly 200 h in Zero grade feed gas, at the optimised conditions, 450 °C and 3 MPa (Fig. 9a). There is a slight drop in activity over the first 50 h before the catalyst stabilises then remains stable over the rest of the entire test. This slight initial drop in activity is expected to be caused by the change in nitrogen content in the promoter material which needs to be stabilised in mixed N2 and H2 at high temperature and high pressure. At high temperature, it is possible that the Ce0.5Sm0.5O2-xNy is partially oxidised by the oxygenate impurities as nitrogen was not detected by XPS for the 80 wt% Fe – 20 wt% Ce0.5Sm0.5O2-xNy sample after the catalytic activity at both 1 and 3 MPa to a temperature up to 600 °C (Fig. 2) although the nitrogen content in Fe - Ce0.5Sm0.5O2-xNy could be too low, beyond the measuring limit for XPS and CHN.The stability of the 80 % Fe – 20 % Ce0.5Sm0.5O2-xNy catalyst at 400 °C and 3 MPa was also measured for over 200 h (Fig. 9b). Similar to the CeO2-xNy promoted Fe-catalyst, an initial slight drop in activity is also observed, remaining stable for the rest of the tested hours. Although the initial drop is at a similar extent to the CeO2-xNy promoted Fe-catalyst, the drop in activity for 80 % Fe – 20 % Ce0.5Sm0.5O2-xNy catalyst continues over a longer period of time, about 100 h. Since the operating temperature is 50 °C (400 °C instead of 450 °C) lower, it could therefore take a longer time to stabilise the nitrogen content in the Ce0.5Sm0.5O2-xNy promoter.In conventional ammonia synthesis catalysts, low operating temperatures amplify the poisoning effect of impurities causing a more significant problem than at high temperatures [17,18]. The stability of 80 % Fe – 20 % Ce0.5Sm0.5O2-xNy catalyst at 300 °C, 3 MPa was therefore also investigated. This catalyst is stable at 300 °C during the measured 290 h in which there is a ∼ 90 h break (Fig. 9c). During the break, the gas flow was stopped, the reactor was cooled down to room temperature with 3 MPa mixed N2 and H2 together with the generated NH3. This experiment indicates that, the oxynitride promoted Fe catalyst exhibits excellent stability in Zero grade feed gas, even at room temperature. This is very different from the conventional industrial fused Fe catalyst as the promoters we used are vastly different. In fused Fe catalysts, the stability relied on the homogeneously distributed Al2O3 additive [17,18]. In our catalyst, there is no added Al2O3 at all. The stability of our Fe-oxynitride composite catalysts relied on the SMSI between iron and oxynitride, prohibiting the growth or sintering of the iron particles. Therefore, our Fe-oxynitride catalyst is even more stable at temperatures as low as room temperature [22].Intermittent operation of the ammonia synthesis reactor is therefore possible as the catalyst will not be damaged during the heating/cooling process. This is particular useful for green ammonia production at a small scale using the surplus intermittent renewable electricity as the energy sources [2]. The dominant technology for ammonia production will be the Haber-Bosch reaction for the foreseeable future, including green ammonia production [2,18,57]. It has been reported that improving the efficiency of water electrolyser and/or developing new catalysts enabling the agile operation of the Haber-Bosch process are the keys to achieving green ammonia industry [58]. From this point of view, if our cation doped cerium oxynitride promoted catalysts are used in the localised green ammonia synthesis plants, then there is potential for this technology to be used for renewable electricity storage providing a better match with the intermittent nature of the renewable resources. During the stability test at 300 °C, no initial activity drop was observed indicating the nitrogen content is fairly stable at this temperature, which implies Ce0.5Sm0.5O2-xNy is not (partially) oxidised by the oxygenate impurity at 300 °C.It has been reported that high pressures can help to prevent the decomposition of oxynitrides to oxides and nitrogen, and high pressures facilitate the oxynitride synthesis process [59]. Therefore at 3 MPa both Ce0.5Sm0.5O2-xNy and CeO2-xNy promoters are expected to have a higher tolerance to decomposition caused by oxygenate impurities in the feeding gas. We tried to test the nitrogen content in the 80 wt%Fe‐20 wt%Ce0.5Sm0.5O2-xNy catalyst after the stability test, however, due to the large Fe content of 80 wt%, the nitrogen content is too low, beyond the measurement limit of the XPS and CHN facility as the total nitrogen in the total composite catalyst is too small. The stability of oxynitrides is related to both the oxygenate concentration and reaction temperature. When the reaction temperature is reduced to 300 °C, it was found that the catalytic activity of Fe-Ce0.5Sm0.5O2-xNy catalyst is stable, indicating that the Ce0.5Sm0.5O2-xNy is stable in the less pure feed gas at 300 °C. It has been reported that the Ni-LaN catalyst is stable for ammonia synthesis when ultrapure feed gas (purity > 99.99995 %) was applied [27]. Ce0.5Sm0.5O2-xNy could also be chemically stable at higher temperatures when ultrapure feed gas is used for ammonia synthesis.It has been confirmed that the 80 % Fe – 20 % Ce0.5Sm0.5O2-xNy catalyst exhibits the highest activity without further purification of the Zero grade H2 and N2 feed gases. In the past, catalyst tolerance towards impurity is normally tested at fixed concentration of O2 such as 5 ppm of an oxygenic compound [20], 1 ppm impurity [19,21]. In order to test the limit of our Fluorite oxynitride promoted Fe catalyst, a much higher concentration of impurities was used. The activities of our 80 wt% Fe – 20 wt% Ce0.5Sm0.5O2-xNy catalyst at 475 °C and 3 MPa with different impurity levels up to 200 ppm were investigated (Fig. 9d). Here the temperature is the measured temperature of the tube furnace hot zone. This investigation into oxygenate tolerance to 200 ppm impurity was conducted in a newly designed reactor capable of achieving higher pressures with a large wall thickness (0.125 inch instead of 0.083 inch). Due to the less effective heat transfer between the thick wall reactor and hot zone of the furnace, the real temperature of the catalyst in new thick wall reactor is slightly lower [22]. An impurity of 10 ppm in the gas is expected to remain as the O2 and H2O traps cannot remove impurities such as CO, CO2, and hydrocarbons [22]. To achieve higher impurity levels, zero grade nitrogen was mixed with nitrogen with 1000 ppm O2 to reach higher quantifiable oxygenate concentrations desired. The activity of the 80 wt% Fe ‐ 20 wt% Ce0.5Sm0.5O2-xNy composite catalyst in Zero grade gas is about 86 % of that after the purification process indicating that the oxygenate impurities still exhibit an effect on activity. Higher activity can be obtained if very pure feed gas is applied in our Fe-oxynitride catalysts. Further increases in the total impurity level to 104 ppm, with known injected 61 ppm oxygen, provides a rate retention of 81 %. Please note 61 ppm oxygen equals to 122 ppm atomic oxygen. This is over 10 times of the maximum allowed oxygenate level (10 atomic oxygen) for industrial fused Fe catalysts [17]. The ammonia formation rate is still more than half (53 %) of the original rate in purified gas when total impurity level was 200 ppm with known injected 158 ppm O2. This experiment indicates the 80 wt% Fe – 20 wt% Ce0.5Sm0.5O2-xNy composite catalyst exhibits excellent oxygenate tolerance properties. When a feed gas with 200 ppm impurities is used, by doubling the amount of catalyst / the size of the reactor, the same amount of ammonia can be produced compared to standard feed gas with 10 ppm atomic oxygenate. If an oxygenate tolerant catalyst, such as Fe – Ce1-zSmzO2-xNy composite is used in the reactor, purification requirements will be lower thus saving on initial facility cost and continued energy inputs significantly improving the overall efficiency for ammonia synthesis. However, there is a risk that the oxynitride may be partially oxidised by oxygenate impurities if their concentration is too high at high reaction temperature. The formed Sm-doped CeO2 will still exhibit high activity and stability but the activity will be slightly lower due to the loss of nitrogen vacancies (Fig. 9b) [22]. Therefore, if we want to take advantage of nitrogen vacancies in Ce1-zSmzO2-xNy to achieve the high activity it is required to minimise the oxygenate concentration. Developing more stable promoters with anion vacancies present in a high concentration, particularly nitrogen vacancies is therefore desired in order to achieve a high stability in oxygenates.The results presented above clearly show that the introduction of a large amount of anion vacancies in CeO2 through cation (Sm3+ ions) and anion (N3− ions) co-doping result in various changes in catalytic properties when used as support for low-cost Fe catalysts for ammonia synthesis. Both stability and activity of the Fe-Ce1-zSmzO2-xNy composite have been significantly improved, which is attributed to the anion vacancies, particularly nitrogen vacancies. The schematic diagram on the interaction between anion vacancies and α-Fe particles and the reactants, H2 and N2 for ammonia synthesis is shown in Fig. 10.As for the reaction mechanism of oxynitride or oxynitride hydride promoted transition metal (TM) catalysts, Kobayashi et al. reported that oxyhydride BaTiO3-xHx improves the activity of the transition metal catalysts through the oxynitride-hydride intermediate, where both lattice N3− ions and H- ions play important roles for the increased catalytic activity [5]. Kitano et al. proposed two possible reaction mechanisms for ammonia synthesis over TM / BaCeO3-xNyHz catalysts. Both are related to Mars − van Krevelen mechanism through anion vacancies with the participation of lattice N3− and H- ions [7]. The key evidence for the Mars − van Krevelen mechanism is the low apparent activation energy, 46−62 kJ/mol for TM/ BaCeO3-xNyHz catalysts [7]. The introduction of N3− ions to the BaTiO3-xHx lattice through an in situ formed oxynitride-hydride intermediate or, to the perovskite oxynitride hydride BaCeO3-xNyHz lattice at the very beginning, will generate anion vacancies from the charge balance. The more N3− ions are introduced into the lattice, the more anion vacancies will be generated. In our Ce1-zSmzO2-xNy, there are N3− ions in the lattice already, similar to BaCeO3-xNyHz. As for the H- ions, it has been widely reported that Ce-H species may be formed when CeO2 is exposed in H2 at high temperature while more Ce-H species can be formed when more oxygen vacancies are presented in the CeO2 lattice [60,61]. It is reasonably deduced that, under the ammonia synthesis conditions, in the presence of high concentration (∼ 75 %) H2 at high temperature, some hydride intermediates may also be formed in our cerium oxynitrides. Following this our oxynitride would be a kind of oxynitride hydride, similar to BaCeO3-xNyHz, although exhibiting a fluorite structure instead of a perovskite structure. The apparent activation energy of the 80 wt% Fe- 20 wt% Ce1-zSmzO2xNy catalysts with z ≥ 0.3 is around 45 kJ/mol, which is fairly close to that for the TM/ BaCeO3-xNyHz catalysts [7]. The low activation energy plus the presence of N3- ions and possible indirectly formed H- ions through the intermediate in Ce1-zSmzO2-xNy, are very similar to the case for BaCeO3-xNyHz, thus they may share the same or similar reaction mechanism.Another important finding in this study is, the catalytic activity seems related to the lattice parameter of the Ce1-zSmzO2-xNy support. The larger the lattice volume with larger ‘free volume’, the higher the catalytic activity (Figs. 3b & 7). This can be considered in two aspects: (a) The size of anion vacancies. As shown in Fig. 10, both N2 and H2 may be adsorbed on the anion vacancies on the surface. It is believed that the bigger the vacancy (charged void), the easier for the adsorption of N2/H2 gases, thus the higher the catalytic activity. (b) The mobility of N3− and H- ions. When N2 and H2 are adsorbed on anion vacancies on the surface of Ce1-zSmzO2-xNy, it may dissociate to form N3- and H- ions, according to the proposed reaction mechanism of oxynitride hydride [7]. High mobility or ionic conductivity of N3- and H- ions will allow the reaction between N3- ions and adsorbed H- species on Fe or, between H- ions and N-species on Fe to happen all over the surface of the composite catalyst, rather than limited to the triple phase (Fe- Ce1-zSmzO2-xNy-gaseous reactants) boundary, similar to the case for the reaction on the anode of a solid oxide fuel cell. This will increase the probability for the formation of ammonia thus result in a higher activity. Large lattice volume means large ‘free volume’ thus the ionic conductivity or mobility of the N3-/ H- ions will be higher, leading to higher catalytic activity [62]. It should be noted that Sm-doped CeO2 is a well-known O2- ionic conductor with high ionic conductivity and has been used as an electrolyte for SOFCs. It is expected that the ionic conductivity for other anions such as N3- and H- ions in the Ce1-zSmzO2-xNy will also increase at increased lattice parameters thus larger ‘free volume’ for easy diffusion of ions, facilitating the ammonia synthesis reaction [56].The high concentration of anion vacancies in Ce1-zSmzO2-xNy will facilitate the nesting/anchoring of Fe particles, resulting in SMSI, which has been described above (Fig. 10). This SMSI can prohibit the sintering of Fe particles even under a strong oxidization environment [22,29]. Therefore, a large amount of anion vacancies in cation doped cerium oxynitride improved both the stability and catalytic activity for ammonia synthesis.It has been reported that, for cubic Fe, (111) plane is the most active for ammonia synthesis reaction. The second most active plane is (211) when exposed to the reactant gases [63]. If a Fe atom is nested onto a surface anion vacancy via SMSI, where the array of anions (O2− and N3- ions) are, ideally the (111) plane is in parallel or close to parallel to the plane of anion arrays. Therefore, the probability for (111) planes to be exposed to the reactants (H2 and N2) is very high, thus can maximize the ammonia production [22]. Theoretically N2 located away from the α-Fe may also be dissociated by nitrogen vacancies not in contact with Fe, then diffuse through the whole Ce1-zSmzO2-xNy particle, then reach the contact point between α-Fe and Ce1-zSmzO2-xNyto catalyze the ammonia synthesis reaction. From all aspects of the ammonia synthesis reaction, oxynitrides with a large amount of anion vacancies, particularly nitrogen vacancies, will benefit the reaction to improve both activity and stability.Cation doped fluorite oxynitrides, new single phase Sm-doped cerium oxynitrides have been synthesized for the first time. For the high Sm-doped cerium oxynitrides, approximately 16.5 % of the anion positions are not occupied. The introduction of nitrogen to form nitrogen vacancies will have better match to the adsorbed N2 in terms of size. The optimised composition is 80 wt% Fe – 20 wt% Ce0.5Sm0.5O2-xNy which showed an activity of 18.8 mmol g−1 h−1 at 400 °C, 3 MPa (WHSV = 16000 mL g−1 h−1) using 99.996 % H2 and N2 as the feed gases, comparable to the industrial fused Fe catalyst at a much higher purity. At 350 °C and 1 MPa, the activity is among the highest in all reported non-Ru based catalysts. The apparent activity energy of our Fe-Ce1-zSmzO2-xNy catalysts with z ≥ 0.3 is in the range of 45 kJ/mol, among the lowest Ea for all reported ammonia synthesis catalysts [5,7,27]. It is believed that the reaction proceeds through the Mars − van Krevelen mechanism mediated by the anion vacancies, similar to perovskite oxynitride hydride BaCeO3-xNyHz. At 3 MPa and 475 °C, the activity retention of the Fe – 20 wt% Ce0.5Sm0.5O2-xNy catalyst is 70 % with known injected 107.5 ppm O2 (150 ppm impurity level). The Fe-Ce0.5Sm0.5O2-xNy catalyst exhibits excellent stability at 300 °C, even after cooling to room temperature implying stability at room temperature. This is suitable for agile operation of localised green ammonia synthesis plants using intermittent energy from renewable electricity. Both catalysts starting from α-Fe2O3- oxynitride are stable in air at room temperature and are thus easy to handle. This article provides a new development strategy for the synthesis of novel oxygenate-tolerant ammonia synthesis catalysts that can be used for both existing large scale Haber-Bosch processes as well as small scale green ammonia synthesis from renewable energy sources. Agile operation is key for small scale ammonia plants utilising intermittent renewable energy, therefore, future work thoroughly investigating the catalyst tolerance to the thermal shock of reactor start-up/shut-down should be investigated based on the promising results highlighted in the stability test. The development of a large weight percent iron based catalyst will have vast cost advantages over expensive Ru and Co based catalysts. However, the replacement of relatively expensive rare earth elements in the oxynitride provides room for further improvements in this regard, which is under investigation.None.The authors thank EPSRC (Grant No. EP/G01244X/1) for funding.Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.apcatb.2020.119843.The following is Supplementary data to this article:

For the first time, new Sm doped cerium oxynitrides with the formula Ce1-zSmzO2-xNy (z ≤ 0.5) are synthesized in order to maximize the concentration of anion vacancies. Single phase Sm-doped CeO2-xNy were confirmed by XRD, HRTEM and Rietveld refinement. These oxynitrides show a great promotion effect for the low-cost Fe catalyst for the ammonia synthesis. At 350 °C and 1 MPa, the activity of 80 wt% Fe- 20 wt% Ce1-zSmzO2xNy is one of the highest reported for non-Ru catalysts for the Haber-Bosch reaction. The apparent activation energy of the 80 wt% Fe- 20 wt% Ce1-zSmzO2xNy catalysts with z ≥ 0.3 is around 45 kJ/mol, which is in the lowest range among all reported ammonia synthesis catalysts. Introduction of nitrogen vacancies through doping may facilitate the mobility of nitrogen vacancies. This study demonstrates doped oxynitrides with a large concentration of anion vacancies, particularly nitrogen vacancies are excellent promoters/co-catalysts for ammonia synthesis.

The production of fuels, such as bio-kerosene for aviation transportation, directly from lignocellulosic biomass is of increasing interest for energy sustainability [1]. Unfortunately, bio-oil produced via thermal pyrolysis of biomass does not meet the requirements desired for a transportation fuel as it is very viscous, dense, corrosive and poorly energetic [2]. Therefore, the bio-oil formed is required to undergo a set of upgrading steps, in which the energetic value is increased by oxygen depletion and by chain elongation. Catalytic fast pyrolysis (CFP) takes place at high temperatures (∼500 °C), in an inert atmosphere (i.e., in the absence of oxygen) and with short residence times (i.e., few seconds) to lower the bio-oil oxygen content [3]. When CFP is performed in the so-called ex-situ mode [4], the catalyst is only contacted with the pyrolysis vapors after conducting the thermal pyrolysis step in a separate reactor. This way of operation favors the formation of aliphatics and olefins over aromatics, which are more interesting for bio-oil production. The bio-oil generated by ex-situ CFP can be further upgraded to advanced bio-fuels by subsequent deoxygenation reactions.Traditionally, zeolite ZSM-5 has been mostly used for the CFP step. This can be done solely with zeolite ZSM-5 as the active phase [5,6] or in combination with promoters, such as Ni or Ga [7,8], which boost its efficiency in obtaining a high-quality deoxygenated bio-oil. The good activity of zeolite ZSM-5 is related to its shape selectivity, acidity and thermal stability [5,6]. However, a common limitation of zeolite ZSM-5 in the CFP process is its deactivation due to coke formation, which results in clogging its micropores. Creating mesoporosity within zeolite ZSM-5, e.g. via desilication [9], can considerably lower the detrimental effects of pore blockage by coke formation [10]. It is important to note that the formation of coke deposits is induced by olefin polymerization [11], as well as other C-C coupling reactions, all of which can abundantly occur during biomass conversion processes. C-C coupling reactions are also required, however, as part of the necessary bio-oil upgrading, with ketonization [12,13] or aldol condensation [14,15] being typical examples, given the large amounts of carboxylic acids, ketones and aldehydes present in bio-oil [16]. Indeed, these reactions are often the topic of (model) bio-oil upgrading and deoxygenation studies and are commonly catalyzed by base catalysts, such as metal oxides (e.g., ZrO2, TiO2 or hydrotalcites) or alkali metal-exchanged zeolites [17,18].For bench-scale and pilot-scale testing, the use of shaped catalyst bodies is required to ensure mechanical strength and to avoid pressure drop issues in the chemical reactors used [19]. To make suitably shaped catalyst bodies, binder materials, such as clay minerals and alumina, are typically added to the powdered catalysts to obtain a mixture that upon extrusion or granulation generates the shaped catalyst bodies. However, binders can modify the catalyst properties and hence activity, e.g. by offering a source of metal cations that can exchange with and alter the catalyst’s active sites, altering for example catalyst acidity within a zeolite [20,21]. Yet, binder materials can also affect the catalyst’s stability. Deactivation of catalyst bodies has been studied, for instance, for fluid catalytic cracking (FCC) particles. These FCC particles are formed via e.g., a spray drying procedure of the different catalyst components and consist of a zeolite material, e.g. zeolite H-Y, promoted or not with active metals (e.g., La), binder and filler materials, such as SiO2, Al2O3 and clay minerals [22,23]. The binder and filler materials improve the stability of the active zeolite component, which would suffer from more structural damage under the hydrothermal conditions applied in the industrial reactor and regenerator systems if the binder and filler would not be present [24]. Common reasons for FCC catalyst deactivation are poisoning by coke [25,26] and metals deposition which leads to catalyst particle agglutination, impeded accessibility of the zeolite’s pore network, and alterations in size and acidity, as illustrated by Meirer et al. [27–29].The nature of the binder material in a catalyst body has a significant impact on its performance. On extruded zeolite-based catalysts, Verkleij et al. reported on the higher resistance against deactivation for zeolite ZSM-5 when blended with Al2O3 rather than SiO2, in the 1-hexene oligomerization reaction. Indeed, the Al2O3-based system better favored elongated oligomers over branched ones, protecting the zeolite pores better against clogging [30]. The same catalyst systems, when applied in the transalkylation of aromatics [31,32], also showed differences in stability. In this case, the Al2O3-based catalyst extrudates showed more and more condensed coke deposits than the SiO2-based catalyst. These deposits were located mainly on the rim of the zeolite crystals, however, while in the case of the SiO2-based catalyst extrudates coke deposits could be found throughout the zeolite crystals, leading to faster deactivation.Indeed, the choice of the binder and its integration with the other components of the technical catalyst – i.e., the manner in which e.g. porosity, crystal size or acidity of the zeolite component is affected - will impact the mode, rate and extent of catalyst deactivation, as previously reported [33]. To reverse (totally or partially) certain causes of catalyst deactivation, oxidative regeneration cycles at high reaction temperatures can be carried out [34,35]. Indeed, by burning-off the coke deposits from spent catalysts, pore or active site blockage can in principle be reversed and the physicochemical properties of the catalyst material, such as surface area, pore volume and acidity, restored. As an example, Michels et al. demonstrated that upon regeneration in flowing air at 550 °C for 3 h, a zeolite ZSM-5-based catalyst extruded with attapulgite as binder material showed a significant recovery of the textural properties and catalytic activity in the methanol-to-hydrocarbons (MTH) reaction [36]. However, other causes of catalyst deactivation, such as morphological and textural changes, could not be repaired by such a regeneration process. Indeed, because of the high temperatures and steam generated upon coke combustion, changes in the structural and textural properties of the catalyst usually worsen [25]. In line with this observation, previous reports [37] emphasized the importance of applying intermediate regeneration temperatures (yet equal or higher than the reaction temperature), to ensure total combustion of the coke deposits formed within the zeolite catalyst but avoid irreversible structural damages.In this work, we report on the different modes of deactivation of two tailored catalysts used in cascade for a more efficient bio-oil deoxygenation in lignocellulose catalytic pyrolysis. This cascade reaction [38] combines the synergistic effect of a solid acid ZrO2/desilicated zeolite ZSM-5/attapulgite clay mineral (further denoted as ZrO2/ds-ZSM-5-ATP) -employed for the ex-situ CFP step- and a solid base K-grafted zeolite USY/attapulgite (further denoted as K-(USY-ATP)) -used in the subsequent bio-oil upgrading step-. While the acidity of the ZSM-5-based catalyst is key to promote cracking and alkylation reactions involved within the catalytic pyrolysis stage, the basicity of the alkaline-grafted USY catalyst is so for the subsequent upgrading of bio-oil via deoxygenating routes, such as aldol condensations [15].Understanding the origin and cause of the deactivation of these catalysts, as well as the extent to which deactivation can be reversed by a suitable regeneration treatment will allow optimizing the catalyst lifetime. To this extent, the catalyst materials were extensively characterized using bulk and spatially-resolved characterization techniques on fresh, spent and regenerated samples in their shaped form. The extent, nature and location of coke formation, the structural integrity and acid/base properties of the catalyst materials, and the regeneration effects on coke removal and recovery of the original properties have been assessed.The experimental details related to catalysts syntheses and characterization are presented in the Supporting Information.The bench scale set-up in which ZrO2/ desilicated ZSM-5-attapulgite catalyst (ZrO2/ds-ZSM-5-ATP) as solid acid and K-exchanged zeolite USY/attapulgite catalyst (K-(USY-ATP)) as solid base were tested is schematically depicted in Fig. 1 c [39]. Biomass feedstock consisted of previously de-ashed wheat straw (4 g) heated to 550 °C and fluidized in 100 NmL/min N2. In a first stage, a catalytic bed of ZrO2/ds-ZSM-5-ATP is exposed to the pyrolytic vapors coming from the thermal pyrolysis stage. The subsequent stage consists of a fixed bed of K-(USY-ATP) catalyst for the for the treatment of the catalytic vapors, which are further upgraded. Both catalytic pyrolysis and upgrading processes operate at 450 °C, and on both processes the catalyst to biomass ratio (C/B) was of 0.6 (excluding the binder into the catalyst weight). The vapors leaving the reactor were condensed to collect the liquid bio-oil for approx. 10 min; non-condensable gases were collected in sampling bags at the end of the line. The energy yield associated with bio-oil product was calculated as the proportion of chemical energy (HHV) retained regarding that of raw biomass.Regeneration of both catalysts was carried out under static air by heating the catalyst bodies in an open crucible at a temperature ramp of 1.8 °C/min up to 550 °C and holding this temperature for 6 h. Fig. 1 shows the activity of the cascade process, consisting of an ex-situ catalytic fast pyrolysis step (i.e., thermal and catalytic pyrolysis) with a ZrO2/ desilicated ZSM-5-attapulgite catalyst (ZrO2/ds-ZSM-5-ATP) as solid acid and a catalytic upgrading step with a K-exchanged zeolite USY/attapulgite catalyst (K-(USY-ATP)) as solid base. The activity is expressed as the catalytic bio-oil deoxygenation as a function of the mass yield (a) and the energy yield (b) for the cascade process (see reactor scheme in Fig. 1c) compared to the single-step CFP process run with ZrO2/ds-ZSM-5-ATP only.After applying this three-step cascade process for a catalyst/biomass (C/B) ratio of 0.6, and for a bio-oil mass yield of 40 wt%, the deoxygenation degree (compared to the non-catalytic thermal bio-oil) was ∼60 wt% (see green triangle data in Fig. 1a). For the ex-situ CFP only experiment, the run with a C/B ratio adjusted at 1.2 to run at equal total amount of catalyst material used. The deoxygenation degree was ca. 15 wt%, clearly inferior to the cascade catalytic process (grey triangle data series). Also, it turned out that at the same weight conditions the energy yield (Fig. 1b) is higher when two catalytic processes are coupled (60%) with respect to only one catalytic step (57%). Accordingly, the cascade process leads to both enhanced bio-oil deoxygenation and energy yield.After ex-situ CFP, 6.8 wt% of coke was formed in the ZrO2/ds-ZSM-5-ATP catalyst, as determined by thermogravimetric analysis-mass spectrometry (TGA-MS) (Fig. S1). The main part of it (6.0 wt%) was highly deficient in hydrogen, indicated by the high temperature of combustion (∼480 °C, Fig. S1a-b) [40]. Given the likely insoluble character of the coke formed which precludes the analysis by chromatography, its polyaromatic nature was confirmed by FT-IR spectroscopy, showing the characteristic stretching bands of condensed ring aromatic structures (νC=C) in the 1560–1600 cm−1 spectral region and the C-H stretching vibrations (νCH) corresponding also to aromatic coke at 3067 cm−1 with contributions of methylene (2930 and 2860 cm−1) and methyl (2970 cm−1) groups [41] (Fig. S2a). Condensations, alkylations and isomerization reactions produced upon CFP may be responsible for the origin of these polyaromatic species. The FT-IR spectrum showed also signatures of aldehydes (νCH at 2745 cm−1, νC=O at 1714 and 1691 cm−1), olefins (νC=C 1636 cm−1) and organic acids (νC=O/νO-H 1616 cm−1), among other compounds, indicating the presence of oxygenates in the carbon deposits including phenols and furans. The corresponding UV–Vis diffuse reflectance spectrum (Fig. S2b) further corroborated the aromatic nature of coke with the presence of species such as pyrenes (with absorption bands at ∼250–350 nm) [42], naphthalenes and anthracenes (with absorption bands at ∼280–400 nm) and more conjugated poly-aromatic carbonaceous species and/or graphite-like coke, likely insoluble, characterized by absorption bands with maxima at >400 nm [43].The size of the coke deposits formed on the spent ZrO2/ds-ZSM-5-ATP catalyst was estimated by Raman spectroscopy, based on the expression introduced by Ferrari and Robertson for disordered graphitic carbons and graphene (see Eq. (1) in the Supporting Information) [44,45]. Based on the D and G integrals (Fig. S3a), the average coke size is estimated to be between 5 and 10 Å. Substituted pyrenes (Fig.S3b) are of this size and have been proposed by Guisnet and Magnoux as average component of the (soluble) coke formed within zeolite H-ZSM-5 material for a coke content close to 9 wt%, with an approximate boiling point close to 400 °C [46].The spent K-(USY-ATP) catalyst extrudate, employed for the catalytic deoxygenation of the bio-oils formed in the first CFP step, showed a lower amount of coke deposits than the ZrO2/ds-ZSM-5-ATP CFP catalyst. This is consistent with the process setup, with the zeolite ZSM-5-based catalyst being directly exposed to the raw pyrolytic vapors, while the K-loaded zeolite USY-based catalyst further upgrades the already treated vapors. TGA-MS analysis of the spent USY catalyst (Fig. S1c,d) showed a coke content of 5.6 wt%, being for the largest part again poly-aromatic [40] (4.9 wt%, combusting at ∼415 °C), while a smaller fraction is attributed to hydrogen-richer coke (Table 1 ). The nature of the coke deposits was further characterized by FT-IR spectroscopy (Fig. S2c) and UV–Vis DRS (Fig. S2d). FT-IR spectroscopy on the spent K-(USY-ATP) catalyst material (Fig. S2c) showed the presence of aromatics (νCH at 3067 cm−1) and coke (νC=C at 1574, 1600 cm−1). As expected, the lower content of aromatics and coke for the K-(USY-ATP) catalyst was indicated by the lower relative intensities of these bands (Fig. S2c) compared to the ZSM-5-based catalyst employed in the first catalytic stage (a). It should be noted that the FT-IR bands at 2745 cm−1 (νCH) and 1714 cm−1 (νC=O) are indicative of aldehydes, showing also oxygenates presence which originate from the deoxygenation activity carried out by the K-(USY-ATP) catalyst. The UV–Vis DRS spectrum of the spent K-(USY-ATP) catalyst material (Fig. S2d) indicates the presence of hydrogen-rich coke types, such as alkylated benzenes, absorbing light in the range of 250–270 nm; hydrogen-deficient coke types, such as naphthalenes and anthracenes, which absorb light in the range of 280–400 nm [43]; and poly-aromatic carbonaceous species above 400 nm.As catalyst extrudates might be subject to diffusion limitations, any coke deposits formed could be non-homogeneously distributed over the shaped catalyst body. To study any spatial distribution of the coke deposits, the spent technical catalysts were studied by Confocal Fluorescence Microscopy (CFM) [47], as schematically depicted in Fig. 2 .Visual inspection after cross-sectioning of the spent ZrO2/ds-ZSM-5-ATP catalyst showed a clear egg-shell distribution of the coke deposits formed. When the surface/shell (Fig. 2a) was irradiated with the excitation lasers of the CFM set-up, no fluorescence could be detected because of the high amount of poly-aromatic or (even) coke, which renders the surface opaque. When moving from the edge to the center of the catalyst body (Fig. 2b), fluorescence was seen, where there were less coke deposits and of softer nature -i.e. H-richer- than on the catalyst surface. Distinguished regions can be observed: brighter ones (highlighted in red, labelled as 2) with an extensive presence of naphthalenes and anthracenes (emitting light at wavelengths below 550 nm), and darker (green, 3) with higher presence of poly-aromatics (with more than 3 aromatic rings) which emit light above 550 nm [47,48]. The ratio between poly-aromatic and anthracene-like carbonaceous species was more pronounced in the regions located closer to the edge of the cross-section, (3, green), confirming the higher presence of more conjugated coke on more external locations of the catalyst body.Zooming in at a central position of the cross-section (1, blue, Fig. 2b and Fig. 2c) revealed various bright spots (A) assigned to the presence of zeolite crystals (0.5–2 µm) where coke preferentially forms in mesopore walls [49,50] given the high concentration of Brønsted acid sites for the zeolite ds-ZSM-5 [51]. By contrast, the darker spots (B) might correspond to the presence of attapulgite crystals (0.5 µm × 30 nm), which given their absence of Brønsted acid sites [52] would produce less coke deposits [53,54].The spent K-(USY-ATP) catalyst extrudate showed an egg-shell distribution of coke deposits too, visually confirmed by the gradient in color over the extrudate cross-section (Fig. S4). Indeed a larger proportion of poly-aromatics [47,48,55] was found on the external surface, while inner spots of the catalyst extrudate contained more coke rich in hydrogen (mainly naphthalenes/anthracenes emitting light between 400 and 500 nm). The shell, with an approximate thickness of ½ mm, was better distinguished when recording spectra on selected spots/regions due to the lower fluorescence intensity compared to the core of the section (Fig. 2d). The catalyst extrudate’s surface (Fig. 2a) also contained regions of higher fluorescence intensity presenting a similar concentration profile as seen in the core of the cross-section, which is possibly associated with local attapulgite agglomeration. Note that for this catalyst material the unique source of (very weak) Brønsted acid sites is the clay mineral given that the zeolite USY is highly dealuminated (i.e., a Si/Al ratio of ∼400) [56].The influence of coke deposits and their subsequent regeneration on the textural properties of spent and regenerated catalyst bodies were determined by physisorption of Ar gas at −196 °C. The results are shown in Fig. S5 and summarized in Table 1.The ZrO2/ds-ZSM-5-ATP catalyst, which showed a type I to type IIb isotherm with a steep H3 hysteresis loop [57,58] when fresh, (Fig. S5a, associated with a microporous material with additional mesoporosity) changed to a type I isotherm with a flatter H4 hysteresis loop after reaction. This was caused by the big loss in micropore and, in particular mesopore volume, partially blocked by the formation of coke deposits. The surface area was also significantly reduced after reaction (∼38% in drop, Table 1). The pore-size distributions plot (Fig. S5b) reveals that the micropores (filled at low relative pressures) and mesopores (filled at higher relative pressures) were partially shuttered after reaction. Notably, with the regeneration procedure applied the original textural properties were recovered to a large extent (up to ∼92%) with micro- and mesoporosity, as indicated in Table 1.The textural properties of the K-(USY-ATP) catalyst extrudate were severely affected upon cascaded bio-oil deoxygenation. The isotherm for the spent catalyst sample revealed a substantial loss in micro- and mesopore volume (black series Fig. S5c). BJH analysis derived from pore-size distribution plot (Fig. S5d) confirmed that the mesopores were partially shuttered after reaction. Besides, an important loss in surface area after reaction was noted too. The affected textural properties were partially restored upon regeneration, as the improved values of micro- and mesoporosity and BET indicate, but to a much lesser extent than with the zeolite ZSM-5-based catalyst material.The acidity of fresh, spent and regenerated catalysts was assessed by FT-IR spectroscopy with pyridine as probe molecule.The Brønsted acid sites (BAS) of the ZrO2/ds-ZSM-5-ATP catalyst originate from the zeolite component [51] and are indicated by the 1545 and 1636 cm−1 bands in Fig. 3 a [59,60]); the Lewis acid sites (LAS) come from the acidic Al3+ ions (indicated by the bands at 1455 and 1620 cm−1 [59–62]), the Zr4+ from the ZrO2 component, and from different cations within the attapulgite (e.g., Fe2+/3+, Mg2+, Ca2+ [62,63] indicated by the multiple bands in the range 1443–1455 cm−1. Quantification of the BAS was based on the integration of the 1545 cm−1 band -for fresh, spent and regenerated catalysts- while the ca. 1448 cm−1 band was used for determining the LAS [61]. After ex-situ CFP reaction the overall acidity (BAS + LAS) of the ZrO2/ds-ZSM-5-ATP catalyst dropped significantly, being specially affected the strong sites (∼75% drop) (Fig. 3b, Fig. S6a). This may indicate that the strong Lewis and Brønsted acid sites are the main contributors to the catalyst activity and that deactivation occurs due to site poisoning [53,54]. Indeed, the induced mesoporosity of the desilicated ZrO2/ds-ZSM-5-ATP catalyst prevents from deactivation by pore occlusion [35,50]. Upon regeneration most but not all acidity could be recovered, however, being the strong acid sites those with better recoverability once combusted the bulky coke.The LAS recovered less than the BAS (Fig. 3b) which might point at likely changes in the attapulgite cations and Zr after reaction and regeneration. Indeed, it should be noted that the band located at 1612 cm−1, attributed to pyridine interacting with cus (coordinatively unsaturated sites) and Zr species [64,65] (Lewis acidity, PyL), was of considerably higher intensity for the fresh than for the regenerated catalyst material (Fig. S6b). However, when compared to the intensity of other cations interacting with pyridine – i.e., band at 1620 cm−1- the relative intensity of PyL-Zr was larger for the regenerated than for the fresh catalyst material. This stronger PyL-Zr interaction is in line with the predicted ZrO2 re-dispersion after reaction and regeneration.Pyridine FT-IR studies showed the very limited acidity of the K-(USY-ATP) catalyst in line with its high Si/Al ratio (∼400). Yet the bands at 1443 cm−1 (attributed to pyridine adsorbed onto Lewis acidic K+ cations [61,66]) and 1447 cm−1 (attributed to pyridine adsorbed onto a smaller cation present in the attapulgite, such as Al3+, Fe3+, Mg2+ and Ca2+) [56] showed the presence of weak LAS, which disappeared after outgassing at 150 ⁰C [61] (Fig. 3c). Upon cascaded bio-oil upgrading and regeneration was observed a good recovery of the sites within the attapulgite clay (indicated by the 1447 cm−1 band, see Fig. 3d). Quantification of the weak LAS, by integration of the 1447 and 1443 cm−1 bands, revealed that the total concentration was very low for the fresh and regenerated samples (Table 1). However, the bands attributed to K+ (1443 cm−1) lost their intensity, indicating some relocation or partial loss after reaction and regeneration. Note that K+-sites, located in the sodalite cages of the FAU structure [15,56], would be significantly hindered/blocked in case structural damage occurs.While acidity is the key to activity in catalytic pyrolysis, basicity is so for carrying out bio-oil deoxygenation. Bulk basicity of fresh, spent and regenerated K-(USY-ATP) catalyst extrudates was determined by CO2-TPD and summarized in Table 1. After performing the catalytic reaction, a decrease of 70% in the number of basic sites was quantified. This dramatic basicity drop might be likely due to the consumption of basic OH groups during reaction (note that the FT-IR spectra of the spent catalyst in Fig. S2c did not show the characteristic stretching band O-H) and to structural damaging [18]. In line with the latter, oxygen vacancies might get clogged and K-OH sites (located in the sodalite cages of the USY zeolite [15,56]) inaccessible.Rather than being recovered after regeneration, basicity dropped further due to likely zeolite structural damage suffered from the gases/steam generated by coke burning. Besides, the attapulgite phase may have been altered upon regeneration, getting affected oxygen anchoring to its basic sites in the form of alkaline cations (such as Mg2+, Ca2+ and K+ [56]).X-ray diffraction (XRD) was employed to measure the structural integrity of the fresh, spent and regenerated catalyst samples.The orthorhombic phase of the zeolite ZSM-5 framework (Pnma space group, PDF 00-044-0003 [67]) remained well preserved after reaction (Fig. 4 a-b). Nevertheless, the shape of some X-ray diffraction peaks changed a bit, including the transformation of the initially split peak at 2θ 26.8° and 27.1° into a single peak, and an intensity increase compared to the peak at 27.6° (Fig. 4b). These changes are typically attributed to the incorporation of organics within the zeolite framework [68,69]. A carbon phase (PDF 00-026-1077) assigned to graphitic coke can be tentatively identified in the XRD pattern of the spent catalyst sample at 2θ 31.5° [70–72], which is absent in the fresh sample. By contrast, the XRD peak seen at 31.0° (2θ) for the fresh sample and assigned to quartz, present as impurity in the binder [73], disappeared after catalysis. This is likely due to either a phase change of attapulgite, or agglomeration of phases due to sintering upon high temperatures.The coke deposited during catalysis led to an increase in the b and c lattice parameters (Table 2 ) [69]. This was counterbalanced by a contraction of the lattice parameter a , likely related to slight dealumination [74], caused by contact with moisture coming from biomass vapors [75]. As response to these mild structural changes, it is observed a slight reduction of the crystal domain size, indicated by the decrease of the LVOL-IB value in Table 2.Upon regeneration, the original crystallite size was almost totally restored, as evidenced by a LVOL-IB value close to the shown by the fresh sample (Table 2). The XRD pattern also showed the re-appearance of the doublet peaks at (2θ) 26.8 and 27.1°, the original shape of the peak at 27.6°, together with the disappearance of the carbon phase at 31.5° (Fig. 4b). Regeneration resulted in lattice parameter changes opposite to those observed after reaction: b and c shortened, an indication of coke depletion, and a expanded back to the initial value of the fresh sample. The regeneration conditions, which expose the sample to high temperature and gas formation during the removal of coke deposits, could have possibly led to phase transformations of the attapulgite component. The absence of a quartz peak at 31.0° (2θ) in the regenerated sample indicates that the re-dispersion or phase change suffered upon reaction was irreversible. In addition, the unit cell size again increased in the regenerated sample, although the value of the fresh sample was not fully recovered.Aluminum speciation in the fresh, spent and regenerated ZrO2/ds-ZSM-5-ATP catalytic material has been assessed by 27Al MAS (Fig. 4c) and 27Al MQ MAS NMR (d) analyses. The main resonances of the fresh catalyst sample were located at 53 ppm (red series, Fig. 4c), assigned to framework tetrahedral coordinated Al species (A, AlIV) [76,77], and at ca. 3 ppm, assigned to extra-framework octahedrally coordinated aluminum species (E, AlVI) [78]. Interestingly, less well-defined extra-framework penta-coordinated Al (C, AlV) species were also present for the fresh sample at ca. 30 ppm. After reaction the signal intensity of framework AlIV species (so-called as A) decreased, in line with a loss in crystallinity, yet with concomitant signal increases in intensity of C and E at 30 ppm and 3 ppm, respectively (black series, c). Besides, the resonance of octahedral AlVI species, E, shifted to a lower chemical shift. This change was better observed in the 27Al MQ MAS NMR spectra (d) and believed to be related to distortions caused by pore coverage by coke [79]. In addition to the shift towards higher field, a shoulder emerged at 10 ppm (D, AlVI’) (Fig. 4c), also attributed to broadening/structural distortions suffered upon coke build-up at high reaction temperature [74–76] (see expansion of D in Fig. 4d).The acquired NMR spectra were fitted according to the different resonances identified (A → F) (Fig. S7) and the areas of the main deconvoluted spectra of the fresh, spent and regenerated catalyst samples were integrated to estimate the ratio between framework and extra-framework Al species (Table S1). The apparent drop of quantified framework Al sites after reaction is in line with the drop in Brønsted acidity noted above [76,80]. Yet, after catalyst regeneration, most of these Al sites seemed to be restored -as indicated by the increased intensity of A (green spectrum in Fig. 4c)-, recovering the framework tetrahedral coordination lost upon reaction. The signal associated with extra-framework AlVI species, E, returned to its initial frequency position (c), while a large downfield distortion of signal D was noted. The increase in distortion of the AlVI species after regeneration compared to the spent sample is likely associated with re-arrangement processes [81,82] of the attapulgite binder material during the thermal treatment, as a result of the gases/steam created upon coke burning. However, the overall ratio between framework and extra-framework Al was notably recovered, as shown in Table S1.The XRD pattern of the K-(USY-ATP) catalyst material (red series, Fig. 4e-f) show the typical cubic pattern of the FAU phase of zeolite USY (PDF 00-045-0112) (see Table S2). As noted above, the diffraction peak observed at ca. 2θ = 31.0° (b) corresponds to the hexagonal phase of quartz (PDF 01-089-1961) [67], present as impurity in the attapulgite clay.No meaningful structural changes are observed for the spent K-(USY-ATP) catalyst (black series in Fig. 4e). Contrarily to what observed for ZrO2/ds-ZSM-5-ATP, no diffraction peak assigned to carbon was seen at 2θ = 31.5° for spent K-(USY-ATP) (black series), pointing at a more amorphous nature of coke.The XRD pattern of the regenerated sample (green series) did show considerable changes regarding the fresh and spent catalysts patterns, such as the lower intensity and the overall broader peak widths. These may be an indication of a drop in crystallinity, compared to the fresh and spent samples. Indeed, significant unit cell shrinkage and reduced size of the crystalline domains (nearly by half) were confirmed upon estimation of the lattice parameters and crystallite sizes (Vcell and LVOL-IB values, respectively, in Table 2). This decrease in crystallinity may have been provoked by the high temperature and the steam formed during coke burning upon catalyst regeneration, likely provoking hydrolysis of the grafted K species (SiOK+ + H2O → SiOH + KOH), and ultimately loss of the catalyst’s basicity. Although smaller these ca. 30 nm crystals do keep their structural properties to a great extent; otherwise, a dramatic decrease of the textural properties would have been observed.When examining the morphology of the ZrO2/ds-ZSM-5-ATP catalyst no significant changes were seen after reaction and regeneration compared to fresh samples. Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images, which are shown in Fig. S8, reveal well-preserved zeolite ZSM-5 grains (of ca. 0.5–2 µm) agglomerated within 4 to 10 µm particles, and embraced by needle-like agglomerates of attapulgite (0.5 µm × 30 nm), accumulating more on edges and surface zeolite defects. Note that quartz impurities [73,83] were detected in the attapulgite in fresh (f) and regenerated samples (h). The TEM images show segregated domains of binder and zeolite, of varied sizes and shapes, unaltered after reaction and regeneration (Fig. S8c-h). Yet, a likely presence of coke-derived carbon flakes [75] were observed in TEM images of the spent catalyst, deposited on the edges of the zeolite crystal (left side Fig. S8g). Also, a high magnification TEM image of the fresh ZrO2/ds-ZSM-5-ATP catalyst (Fig. S8f) revealed the presence of nanosized ZrO2 particles distributed on the zeolite crystals where this component was originally deposited via impregnation, but also on the attapulgite clay binder material. This observation could indicate that ZrO2 was re-dispersed during granulation and calcination. Furthermore, ZrO2 might have re-dispersed after regeneration, as illustrated by the aggregated clusters formed onto the regenerated ZrO2/ds-ZSM-5-ATP catalyst (Fig. S8h).Re-dispersion effects were further studied by µ-XRF (Fig. 5 a-b). On the fresh catalyst body (a) can be seen that Zr species preferably remained close to zeolite domains. It should be noted that zeolite domains are easily identified by the higher presence of Si -predominant in the zeolite- and in particular by the absence of a Mg signal which is unique for the attapulgite clay. On the contrary, Zr species seem to re-disperse upon reaction and regeneration, as indicated by the Zr signal of the regenerated catalyst body (Fig. 5b) which is also detected at the attapulgite domains.SEM (Fig. S9) and TEM images (Fig. 5c,f) of the fresh K-(USY-ATP) catalyst material showed particle sizes of 300–500 nm. Attapulgite aggregates, identified by its characteristic needle-like morphology, were found on the zeolite crystal edges. Compared to the fresh sample, the zeolite particles seemed irreversibly clustered and of smaller size after reaction (d,g) and regeneration (e,h), in support of the structural modification noted above.In the high magnification image of the fresh catalyst (Fig. 5f) mesopores -generated via dealumination- with dimensions of ca. 20 nm can be distinguished. These were not observed with any clarity in the spent catalyst (g), however, presumably due to coverage by coke deposits. In the case of the regenerated sample (h), cavities can be seen, likely created by interconnection of mesopores [84].Upon ex-situ CFP, the ZrO2/ds-ZSM-5-ATP catalyst developed coke consisting of highly polyaromatic deposits, as indicated by TGA-MS, FT-IR and UV–Vis DRS. Confocal fluorescence microscopy mapping showed that the coke deposits are distributed heterogeneously, i.e., with an egg-shell pattern, over the technical catalyst: being more poly-aromatic and abundant on the surface, and its concentration diminishing progressively towards the inside of the body. By applying Raman spectroscopy, the soluble fraction of coke was estimated to range between 5 and 10 Å in size. These large coke species locate in the zeolite’s mesopores, partially blocking the strong acid sites, but also form externally, affecting the catalyst’s textural properties too. This was demonstrated with Ar physisorption and py-FT-IR spectroscopy, which showed drops in both (meso)pore volume, surface area and acidity, thereby especially affecting the strong (and Lewis and Brønsted) acid sites. Coke deposition also led to structural changes, as observed by XRD with the expansion of the lattice parameters.Another structural change pointed out by XRD is the re-distribution/phase change of attapulgite, evidenced by the observed contraction of the unit cell of the ZrO2/ds-ZSM-5-ATP catalyst after reaction. Such a change was also observed in the 27Al MAS NMR spectra, which showed an enhancement of the EFAl species in the spent catalyst at the expense of the tetracoordinated framework Al species. Penta- and, in particular octahedral Al species, increased in amount and extension, i.e., broadened, suggesting relocations upon catalytic reaction. No significant morphological changes were seen in the TEM images of the spent sample.Regeneration, carried out in static air for 6 h at 550 °C to burn-off the coke deposits, efficiently restored the textural properties and acidity previously affected by the formed coke, ruling-out any significant structural collapse. TNH3-TPD and FT-IR spectroscopy with pyridine as probe molecule did, however, confirm some irreversible loss in the acidity for the regenerated catalyst material. FT-IR also showed a relative larger intensity of the PyL-Zr band for the regenerated catalyst relative to the fresh catalyst, which suggests ZrO2 re-dispersion, as confirmed by µ-XRF.The 27Al MAS NMR measurements indicate that the distortions created after regeneration are more pronounced than after reaction, likely due to re-arrangement processes of the attapulgite as a result of the gas/steam created under the regeneration conditions. Structural changes after regeneration were also observed by XRD, as indicated by the disappearance of the quartz impurity in the attapulgite binder. Despite some distortion of the EFAl species, the framework Al species were mostly recuperated. The attapulgite phase is believed to play an important role in this regeneration given its high content in SiO2 and Al2O3, which serve as reservoir upon Al distortion effects suffered during reaction.After bio-oil upgrading the K-(USY-ATP) catalyst extrudates suffered structural damage and pore blockage by coke. It was determined by TGA-MS, FT-IR, UV–Vis DRS and confocal fluorescence microscopy that the coke deposits consisted of a large proportion of naphthalenes/anthracenes and poly-aromatics, distributed again in an egg-shell manner over the spent catalyst extrudate. However, these coke deposits were of a softer nature (i.e. H-richer) than the coke deposits formed on the ZrO2/ds-ZSM-5-ATP catalyst used for the catalytic fast pyrolysis stage. This is in line with the latter being upstream and directly exposed to the crude pyrolysis vapors, while the K-USY-based catalyst further upgrades the already treated vapors. Another factor for which less poly-aromatic coke formed onto the K-USY-Attapulgite catalyst is the lack of strong acid sites where it develops more easily. The textural properties of the spent catalyst nevertheless were severely affected by coke formation and morphological damages, as determined by physisorption and TEM measurements. The observed clustering had detrimental consequences for the required basicity of the catalyst material, as determined by CO2-TPD, by means of clogged oxygen vacancies or hindering accessibility to the K+ and OH– sites.Loss of structural integrity was enhanced upon catalyst regeneration due to the steam formed during coke burning, with the complete loss of basic sites. XRD measurements revealed the shrinkage of the zeolite unit-cell as well as a loss of crystallite size in the regenerated catalyst. The original textural properties, affected by coke deposition and regeneration, could only be partially recovered, less so than with the CFP catalyst. The Py-FT-IR spectroscopy studies indeed suggested some loss in K-loading after reaction and/or regeneration. The changes in physicochemical properties correlated with the morphological changes, as observed by TEM, for the zeolite material regenerated.A cascade process, consisting of a thermal pyrolysis followed by a two-step catalytic ex-situ catalytic fast pyrolysis (CFP) -catalyzed by a ZrO2/desilicated zeolite ZSM-5-/attapulgite material as solid acid- and the subsequent catalytic upgrading/deoxygenation of the formed oil -catalyzed by a K-(zeolite USY-attapulgite) material as solid base- was studied for the production of bio-oil from lignocellulosic biomass. A high bio-oil mass yield was achieved (40 wt%) with a remarkable deoxygenation degree (60 wt%), compared to a non-catalytic thermal bio-oil.Upon ex-situ CFP the solid acid ZrO2/desilicated zeolite ZSM-5-attapulgite catalyst suffers acid site coverage by the build-up of coke deposits and (reversible) changes in the Al coordination. In the case of the base K-(zeolite USY-attapulgite) catalyst, employed to further upgrade the formed bio-oil via CFP, mild pore blockage by coke formation and a partial loss of structural integrity was observed. Yet, the main deactivation cause is ascribed to clustering of the crystallites which hinders the catalyst’s basicity.Regeneration of the deactivated catalysts by coke burn-off to a large extent reverted the negative effects of the coke deposition on the ZrO2/desilicated zeolite ZSM-5-attapulgite catalyst. Although the structural distortion suffered by the catalyst upon pyrolysis can be considered mild and reversible, the small losses in framework Al and acidity upon a complete reaction plus regeneration cycle will progressively attenuate its activity. This, together with the observed ZrO2 migration after reaction and regeneration will eventually require catalyst replacement by fresh material after few reaction cycles.In contrast, the regeneration procedure caused irreversible structural change in the K-(zeolite USY-attapulgite) catalyst. It is believed that the steam produced upon coke burning provokes hydrolysis of the grafted K species, with the KOH produced attacking the zeolite’s structure, which ultimately experiences a total loss of basicity. This implies that after only one reaction cycle the K-(zeolite USY-attapulgite) catalyst would need to be replaced, with its associated economic implications for the process.Alternative regeneration procedures which efficiently restore the initial properties of the alkaline-exchanged USY catalyst material or the revision of the grafting procedure which, as seen, negatively impacts on the structure of the catalyst in the downstream process, might offer an alternative here. Among them, performing the regeneration at milder temperature conditions and in a flow reactor rather in a muffle furnace at static conditions would shorten the long exposure of the catalyst to the produced water vapors upon coke combustion, likely preventing structural damage and basicity loss. Also, to correct for hindered basicity a newly alkaline grafting process could be accomplished after regeneration, replenishing the extinguished K-OH sites. Testing of these newly proposed recoverability methods and the determination of the catalysts’ lifespans, i.e., the number of regeneration cycles that the catalysts can survive before getting replaced, could be topics for future investigation.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.Authors gratefully acknowledge the financial support from the European Union Seventh Framework Programme (FP7/ 2007-2013) under grant agreement n°604307 (CASCATBEL project). Dr. T. Hartman (Utrecht University, UU) is thanked for recording the Raman spectra, while Dr. A.-E. Nieuwelink (UU) is acknowledged for performing the µ-XRF analysis. The NMR experiments were supported by the Netherlands Organization for Scientific Research (NWO) within the Middelgroot program (no. 700.58.102 to M.B.), and uNMR-NL, an NWO-funded National Roadmap Large-Scale Facility for The Netherlands (no. 184.032.207).A.M.H.G. contributed to the idea of this study, conducted experimental work, processed the results and wrote the manuscript. R.M.D. and E.T.C.V. calculated the XRD lattice parameters, interpreting the results. H.H. and D.P.S. run the catalytic tests and participated in the interpretation of the results. K.H. and M. B. performed the 27Al (MQ) MAS NMR measurements. P.C.A.B. contributed to the idea of this work and aided on the discussion of results and manuscript writing. B.M.W. contributed to the idea of this study, including manuscript design and writing.Supplementary data to this article can be found online at https://doi.org/10.1016/j.jcat.2021.09.029.The following are the Supplementary data to this article: Supplementary data 1

The modes of deactivation -and the extent to which their properties can be restored- of two catalyst bodies used in cascade for bio-oil synthesis have been studied. These catalysts include a solid acid granulate (namely ZrO2/desilicated zeolite ZSM-5/attapulgite clay) employed in ex-situ catalytic fast pyrolysis of biomass, and a base extrudate (K-exchanged zeolite USY/attapulgite clay) for the subsequent bio-oil upgrading. Post-mortem analyses of both catalyst bodies with Raman spectroscopy and confocal fluorescence microscopy revealed the presence of highly poly-aromatic coke distributed in an egg-shell manner. Deactivation due to coke adsorption onto acid sites affected the zeolite ZSM-5-based catalyst, while for the base catalyst it is structural integrity loss, resulting from KOH-mediated zeolite framework collapse, the main deactivating factor. A hydrothermal regeneration process reversed the detrimental effects of coke in the acid catalyst, largely recovering catalyst acidity (∼80%) and textural properties (∼90%), but worsened the structural damage suffered by the base catalyst.

The current energy landscape based on fossil fuels created enormous economic and environmental problems such as greenhouse emission, climate change, solid waste pollution et al. Hydrogen economy is a clean alternative to fossil fuels [1–7]. Looking for an efficient, low-cost and safe hydrogen storage method is the prerequisite to the large-scale applications of hydrogen economy [8]. Mg stand as a candidate for hydrogen storage due to its large hydrogen storage density(7.6 wt%, theoretical value), fair accessibility and low-cost [9–11]. However, because of its sluggish reaction kinetics and high thermodynamic stability, Mg/MgH2 is not able to meet the requirements for practical application [12,13].Researches of past decades attempted many hopeful efforts to improve its hydrogen absorption/desorption properties, including nano-structuring [14,15], catalytic doping using transition metals [16–20], metal oxides [21–23], and intermetallic compounds [24–27]. Tremendous physical and chemical methods such as ball-milling [28,29], thin-film deposition [30,31] and combustion synthesis [32], have been attempted to synthesize Mg-based composites. Since the last couple of years, high-energy planetary ball milling equipment is applied to produce structural defects in many kinds of materials [33,34]. Nevertheless, the particle size of magnesium will increase after ball milling because of its agglomerating and welding property. Thus, carbon materials, which are anti-welded, are frequently used as the additive during ball milling to prevent magnesium from agglomerating and welding.Many carbon materials such as carbon nanotubes (CNTs), carbon nanofibers and graphite themselves are hydrogen storage materials [35]. In order to improve the hydrogen storage performance of carbon nanotubes, Yoo et al. [36] introduced defects and doped Pd on CNTs, the defective CNTs with Pd particles at 1 atm and 573 K stored 1.5 wt% hydrogen. Similarly, Hirai et al. [37] produced Pd-doped graphite as the hydrogen storage material with PdCl2-graphite as the precursor. However, the limitations of carbon material for hydrogen storage are obviously. In order to store hydrogen in molecular state by physisorption, most studies have been carried out at high pressures (1–16 MPa) and low temperatures (80–133 K). Although chemisorbed hydrogen concentration can reached up to a higher level, the chemically absorbed hydrogen cannot be desorbed reversibly at room temperature because of their strong CH covalent bonds [38]. Therefore, carbon material seems unfavorable to use as hydrogen storage material solely. Developing Mg-carbon materials by ball milling for hydrogen storage have drawn considerable research interest. Huang et al. [39] investigated the effects of different carbon additives such as carbon black, graphite and multi-walled carbon on the hydrogen storage properties of magnesium and found that the composite containing graphite displayed a remarkable decrease in the desorption temperature. Liu et al. [40] prepared MgH2 cluster with size below 5 nm. Multi walled carbon nanotubes and graphene nanoparticles were used to limiting the dimensions of clusters. The small clusters significantly reduced the hydrogen desorption temperature comparing with the bulk MgH2. Zhou et al. [41,42] prepared crystallitic carbon from anthracite and used as additive during ball milling. The magnesium particles were milled to 20–60 nm and hydrogenated to β-MgH2. With the increase of milling time, γ-MgH2 of orthorhombic crystal is formed. The endothermic peak of γ-MgH2 is 53 °C lower than that of β-MgH2. Recently, Han et al. [43] successfully prepared oxygen-rich activated carbon by the mulch-assisted ambient-air synthesis for hydrogen storage. In Pd-supported carbon-Mg hydrogen storage composites, carbon materials are considered as spillover agents [44]. First, H2 dissociates on Pd surface, then, H atoms spill onto the carbonaceous materials towards Mg bulk.Among the transition metals, Ni exhibit excellent catalysis for the hydrogen absorption/desorption of MgH2. Shi et al. [45] attempted to elaborate the synergistic mechanism between Ni and carbon aerogel for Mg-based hydrogen storage composite. The results show that the synergistic catalytic effect was attributed to the charge transfer between Ni, carbon aerogel and MgH2. After the introduction of Ni/carbon aerogel, the dehydrogenation activation energy of the Mg-Ni composite was reduced to 86.3 kJ mol−1. Yao et al. [46] anchored uniform-dispersed Ni nanoparticles on grapheme oxide(GO) to prepare Ni@rGO as the catalyst for MgH2. The activation energy for the rehydrogenation of MgH2 Ni@rGO reduced to 47.6 ± 3.4 kJ mol−1. Coincidentally, Liu et al. [47] also used rGO as the supporter and anchored Ni3Fe on it to prepare the catalyst Ni3Fe/rGO. Ni3Fe/rGO shows an excellent synergistic effect on the hydrogen storage performance of MgH2. Ouyang et al. [48–50] did many works about the synergistic catalysis on Mg-based alloys. They found out that adding metals and their hydrides, such as In and Ce, was an effective way to improve the hydrogen storage properties of Mg-based hydrogen storage materials. The sluggish kinetics of Mg/MgH2 has been significantly improved.The previous literatures show that Ni and carbon materials exhibit remarkable catalysis for the hydriding reaction of Mg. Moreover, hybrid catalysts usually show enhanced catalytic performance comparing with the single-phase catalysts. Herein, practical experiments show that C atom and Ni atom can be incorporated into Mg crystal. In the present study, the microphysical processes of H2 absorption, H2 dissociation and H diffusion on Ni/C synergistic incorporated Mg(0001) were studied by first-principles calculations. The mechanism of Ni/C atoms for catalyzing hydrogen storage of Mg crystal were comparatively discussed.Mg powder with a particle size of < 0.074 mm was purchased from Tianjin Ruijinte Chemical Company, China, and used as received. Anthracite used as milling aid was purchased from Rujigou Mine, China. The anthracite has low volatile matter content (6.60 wt%, air dry basis), low ash content (8.55 wt%) and high fixed carbon content (83.00 wt%). Nano-nickel with a particle size of 20 to 100 nm was purchased from Aladdin Industrial Corporation. The hydrogen with a purity of > 99.999 vol.% was purchased from Jinghui Gas Company, Beijing, China.The anthracite coal which purchased from Rujigou Mine was used as the precursor to prepare the crystallitic carbon by demineralization and carbonization (supporting information). The Ni/C co-incorporated Mg was prepared by ball milling on a Fritsch Pulverisette-6 planetary ball-mill. The number of stainless steel balls with diameter of 3, 5, 10 and 20 mm was 500, 20, 2 and 2, respectively. The work revolution of the mill used in this study were speed of 300 r min−1, ball to sample weight ratio of 30:1, milling time of 3.0 h, weight ratio of Mg powder to nano-nickel to carbon was 7:1:2. As a contrast, C-incorporated Mg was prepared with the Mg powder to carbon weight ratio of 8:2 and Ni-incorporated MgH2 was prepared with the MgH2 to nano-nickel weight ratio of 9:1. The atmosphere in the vial during milling was Ar.X-ray diffraction (XRD) measurement was carried out on a D8 Discover X-ray diffractometer with Cu Kα radiation using a step of 0.02°. Transmission electron microscopy (TEM), high-resolution transmission electron microscopy (HRTEM) and selected area electron diffraction (SAED) were carried out on a JEOL JEM-2100F electron microscope operating at 200 kV.All the first-principles calculation based on density functional theory (DFT) was carried out on the DMol3 program package of Materials Studio 7.0. The generalized gradient approximation (GGA) parameterized by Perdew-Burke-Ernzerhof(PBE) was adopted for the exchange-correction calculation [51,52]. A (6 × 6 × 6) Monkhorst-Pack k-points was used in the optimization of Mg cell, while a (4 × 4 × 1) k-points was used in supercells optimization, and the vacuum space of different Mg(0001) supercells was set at 20 Å as a periodic boundary condition [53].The Mg(0001) supercell was constructed by 5-layers of (3 × 3) Mg(0001) slabs containing 45 Mg atoms, as shown in Fig. S1(supporting information). There are four nonequivalent interstitial sites for the absorption of C and Ni atoms, namely top, bridge, hcp and fcc (Fig. S1, supporting information) [54,55]. The supercells of the Ni/C incorporated Mg(0001) were constructed by absorbing Ni/C atoms on the topmost surface of clean Mg(0001). Finally, fcc was preferred as the absorbing site to construct the incorporated supercells (supporting information).The convergence criteria were set as energy tolerance of 1.0 × 10−5 Ha atom−1, self-consistent field (SCF) tolerance of 1.0 × 10−6 Ha atom−1, maximum force gradient of 0.002 Ha Å−1, and maximum atomic displacement of 0.005 Å. The Ni and C atoms were treated with spin polarization and performed using different orbitals for different spins. For clean and C-incorporated Mg(0001), the core treatment of All Electron and the smearing broaden of 0.005 Ha were chosen. While, for the Mg(0001) with Ni atom, the core treatment of DFT Semi-core Pseudopots and the smearing broaden of 0.01 Ha were chosen to make sure that the calculation processes are convergent and the results are reasonable. The Double-Numeric plus d-function was chosen as the global basis set.To clarify the structure of incorporated Mg models in calculation, the morphology and crystal structure of the crystallitic carbon and incorporated Mg samples are investigated. After demineralization and carbonization, the particles of crystallitic carbon become structured. Crystalline structure of the carbon shows that diffraction peak for graphite(0002) is obvious, suggesting that the carbon possess similar atomic arrangement as graphite. Moreover, the diffraction peak for graphite(10 1 ¯ 1) which is the characteristic peak of the amorphous carbon is appeared. After structural simulation according to the diffraction, the detailed lattice parameters are listed in Table S2(supporting information). Compared with anthracite coal, a/b of the crystallitic carbon is increased and c is decreased. Correspondingly, the length of the pi bond increases from 1.35 to 1.40 Å and the interlayer spacing decreases from 3.28 to 3.26 Å. Demineralization eliminates the mineral salt in the coal so that the interlayer spacing tends to shrink. Carbonization made the arrangement of C atoms regular and stretch pi bond. The crystallitic carbon can prevent Mg particles from cold welding and agglomerating and break the particles by its hard edges and protrusions to reduce the particle size of Mg [43]. Therefore, ball milling can crush the particle of Mg effectively. There are amounts of hexagon particles in the TEM image (Fig. 1 B) which are Mg particles. This structure is often seen in the Mg-based materials [56]. It indicates that monocrystal Mg flakes are peeled off during ball milling. The EDS mappings of Mg and C (Fig. 1C,D) shows that Mg and carbon is mixed well and carbon is distributed on Mg evenly.In the Ni/C co-incorporated Mg sample, Mg and Ni are the main phases and Mg2Ni cannot be detected, as shown in the XRD pattern in Fig. 2 A1. After hydrogenation, Mg is hydrogenated and formed MgH2. The diffraction peaks of Mg2NiH4 is appeared. This means atomic Ni which incorporates into the Mg surface diffused into the crystal grain of MgH2. MgH2, Ni and Mg2NiH4 became the main phases as shown in Fig. 2A2. Then, after dehydrogenation, the component of the material became complex. The diffraction peaks of Mg, Ni and incomplete dehydrogenated MgH2 are apparent. Mg2NiH4 releases H2 and turns into Mg2Ni indicating the incorporation of Ni is stable. The TEM image of Ni/C co-incorporated Mg after hydrogenation shows that MgH2 nano-crystals with tetragonal (P42/mnm) space group evenly distribute in the material (Fig. 2B). Based on the HRTEM observation (Fig. 2C), crystal domain with lattice fringes of 0.166 and 0.125 nm are apparent and can be indexed to (211) and (202) planes of MgH2, respectively. Diffraction spots of MgH2, Mg, Mg2NiH4 and Ni are presented in Fig. 2D. The results agree with the previous literatures [46,57]. With hyriding and dehyriding reaction, Mg2Ni/Mg2NiH4 can in-situ form in the ball-milled Ni-Mg/MgH2 samples. The addition of Mg2Ni alloys can improve the de/hydrogenation performance of Mg/MgH2 system [58].The TEM image of Ni/C co-incorporated Mg shows the irregular shape of Mg crystallites (Fig. 3 A) and EDS mapping analysis shows the homogeneous distribution of Ni and C over Mg (Fig. 3A1–A3). Herein, Ni/C co-incorporated Mg(0001) is constructed to investigate the behavior of H2 on Mg surface. Fig. 3B1–B4 shows the clean Mg(0001), C-incorporated Mg(0001), Ni-incorporated Mg(0001) and Ni/C co-incorporated Mg(0001) after geometry optimization. Obviously, C atom stably adsorbs on the second layer of Mg(0001) and Ni atom adsorbs on the first layer of Mg(0001).The adsorption energy of hydrogen is calculated by Eq. (1) as follows: (1) E ads = − ( E H 2 + E Mg − E total ) where Eads is the absorption energy of hydrogen, EH2 is the energy of hydrogen molecule, EMg is the energy of different Mg(0001) models and Etotal is the total energy of Mg(0001) model with H2. The results in detail are listed in Table S3(supporting information). It is undisputed that the incorporation of C and Ni atoms has an obvious effect on the H2 adsorption. With the incorporation of C and Ni atoms, the energy which emitted from the H2 adsorption decreases. The total charge density maps of different Mg(0001) with H2 adsorption are shown in Fig. 4 A–D. The electron clouds of H2 and clean Mg(0001) has no obvious interaction. However, with the incorporation of C and Ni, the electron clouds of H2 and different Mg(0001) become overlapped. After C atom enters into Mg crystal lattice, the electronic structure of Mg(0001) surface is changed. The interaction between the electron clouds of H and Mg gradually become clear. While Ni atom incorporates on Mg(0001) surface, it has a direct impact on the adsorption of H2. As an anchor, the electron cloud of Ni has obvious overlap with that of Mg and H2. To clarify the role of C and Ni atoms in the Ni/C co-incorporated Mg(0001), the deformation charge density maps and density of states (DOS) are studied. Obviously, the electron enrichment regions are distributed in the surrounding of C and Ni atoms and the electron depletion regions are around Mg atoms. It shows that C and Ni atoms obtain electrons from Mg. The unoccupied orbits of Ni and H2 has obvious overlap. Meanwhile, the DOS of Ni/C co-incorporated Mg(0001) with H2 adsorption indicates that C s has obvious hybridization with Mg s and p at an energy of −10.0 eV, and C p has evident hybridization with Mg s and p at the range of −6.2 to 1.3 eV. While the Ni d has apparent hybridization with Mg p at an energy of −0.8 eV. The orbital hybridization verifies that Ni and C atoms influence the electronic structure of Mg(0001), while the σ-bond of H2 still firm.The incorporation of C and Ni can reduce the barrier energy of H2 dissociation in different levels, as shown in Fig. 5 . The corresponding deformation charge density maps are shown in Fig. S2(supporting information). After dissociation, H atoms accept electrons from Mg. With the incorporation of C, the Mulliken charge of H reduces from −0.248 to −0.271. While when it comes to Ni, the Mulliken charge of H increases from −0.248 to −0.057. It indicates the influence of C and Ni incorporation for the electronic structure of H is opposite. Thus, the Ni/C co-incorporation may has an eclectic effect for the hydriding reaction of Mg. Actually, the incorporation of C can modulate the electronic structures of Ni and H, as shown in Fig. S2D (supporting information).With the incorporation of C, the barrier energy (Eb) of H2 dissociation on Mg(0001) reduced from 104.8 to 95.3 kJ mol−1. But when it comes to Ni, the barrier energy significantly reduces to 0.9 kJ mol−1, indicating that Ni catalyzes the dissociation of H2 effectively. However, the catalysis of C incorporation seems to be finite. While the Ni/C co-incorporation also made the barrier energy reduced dramatically due to the catalysis of Ni.The catalytic mechanism of Ni atom on H2 dissociation is clarified by the DOS and deformation charge density calculation as shown in Fig. 6 . The Ni d orbit and H s orbit both emerge at −5.6, −2.8 and −0.85 eV in DOS (Fig. 6A), suggesting that the two orbits have obvious overlapping effect. To visualize the fix action of Ni d orbit on hydrogen absorption, deformation charge density distributions of the Ni/C co-incorporated Mg(0001) after H2 dissociation were calculated. It is found that Ni and H atoms combine together after H2 dissociation. The Ni dz 2 orbit and H s orbit accept the electrons (red color regions), while the electron in Ni dxy orbit is concentrated around the regions closing to Mg atoms due to the delocalization of the free electrons of Mg (red color regions) and the dxy orbit which far from Mg is electron deficiency (blue color regions). Mg atoms which surrounding Ni atom, meanwhile, contribute the electrons (blue color regions). Ni d orbit plays an important role during H2 dissociation, as other transition metals such as Fe [59]. In contrast to the Ni/C co-incorporated Mg(0001), the Ni atom in Ni-incorporated Mg(0001) shows a closer combination with H atoms, as shown in Fig. 6C,D. The electron deletion region of Ni dxy orbit is contracted around x axis due to the strong electron obtain ability of Ni. Without C atom, the Milliken charge of Ni become more negative (from −0.324 to −0.357), indicating that C can weaken the electron obtain ability of Ni which is beneficial for the H diffusion.After H2 dissociation, H atoms need to diffusion into the crystal lattice of Mg to form MgH2, as shown in Fig. 7 . The energy barrier (Eb) of H diffusion on clean Mg(0001), C-incorporated Mg(0001), Ni-incorporated Mg(0001) and Ni/C co-incorporated Mg(0001) are 43.0, 47.8, 69.0 and 35.8 kJ mol−1, respectively. The diffusion of H in the crystal lattice of Mg is an endothermic process. Thus, the energy of reaction (Er) is positive. With the incorporation of C or Ni, the energy barrier of H diffusion increases while it compared with that of H diffusion on clean Mg(0001). Despite the incorporations of C and Ni can reduce the barrier energy of H2 dissociation, the incorporations hinder the diffusion of H at the same time. Though the incorporation of C hinders the diffusion of H, the energy barrier of H2 dissociation still higher than that of H diffusion which means the dissociation of H2 is the limiting step during Mg is hydrogenated. Therefore, the hindrance of C incorporation for H diffusion has no obvious effect on the hydrogenation of Mg. However, Ni incorporation can significantly reduce the energy barrier of H2 dissociation and makes the H diffusion become limiting step. During the diffusion of H, Ni atom plays a role of anchor and fixes H atom around it, as shown in Fig. 6. So that the incorporation of Ni hinders the H diffusion on Mg(0001). The increase of the energy barrier of H diffusion against the hydrogenation of Mg. The Ni/C co-incorporation, which is different from the single incorporation of Ni and C, can reduce the energy barrier of H diffusion. During the dissociation of H2 on Mg, the Ni/C co-incorporation, similar to the Ni incorporation, can significantly reduce the energy barrier. Then, H diffusion becomes the limiting step. However, the Ni/C co-incorporation can reduce the energy barrier of H diffusion at the same time. Thus, the Ni/C co-incorporation shows the best catalysis for the improvement of the H2 absorption performance of Mg compared with the single incorporation of Ni and C.The analysis of H2 dissociation and H diffusion on different Mg(0001) show that H2 dissociation is the limiting step of the hydriding reactions of clean and C-incorporated Mg(0001) and H diffusion is the limiting step of the hydriding reactions of Ni- and Ni/C co-incorporated Mg(0001). Thus, the energy barriers of the critical step of clean, C-incorporated, Ni-incorporated and Ni/C co-incorporated Mg(0001) hydriding reaction are 104.8, 95.3, 69.0 and 35.8 kJ mol−1, respectively. Obviously, the Ni/C co-incorproated Mg(0001) shows the best hydriding performance. Moreover, the value of its energy barrier is very close to the hydriding activation energy of Mg with Ni4@rGO6 in Refs. [46,39].In addition, the incorporation of C also can influence the incorporation of Ni on the co-incorporated Mg(0001) surface. In the experiment, Ni enters the Mg crystal and in-situ forms Mg2Ni/Mg2NiH4. Herein, the effect of C incorporation for Ni diffusion in Mg is investigated. The energy paths of Ni diffusion on clean Mg(0001) and C-incorporated Mg(0001) are shown in Fig. 8 . Obviously, the incorporation of C can significantly reduce the barrier energy of Ni diffusion on Mg(0001) surface. The barrier energy is reduced from 37.1 to 7.2 kJ mol−1. This means C-incorporated structure is beneficial for the incorporation of Ni on Mg surface. It is conducive to the formation of Mg2Ni/Mg2NiH4. In addition, the deformation charge maps and Mulliken charge in Fig. S3(supporting information) show that during the diffusion of Ni, the Mulliken charge of C increased stepwise and Ni became more negative at the same time. This means electrons is transferred from C to Ni.In summary, the first-principles calculations are performed to investigate the effect of Ni/C incorporation for the hydriding reaction of Mg(0001). The morphology and crystal structure of the Ni/C co-incorporated Mg sample show that the pi bond of carbon is stretched and monocrystal Mg flakes are peeled off by ball milling. Carbon is distributed on Mg evenly to create conditions for the formation of C-incorporated structure. With Ni incorporation, Mg2NiH4/Mg2Ni is in-situ formed in the Ni/C co-incorporated Mg sample after hydrogenation and dehydrogenation. On the incorporated Mg(0001), the H2 molecule exhibits a better dissociation performance than that on clean Mg(0001). The catalytic effect of Ni on H2 dissociation can be ascribed to the bridging effect of Ni dxy orbit.However, the single incorporation of Ni or C is unfavorable for the H diffusion. The strong interaction between Ni and H hinders the activity of H on Mg(0001). This has a considerable effect on the hyriding reaction of Ni-incorporated Mg(0001) due to the H diffusion become limiting step of the hyriding reaction. But the incorporation of C can weaken the constraint of Ni for the H diffusion when it comes to Ni/C co-incorporated Mg(0001). The Ni/C co-incorporated structure in Mg(0001) shows an eclectic performance in H2 dissociation and H diffusion. The Ni/C co-incorporated Mg(0001) shows the best performance during hyriding reaction. The catalysis of Ni can effectively reduce the barrier energy of H2 dissociation and the incorporation of C can improve the H diffusion performance. Moreover, the incorporation of C is beneficial for the formation of Ni-incorporated structure. The present paper is helpful to clarify the catalytic roles of Ni and C in co-incorporated system on hydriding of Mg crystal.The authors would like to acknowledge the National Supercomputing Center in Shenzhen for their technical support of Materials Studio.This work is supported by the National Key R&D Program of China (Grant No. 2017YFB0103002), National Natural Science Foundation of China (Grant Nos. 51771056, 51371056, 51701043 and 52071141), Equipment Pre-research Field Foundation (Grant No. 6140721040101), Equipment Pre-research Sharing Technology (No. 41421060201), Changzhou Leading Talents Project (Grant No. CQ20183020), 333 Project in Jiangsu Province and the Priority Academic Program Development (PAPD) of Jiangsu Higher Education Institutions, Fundamental Research Funds for the Central Universities (Grant No. 2021MS051), Interdisciplinary Innovation Program of North China Electric Power University (grant number XM2112355).

Ni and carbon materials exhibit remarkable catalysis for the hydriding reaction of Mg. But the underlying mechanism of Ni/C hybrid catalysis is still unclear. In this work, density functional theory (DFT) calculation is applied to investigate the effect of Ni/C co-incorporation on the hydriding reaction of Mg crystal. The morphology and crystal structure of the Ni/C co-incorporated Mg sample show that the co-incorporated structure is credible. The transition state searching calculation suggests that both the incorporations of Ni and C are beneficial for the H2 dissociation. But Ni atom has a dramatic improvement for H2 dissociation and makes the H diffusion become limiting step of the hyriding reaction. The Ni dz 2 orbit and H s orbit accept the electrons and combine together compactly, while the Ni dxy orbit is half-occupied. The catalytic effect of Ni on H2 dissociation can be ascribed to the bridging effect of Ni dxy orbit. The incorporation of C can weaken the over-strong interaction between Ni and H which hindered the H diffusion on Mg(0001). The Ni/C co-incorporated Mg(0001) shows the best performance during hyriding reaction compared with the clean and single incorporated Mg(0001).

Currently, the production of sustainable and low-cost energy resources has become an urgent necessity to overcome the most serious issues that face the contemporary world about the future availability of fossil fuels (Sayed et al., 2020; Basyouny et al., 2021). Biodiesel or the extracted diesel from the bio-resources was assessed extensively to be introduced as an alternative fuel for the fossil fuel that can be produced by simple, low cost, and sustainable techniques (Lee et al., 2020; Abukhadra et al., 2020). Most of the addressed biodiesel products were extracted thermo-chemically from green resources (algae, biomass, vegetable oil, and extracts of plants) in addition to the animal fats (Abukhadra et al., 2021). Technically, the determining viscosity, lubricity, cetane number, and flashpoint of the biodiesel products are of acceptable values to be used directly as fuels in the engines or to be mixed with other types of fossil fuels in blends (Lawan et al., 2020; Zhong et al., 2020).Chemically, the biodiesel terms refer to series of fatty acid methyl esters that can be extracted from the transesterification process of triglycerides mainly in the vegetable and plants oils, or by esterification of free fatty acids (Al Hatrooshi et al., 2020). The spent or the waste products of cooking oil like corn oil and sunflower oil were evaluated as low cost, recyclable, and commercial raw oil instead of the fresh oils that are of expensive costs (Ibrahim et al., 2020). The transesterification process involved the chemical conversion of the studied types of oil either edible or non-edible using alcohol and in the presence of a potential catalyst (Sayed et al., 2020; Bin Jumah et al., 2021). The heterogeneous catalysts were recommended to be used considering the several technical, economic, and environmental drawbacks of the commonly used homogeneous catalysts (Rabie et al., 2019; Betiha et al., 2020). The solid heterogeneous catalysts that are of active basic and acidic functional groups were assessed as promising catalysts for their low cost, high reactivity, high thermal stability, significant recyclability, strong mixing, and dispersion properties, low hazardous byproducts, and facile separation properties (Sayed et al., 2020; Mohadesi et al., 2020).The acidic forms of the heterogeneous catalysts were assessed widely as promising environmental catalytic materials that can be used in both esterification and transesterification processes with significant activity (Nata et al., 2017; Wang et al., 2019). The applications of the acidic catalyst were recommended than the basic forms as it is of no saponification effects especially with the transesterification systems that contain low-grade oils as feeds of significant free fatty acid content (Wang et al., 2019). The fabrication of the acidic catalysts involved acidification and acidic functionalization of different types of inorganic materials as montmorillonite, mesoporous silica, mica, zeolite, clay minerals, and some metal oxides (Betiha et al., 2020; Farabi et al., 2019a; Cheng et al., 2019; Silva et al., 2018; Negm et al., 2019). The acidic inorganic catalysts exhibited some technical drawbacks during the transesterification processes related to their cost, low acidic densities, poor operational stability, and small pore sizes (Zailan et al., 2021; Mendaros et al., 2020).The catalysts that were prepared by acidification or sulfonation of carbon and carbonaceous materials showed significant catalytic activities in addition to their promising mechanical and thermal stability during the transesterification processes (Wang et al., 2019; Mateo et al., 2021; Shi et al., 2018). Most of the investigated carbon-based acidic catalysts were prepared by acidification processes for carbonized biomasses, biochar, hydrochar, activated carbon, polymers, cellulose, graphene oxide (Betiha et al., 2020; Flores et al., 2019; Fonseca et al., 2020). Therefore, there are several limitations about the possible use of such catalysts considering the availability and the price of the precursors in addition to the cost of the carbonization processes (Mendaros et al., 2020; Flores et al., 2019).Natural coal is of very high availability and has huge reserves in several countries with different ranks and qualities (Shaban et al., 2017; Yu et al., 2018). Structurally, it is composed of polycyclic sheets that are of aromatic hydrocarbons in the polymeric form (Tang et al., 2019). These aromatic rings are bonded to alkyl side chains and oxygen functional groups as carbonyl and hydroxyl (Yu et al., 2017). The structure of coal contains several organic compounds related to its content cellulose, lignite, resign, and the other macerals. These are associated with several organic chemical groups as carbonyl, carboxyl, and hydroxyl groups (Tang et al., 2019). Additionally, the present bridge bonds of -CH- or -O- induce the structural flexibility of coal which enhances the active groups loading process and reduces the steric hindrance impact (Yu et al., 2018; Tang et al., 2019). Therefore, it is an ideal structure for the synthesis of highly active acidic catalysts for the transesterification and esterification processes. However, little has been introduced to evaluate the possible acidification of coal-based activated carbon in the esterification reactions; no previous studies have been developed to evaluate the acidification of the raw coal itself as an acidic catalyst for the transesterification reactions (Tang et al., 2019). Egypt was pleased with about 50 million tons of coal as a geological reserve in the El-Mghara area, Sinai that is of no economic value until now.Therefore, the introduced study involved a novel investigation for the Egyptian raw coal as a potential acid catalyst without carbonization in the transesterification of waste sunflower oil either at low-temperature conditions or at high-temperature conditions. The acidification processes involved systematic sulfonation of the coal structure using different concentrations of H 2 SO4 with detailed inspection for the effect of the modification processes. The sulfonated coal catalyst (S.Coal) was applied in the transesterification processes considering different experimental factors. Additionally, the catalytic mechanism and the kinetic behaviors were addressed in the study. The raw coal samples were obtained from El-Maghara coal deposits, Sinai, Egypt. The full ultimate and proximate chemical composition of the used sample was presented in Table 1. Commercial sunflower oil and ultrapure methanol (99.8% purity; cornel Lab Company) were used in the transesterification studies. Analytical grade sulfuric acid (95%–98% purity; cornel Lab Company) was used as an acidification reagent for the coal sample. Commercial waste products for sunflower cooking oil were collected from the local restaurants and applied in the transesterification process. The fatty acid content as well as the physical properties of the studied oil sample was emphasized in Table S1.The raw coal fractions were ground gently using a normal home blender to size range from 20 μ m to 70 μ m . After the grinding, 10 g of the coal powder were immersed within 100 mL of sulfuric acid at 150 °C for 60 min at the inert condition of the nitrogen atmosphere. Then, the resulted suspension was cooled down to low-temperature conditions (40 °C) and washed extensively with distilled water until attending the neutral pH. Finally, the modified sample was dried for 180 min at 70 °C, kept in specified tubes, and labeled as SO3H/coal to be used in the further steps.The catalyst structure was investigated based on X-ray diffraction patterns utilizing a PANalytical X-ray diffractometer (Empyrean) with Cu-K α radiation considering the 2 Theta angle from 5° to 70°. The morphological changes under the sulfonation effect were followed using SEM images of the Scanning electron microscope (SEM, (Gemini, Zeiss-Ultra 55)). The petrographic properties of raw coal, as well as the S.Coal sample, were addressed by inspecting their thin sections under a Nikon polarizing transmitted microscope. The chemical structure before and after the modification reactions were investigated based on their FT-IR spectra using Fourier Transform Infrared spectrometer (FTIR-8400S) within measuring range from within determination range 400 cm−1 to 4000 cm−1. The species of the bonded ions to the surface of S.Coal as catalyst were determined utilizing X-ray photoelectron spectroscopy on a Thermo scalable Thermofihsre instrument (250 Xi, USA) occupied with Al K radiation source of monochromatic properties (1468.7 eV) with binding energies measuring range from 0 up to 500 eV. The densities of the incorporated acid groups in the structure of the synthetic S.Coal were determined based on the Boehm titration method. 0.5 g of the prepared S.Coal catalyst was added to solutions of NaHCO3 (0.05 M; 17 mL), Na2CO3 (0.05 M; 17 mL), NaOH (0.05 M; 17 mL), and Na2SO4 (1.0 M; 20 mL). After that, the mixtures were shaken for 24 h and the solid particles were separated by the filtration process. Then, 5 mL of the four aliquot solutions were acidified using diluted HCl (0.05 M) and the present acidic groups were determined by simple titration using NaOH (0.05 M) in the existence of indicator (phenolphthalein). The surface area of S.Coal as catalyst was inspected considering its N 2 adsorption/desorption curve using Beckman Coulter surface area analyzer (SA3100 type).The transesterification experiments were conducted inside an airtight autoclave (150 mL) connected to a digital heater of magnetic stirrer (600 rpm). The affecting factors on the activity of the S.Coal catalyst were followed from 5 min to 120 min as the reaction time, 5:1 to 25:1 as the methanol–oil​ ratio, 1 g to 5 g of S.Coal as the weight of the catalyst, and 40 °C to 120 °C as temperature. The experimental procedures were accomplished considering the essential step of careful filtration of the oil sample (38 mL) followed by gentle heating at 75 °C to remove the solid suspension and to avoid the side effect of the humidity. After that, a selected quantity of S.Coal was added to the oil sample and stirred for about 15 min to achieve homogeneous dispersion for the S. Coal within the sample. Then, the studied methanol volumes were added to the reactants at adjusted transesterification temperature for a certain time interval.By completing the selected time interval, the liquid phases were separated from the solid S. Coal fractions by filtrations using Whitman filter paper ( 40 μ m ). Then, the sample was poured carefully within a separating funnel to confirm the separation of the glycerol byproducts at the bottom of the funnel. After, the complete removal of the glycerol layers, the oil sample was heated at about 75 °C for 4 h to remove the residual alcohol content. The formed fatty acid methyl esters were inspected using gas-chromatography (Agilent 7890 A). The determination was performed after the controlled dilution of the biodiesel sample with n-hexane in the presence of methyl heptadecanoate as the reference standard. Based on the determined values of fatty acid methyl ester (FAME), the biodiesel yields were calculated directly from Eq. (1). (1) Biodiesel yield ( % ) = ( W e i g h t o f b i o d i e s e l × % F A M E ) W e i g h t o f o i l × 100 The gas chromatography technique of the Agilent 7890 A type was applied for the qualitative detection of the formed fatty acid methyl ester (FAME). The essential determination procedures included firstly controlled dilution of the collected oil samples after the tests using n-hexane. After that, the formed FAME components were determined using an Agilent-7890 A Series gas-chromatograph system which was connected with a split/splitless injector, flame ionization detector, and DB WAX capillary column (30 m × 0.25 m × 0. 25 μ m ) containing inert gas ( H 2 ) as a carrier. The temperature of both the detector and injector was adjusting during the analysis at 280 °C. The oven temperature was adjusted to be or regular increment from 120 °C up to 260 °C considering the increasing rate at a fixed value of 10 °C/min. The quantity of the present FAME was estimated considering the internal standard of methyl heptadecanoate injection.The physicochemical properties of the produced biodiesel in the studied system were followed based on the values of flash point, viscosity, cloud point, cetane number, pour point, density, and acid value. The actual value of viscosity was detected by Chongqing viscometer based on the recommended ASTM D445 test method. The flashpoint value was measured by the Penksy-martins flash tester based on ASTM D93 test method. The value of cloud point as well as pour point were measured by Lawler cloud point and pour point analyzer, respectively based on ASTM D2500 (cloud point) and ASTM D97 (pour point) tests. The value of the cetane number was determined by the Ignition quality tester based on the ASTM D613 test. The value of density was obtained by density hygrometer based on the ASTM D941 method. The acid value of the sample was determined by an automated titration system based on the ASTM D664 method.Regarding the structural properties, the XRD pattern of raw sub-bituminous coal declared the amorphous structure of the sample (Fig. 1). This was confirmed by detecting the broad peaks of amorphous carbon at 2Theta angles of 8°–30° and 40°–50° which was assigned to the amorphous structure of the aromatic carbon sheets of lattice plane (002) and (101) (Fig. 1A) (Akinfalabi et al., 2017; Wong et al., 2020; Farabi et al., 2019b). Such aromatic sheets of carbon are of random orientation in the carbonaceous components of coal (Akinfalabi et al., 2017; Farabi et al., 2019b). The SO3H functionalized coal (S.Coal) showed noticeable changes in the pattern as the first broad peak was shifted to a high position (10°–32°) and the second peak declined strongly (Fig. 1B). This demonstrates a change in the structure of the sample and a possible increase in the amorphization degree (Farabi et al., 2019b; Niu et al., 2018b). This behavior is related to the effect of the sulfonation process in breaking the C-O-C cleavage bonding in the carbonaceous components of coal and the significant increase in the basal spacing (Araujo et al., 2019). As a result, the sulfonated carbon altered significantly to be of more rigid and amorphous properties under the continuous disordering in the carbon units of the carbonaceous components of the coal (Farabi et al., 2019b). Additionally, the expected dehydration effect of the carbonaceous components to form polyaromatic/carbon on the surface of the coal sample is of significant impact on such changes in the XRD pattern (Akinfalabi et al., 2017). The obtained pattern for the spent is of no observable changes in the previously detected peaks of the amorphous carbon, only reduction in the intensity which declared the effect of the adsorbed oil or the glycerol on the surface of the S. coal catalyst (Fig. 1C).Based on the FT-IR spectra of raw coal and S. Ccoal samples prepared at different concentrations of H 2 SO4 were presented in Fig. 2. The raw sample demonstrated the common identification chemical groups of commercial coal (Fig. 2A). The essential chemical groups were detected are O-H groups (3000 cm −1 to 3600 cm−1), aliphatic –CH2 (2858 cm −1 and 2940 cm−1), ketone groups (2325 cm −1), C=C stretching (1716 cm−1), CO stretching (1616 cm−1), bending vibration of C-H of a methylene group (1450 cm−1), bending vibration of C-H in the methyl group (1372 cm−1), C-O stretching (1000 cm−1 to 1200 cm−1), and deformed C-H in the aromatic planes (500 cm−1 to 900 cm−1) (Fig. 2A) (Shaban et al., 2017; Yu et al., 2018). After the functionalization process, the recognized spectrum of the S. Coal declared a strong increment in the intensities and the broadness of the O-H related band around 3398 cm −1 (Fig. 2B to F). This is related to the stretching vibration of the COOH groups that were formed as a result of the enrichment in the acidic chemical groups within the structure of the coal sample during the strong oxidation of the carbonaceous components by the sulfonation reactions (Mateo et al., 2021; Wong et al., 2020). The intensities of the –OH-related bands increased significantly with the different values of H 2 SO4 concentrations reflecting increment in the incorporated acidic groups (COOH) with testing higher concentrations of the acid (Fig. 2B to F) (Farabi et al., 2019b; Araujo et al., 2019). This was observed also for the absorption band that related to the CO vibrational modes of COOH groups around 1695 cm−1 (Mateo et al., 2021; Yu et al., 2018). It appeared at higher intensities for the samples with were treated with the highest H 2 SO4 concentrations. The incorporated sulfur-bearing groups were identified by SO3 stretching (1174 cm−1) of the –SO3H groups, symmetric (1071 cm−1), and asymmetric (1001 cm−1) in addition to C-S groups (578 cm −1) (Fig. 2B to F) (Mateo et al., 2021; Fonseca et al., 2020). The essential bands of raw coal as the C=C stretching (2610 cm−1) and the C-H plane bending (1296 cm−1) were detected as reduced bands especially with using high concentrations of the acid (Fig. 2B to F) (Farabi et al., 2019b; Araujo et al., 2019). This might be related to the predicted destruction of the cyclic structure of cellulose and the oxidation of the present methyl groups in the structure during the oxidation process as they were converted into COOH groups (Mateo et al., 2021). Also, the intensity of the CO stretching-related band show (1629 cm−1) an obvious increment as a result of the oxidation process and the formation of new carboxylic groups (Fig. 2B to F) (Fonseca et al., 2020).XPS analysis was also performed to investigate the functional groups which were formed during the formation of S.Coal as an acidic catalyst in the conversion of sunflower oil (Fig. 3). The survey scan was estimated within the range from 0 to 1400 eV to confirm the change in the structure of the coal skeleton after the incorporation of –SO3H functional groups. The high resolutions scans clarify the existence of the identification peaks of C 1s, O 1s, and S 2p (Fig. 3A). The first spectrum of O1s spectrum reflected the existence of two chemical states for the oxygen ions at 531.7 (C=O bond) and 533.2 eV (S-O, S-OH, C–OH bonds) (Fig. 3B) (Yu et al., 2018; Tang et al., 2019). The first spectrum of C 1s demonstrates the presence of three chemical states at 288.6 eV (to C=O/O–C=O bond), 286.6 eV (C–O bond), and 284.5 eV (C-C/C-H bond) (Fig. 3C). Additionally, the S 2p spectrum reflects the chemical state of -SO3H at 168.7 eV confirming the incorporation of the sulfonic groups within the coal structure (Yu et al., 2018) (Fig. 3D). The optical properties of the investigated coal sample before and after the modification reactions were evaluated based on the obtained images from the transmitted polarized microscope (Fig. 4A, B, and C). The raw sample appeared with a characteristic macerals structure of coal that is composed mainly of the carbonized wood tissue (vitrinite) as the essential component. The present vitrinite contains several varieties of other species of macerals (Liptinite) as pollen, spores, resin, and cuticle of leaves (Fig. 4A and B). Additionally, other inorganic components were identified as cryptocrystalline silica, pyrite, and clay impurities. The SO3H-functionalization and the related oxidation process resulted in strong changes in optical features of the coal sample (Fig. 4C). The sample showed a strong reduction in the present inorganic impurities and the maceral structure appeared in a dark tone reflecting the influence of the oxidation processes (Fig. 4C).Regarding the morphology, the SEM image of raw bituminous coal reflected the presence of the sample as compacted layers (Fig. 4D). These layers are composed of irregular forms and stacked randomly above each other which are related mainly to the compression of the macerals and the wood tissue components (Fig. 4D). The modification reaction resulted in strong changes in the morphologies and the surficial features of the coal sample (Fig. 4E and F). The surface appeared to be of rugged and irregular topography with numerous nano nudes (Fig. 3G and H). Moreover, the structure appeared to be of observable microporosity which might be related to the dissolving of the inorganic components (Fig. 4H). This gives the sulfonated products a higher surface area and more exposed sites than the raw coal which will be of strong contribution in inducing the catalytic performance of the sample. Regarding the SEM image of the S.Coal after the transesterification process, the inspected featured reflected partial coating of the S.Coal particles with irregular materials that might be related to the adsorbed glycerol or the components of the studied S.SFO (Fig. 4I).The weight of the used catalyst is of a significant influence on the biodiesel yield which can be obtained by the transesterification of the spent sunflower oil over S.Coal catalyst (95% H 2 SO4). The effect of the catalyst dosage was followed from 1 g up to 5 g and the other controlling factors were fixed at 60 min as reaction time, 15:1 as a methanol-to-oil ratio, 600 rpm as stirring speed, and 40 °C as low-temperature conditions (Fig. 5A).There is a noticeable enhancement in the determined yields from 93.2% up to 98.4% with increasing the catalyst content from 1 g up to 3 g (Fig. 5A). This behavior can be explained based on the predicted increase in the catalytic active sites and the interacted surface area which induce the catalytic performance of the conversion process (Basyouny et al., 2021; Bhatia et al., 2020). Beyond 3 g, the used dosages of S.Coal (4 g and 5 g) are of negative effects and the produced yield decreased to 95.5% (4 g) and 89.2% (5 g) (Fig. 5A). The further increases in the catalyst weight above the limit cause a reduction in the homogeneity properties of the reaction system causing a reduction in the resulted yields (Sayed et al., 2020; Bin Jumah et al., 2021). Additionally, the unreacted particles of S.Coal increase the mass transfer resistance between the different reaction components which diminish the conversion yield of the reaction (Sayed et al., 2020).According to the stoichiometric equation of the transesterification reaction, each molecule of triglycerides reacts with three methanol molecules to create its FAME so the adjustment of the ratio between triglyceride and methanol is of a significant effect on the resulted yields (Negm et al., 2017). The influence of the methanol-to-oil ratio as a controlling factor on the transesterification process of sunflower oil by S.Coal (95% H 2 SO4) was investigated within the range from 5:1 up to 25:1. The other affecting parameters were fixed at 3 g as catalyst dosage, 60 min the reaction time, 600 rpm as stirring speed, and at low-temperature conditions (40 °C) (Fig. 5B).In the studied transesterification process for S.SFO over S.Coal as an acidic catalyst, the increment in the methanol ratio from 5:1 up to 20:1 resulted in an increase in the yields up to 98.8 % (Fig. 5B). The accurate amount of methanol is of vital effect in accelerating the reaction in the forward direction and in reducing the mass transfer resistance between the different phases of the reaction (Bin Jumah et al., 2021; Rabie et al., 2019). Moreover, the suitable methanol content enhances the solubility of the reactants and decreases the viscosity inside the reaction system (Singh et al. 2020). All the previous effects are of valuable impact in enhancing the catalytic activity of S.Coal and the achieved biodiesel yields for the studied S.SFO. At the studied methanol ratio of 25:1, the observed yield was declined slightly down to about 97.4% i.e. the best methanol-to-oil ratio is 20:1 (Fig. 5B). The reversible effect for the excessive content of methanol on the produced yields is related to the impact of the unreacted alcohol molecules in the deactivation of the catalytic sites and the dilution of the mixture which direct the reaction to the backward direction (Abukhadra et al., 2019).The transesterification time interval is a vital factor in controlling the miscibility and the reaction balance (Sayed et al., 2020). The influence of the time interval was inspected within the experimental range from 10 min up to 80 min for different temperature values from low-temperature conditions (40 °C) up to 120 °C as high-temperature conditions. The transesterification conditions were studied at 20:1 for the methanol/oil ratio, 3 g for the used S.Coal dosage (95% H 2 SO4), and 600 rpm for the stirring speed (Fig. 5C).At the low-temperature conditions (40 °C), the yields increased gradually with increasing the transesterification interval up to 60 min and the obtained yield attend 98.8%. The further expansion of the reaction time resulted in diminishing the catalytic activity and the produced biodiesel yield declined to be 98% after 80 min (Fig. 5C). This behavior is ascribed to the immiscible nature between the different reaction components which would hinder the catalytic activity of the S.Coal at the starting periods (Toledo Arana et al., 2019). Therefore, the reaction requires a considerable period to reach the equilibrium point (60 min) at which the mixing periods achieved the best homogeneity and low miscibility between the components. The elevation in the time to a higher interval than the equilibrium time (60 min) showed a desirable impact on the produced yield as the reaction was driven to the reversible direction (Basyouny et al., 2021).With increasing the temperature, the reactions show similar behavior as observed at the low-temperature conditions regarding the enhancement of the yield with time (Fig. 5C). There is a noticeable enhancement in the achieved yield at all the studied intervals as compared to the normal conditions of low-temperature conditions (40 °C). Moreover, the equilibration time and the required intervals to achieve the best yields declined strongly after the regular increase in the studied temperature. At 120 °C, biodiesel yields of 97.3% and 99.3% were determined after 10 min and 20 min, respectively (Fig. 5C). Additionally, the reaction equilibration was attained after 30 min achieving the highest yield (99.6%). Based on these results, it can be concluded that the performance of the transesterification reaction inside an airtight autoclave under high-temperature conditions is of strong impact in enhancing the catalytic activity of S. Coal and decreasing the required time for effective conversion processes. This behavior is attributed to the endothermic nature of the transesterification reaction and the increase in the reaction temperature cause increase in the kinetic energies of the system. As a result, the mass transfer between the different phases of the reaction will increase strongly and the rate of reaction will be accelerated significantly (Toledo Arana et al., 2019; Seela et al., 2020).The sulfonation conditions are of great influence on the stability of –SO3H and the number of active sites within the produced catalyst (Fonseca et al., 2020). Therefore, the effect of the sulfonation conditions was inspected as function of the sulfuric acid concentration (70%, 75%, 80%, 85%, 90%, and 95 %). The transesterification conditions were studied at 20:1 for the methanol/oil ratio, 3 g for the used S.Coal dosage, 600 rpm for the stirring speed, 60 min as the interval at low-temperature conditions (40 °C), and 30 min as the interval at high-temperature conditions (120 °C) (Fig. 5D).Based on the determined biodiesel yields, the highest percentages were obtained for the coal samples which were treated with the highest H 2 SO4 concentrations (85%, 90%, and 95%) either at the low-temperature conditions (40 °C) or at high-temperature conditions (120 °C) (Fig. 5D). It can be detected that using the sulfuric acid at a concentration above 90% is of slight effect on the catalytic performance of S.Coal. Such enhancement in the catalytic activity of S.Coal with using high concentrations of H 2 SO4 is related to the vital role of the acid in inducing the stability and the acidity properties of the catalyst (Tang et al., 2019). This can be clarified according to Luciatalia,s principle of equilibrium as the sulfonation process is a reversible exothermic reaction. Thus, decreasing the H 2 SO4 concentration would direct the reaction to the opposite direction and diminish the quantities of the incorporated sulfonated groups within the coal sample (Wong et al., 2020). This was supported by the previously investigated FT-IR analysis that reflected a significant increment in the intensities of the identification bands of sulfur-bearing chemical groups as O=S=O, SO3, -SO3H, and C-S groups.Such results also are of high agreement with the measured acid densities of the coal samples after the sulfonation processes with the different concentrations of H 2 SO4 (Tabel.2). The measured densities increased from 2.95 mmol/g up to 7.53 mmol/g with increasing the H 2 SO4 concentration from 70% up to 95%. Additionally, the increase in the sulfur content with treating the coal samples with high H 2 SO4 concentration confirms the effect of the concentration in inducing the stability of the incorporated sulfur-bearing groups (Table 2). Moreover, the slight enhancements in the surface area and the porosity with regular increase in the concentration of the used acid are of remarkable effect in inducing the activity of the S.Coal catalyst which was synthesized at high concentrations of H 2 SO4 (85%, 90%, and 95%) (Table 2). The recyclability properties of the produced S.Coal as acidic catalyst were assessed for five reusing cycles. The used S.Coal particles after the transesterification tests were collected washed carefully using distilled water three times and each time consumed 10 min. After that, the fractions were dried gently in an electric drier for 8 h at 60 °C to be reused in the next transesterification cycle. The transesterification conditions were selected at 600 rpm as the stirring speed, 20:1 as the adjusted methanol to oil ratio, 3 g as the incorporated S.Coal quantity, 60 min as the interval at low-temperature conditions, and 30 min as the interval at high-temperature conditions (120 °C) (Fig. 5E).The achieved yields reflect the considerable stability of the S.Coal as an acidic catalyst for the transesterification of S.SFO considering the addressed five cycles either at low-temperature conditions or at high-temperature conditions (Fig. 5E). The obtained yields for reusing cycles of S.Coal at the low-temperature conditions (40 °C) are 98.8% (Cycle 1), 98% (Cycle 2), 97.3% (Cycle 3), 95.4% (Cycle 4), and 93.5% (Cycle 5) (Fig. 5E). The obtained values at high-temperature conditions (120 °C) are 99.6% (Cycle 1), 99.3% (Cycle 2), 97.4% (Cycle 3), 94.53% (Cycle 4), and 89.6% (Cycle 5) (Fig. 5E). The linear decrease in the S.Coal with repeating the assessed cycles reflects the possible deactivation of its catalytic sites by the adsorbed glycerol and oil during the transesterification processes. Additionally, the expected leaching of the - S 3 OH groups during the tests affected negatively the activity of the S.Coal catalyst during the recyclability tests (Araujo et al., 2019). The observed high stability of the S.Coal sable at the low-temperature conditions (40 °C) as compared to high-temperature conditions (120 °C) demonstrate the effect of high-temperature and pressure conditions in accelerating the leaching of the –SO3H groups from the S-Coal sample.The effective transesterification processes that involve the promising conversion of S.SFO into biodiesel occur according to three steps as declared in Eq. (2), Eq. (3), and Eq. (4). The three equations represent successive chemical interactions between the present triglyceride (TG) as well as diglyceride (DG) and monoglyceride (MG) with about 1 mole of methanol which resulted in the formation of FAME (1 mole). Finally, the reactions resulted in about 1 mole of glycerol and 3 moles of FTAM as declared in Eq. (5) (Naeem et al., 2021). Considering the assumption of Eq. (5), the transesterification rate is controlled essentially by the concentrations of the liquid reactants either the TG or methanol content. (2) T G + Methanol ↔ D G + M E (3) D G + Methanol ↔ M G + M E (4) M G + Methanol ↔ G R + M E (5) T G + 3 Methanol ↔ G R + 3 M E The transesterification as a chemical reaction is of reversible properties which make the high alcohol content is of valuable impact in directing the reaction in the forward side until the equilibrium (Abukhadra et al., 2020). As the higher concentrations are of no significant impact without the existence of suitable concentrations of TG, the transesterification reactions appear to depend only on the availability of TG molecules. Therefore, the occurred reactions during the conversion of S.SFO over S.Coal are of pseudo-first-order kinetic properties (Roy et al., 2020). The Pseudo-First order equation in its linear form can be expressed by Eq. (6). (6) − L n 1 − Y M E = k t The kinetic investigation was conducted considering the time from 10 min up to 60 min, the temperature from 40 °C up to 120 °C, the methanol/oil ratio at 20:1, the S.Coal dosage at 3 g, and the stirring speed at 600 rpm (Fig. 5F). The obtaining fitting degrees (R2 > 0.9) reflected strong agreement between the kinetic behaviors for the transesterification of S.SFO over S.Coal and the assumption of the Pseudo-first order kinetic model (Fig. 5F; Table.S2). Such fitting results suggested the operation of the adsorption-surface reaction–desorption mechanism during the process and the pseudo-First​ order behavior is reasonable to represent the transesterification systems of S.Coal (Li et al., 2019).The value of the activation energy (41.05 kJ/mol) was obtained as a theoretical parameter based on the Arrhenius equation (Eq. (7)) and the conducted linear regression plots of ln k vs 1/T (Eq. (8)). This declares the suitability of the S.Coal transesterification system for S.SFO to operate effectively at mild conditions and low energy (Naeem et al., 2021) (Fig. S1). (7) k = e x p − E a R T (8) L n k = A . − E a R T + L n A The sulfonation mechanism depends strongly on the presence of electrophilic species to functionalize the aromatic components within the coal structure (Xiao and Hill, 2020). During the sulfonation process, the molecules of the used sulfuric acid will be affected by the protonation process as a result of the contentious reaction between them. This is related to the affinities of the -OH groups to grab hydrogen ions which causes a break for the present oxygen–hydrogen bonds forming good leaving groups of H 2 O (Tang et al., 2019). On the other hand, the protonated oxygen during this step will interact with one of its lone pair electrons creating bonds with the sulfur ions which leads to the formation of protonated trioxide (Tang et al., 2019). Such positively charged sulfur ions in their trioxide forms are of high electronegativity and act as strong effective electrophiles that are of significant role during the sulfonation of the coal structure (Yu et al., 2018) (Fig. 6A and B). The attack of the sulfur ions on the benzene rings resulted in destruction for the double bonds in the aromatic ring and this induced the incorporation of sulfur-bearing groups as –SO3H (Fig. 6A and B) which was confirmed by the FT-IR analysis (Fig. 2). On the other hand, two reacted molecules of sulfuric acid produced HSO 4 − radicals which act as a base to remove the hydrogen protons (Xiao and Hill, 2020). At the same time, there is strong oxidation process occurred for the OH groups converting them into carboxyl groups (COOH) (Fig. 6A and B) as observed in the FT-IR spectrum. The existence of COOH and –SO3H in addition to OH as highly active catalytic groups after the sulfonation process induced the catalytic efficiency of S.Coal as an acidic catalyst in the transesterification processes (Fig. 7). The observed declination in the catalytic efficiency of S.Coal samples with decreasing the concentrations of H 2 SO4 can be illustrated based on Luciatalia , s principle. According to the principal, such sulfonation reactions are of reversible properties and the decrease in the H 2 SO4 concentration directs the reversible direction which reduces the entrapment efficacy of the sulfuric groups in the structure of coal (Mateo et al., 2021; Fonseca et al., 2020). Moreover, the oxidation processes at the diluted concentrations of H 2 SO4 are of faint effect which resulted in the formation of aldehyde groups instead of the active carboxyl groups which affect negatively the efficacy of the catalyst.For the transesterification conversion of S.SFO over S.Coal as an acidic heterogeneous catalyst, it can be illustrated based on the chemistry of the oil. Chemically the vegetable oils are unsaturated and saturated monocarboxylic acids are known as triglycerides (Fig. 7) (Abukhadra et al., 2020). The triglycerides can be reacted with alcohol molecules during the transesterification process forming monoesters in the presence of the catalyst. This reaction involves essentially hydrolysis reaction for the ester groups between the present fatty acids and the glycerol (Fig. 7) (Basyouny et al., 2021). This causing generation of new types of ester bonds between the fatty acids and the alcohol molecules and then replacement for the glycerol by three monohydric alcohols. The predicted mechanism for the transesterification of S.SFO over the synthetic sulfonated coals as acidic catalyst (S.Coal) and the expected nucleophilic attacks of the reacted ester groups was emphasized schematically in Fig. 7 (Negm et al., 2017).The physical properties of the resulted biodiesel product at the best conditions over S.Coal as acidic heterogeneous catalyst were evaluated based on the international requirements for the suitable biodiesel as biofuels standards (ASTM D-6751 and EN 14214) (Table 3). The viscosity and the density of the tested biodiesel sample are of high agreement with the suggested values for the recommended biodiesel products. Additionally, the cetane index was measured at a promising value (more than 45) which declares the safety properties of the product if it was applied directly in the engines (Basyouny et al., 2021). Moreover, the measured value of flashpoint is suitable for the handling and the transport of the biodiesel product as fuel (Table 3). The results of the GC–MS analysis reflect the formation of palmitoleic acid methyl ester, oleic acid methyl ester, and linoleic acid methyl ester as the essential fatty acid methyl esters (FAME) (Table 3). Additionally, other species were detected at minor content including myristic acid methyl ester, palmitic acid methyl ester, eicosanoic acid methyl ester, stearic acid methyl ester, caprylic acid methyl ester, and behenic acid methyl ester (Table 3).The activity of S.Coal as a heterogeneous catalyst for the transesterification of waste cooking oil was compared with other studied catalysts in literature either the basic catalysts or the acidic catalysts (Table 4). The synthetic S.Coal as a heterogeneous catalyst displayed higher activity than several studied basic catalysts including zeolite-X, NiO, CaO, CaO/SiO2, KOH/Clinoptilolite, and, synthetic apatite. Additionally, it appears as a high active acidic catalyst as compared to the other acidic catalysts including some inorganic acidic catalysts (SO4/Fe-Al-TiO2, Fe2O3-MnO-SO4/ZrO2, and Ti(SO4)O) and sulfonated carbonaceous catalysts (Sulfonated graphene and Sulfonated AC from bamboo) (Table 4). The resulted high yield in addition to the determined acid densities in comparison with the other acid catalyst which were prepared by sulfonation of carbonaceous materials demonstrates the flexibility of the raw component to affect strongly by the sulfonation reactions in their natural form without carbonization. It was reported that the carbonization process is of negative effects on the carbon skeleton. The carbonization temperature causes accumulation for the structural carbon layers in a highly random orientation as a result of the destruction and collapse of carbon frameworks (Yu et al., 2017). This reduces the quantities of the present hydroxyl and carboxyl active groups during the formation of the catalyst (Flores et al., 2019). Therefore, raw coal without carbonization is a favorable precursor for the synthesis of acidic heterogeneous catalysts with significant acidic density and promising catalytic activity.Raw sub-bituminous coal was treated with sulfuric acid in a sulfonation process to produce potential acidic catalysts for the transesterification of S.SFO. The synthetic S.Coal catalyst using H 2 SO4 (95%) exhibits the best catalytic activity, acid density (8.4 mmo/g), and surface area (26.4 m2/g. At low-temperature conditions (40 °C), the S.Coal achieved a yield of 98.8 after 60 min using 3 g of the catalyst and a methanol/oil ratio of 20:1. However, at high-temperature conditions (120 °C), the yield was enhanced to 99.5%, and the time interval was reduced to 30 min only. The synthetic S.Coal catalyst is of considerable reusability and higher catalytic performances some addressed basic and acidic catalysts. Moreover, the extracted biodiesel at the best conversion conditions is of acceptable technical qualifications according to the international requirements. The results declare higher efficacy of the sulfonation process on the raw coal than the carbonized products. Sherouk M. Ibrahim: Visualization, Formal analysis, Writing – original draft, Writing – review & editing. Ahmed M. El-Sherbeeny: Conceptualization, Project administration, Visualization, Writing – original draft, Writing – review & editing. Jae-Jin Shim: Writing – original draft, Writing – review & editing. Ali A. AlHammadi: Formal analysis, Writing – original draft, Writing – review & editing. Mostafa R. Abukhadra: Conceptualization, Project administration, Visualization, Formal analysis, Writing – original draft, Writing – review & editing.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.The authors extend their appreciation to King Saud University, Saudi Arabia for funding this work through researchers supporting project number (RSP-2021/133), King Saud University, Riyadh, Saudi Arabia Supplementary material related to this article can be found online at https://doi.org/10.1016/j.egyr.2021.11.139.The following is the Supplementary material related to this article. MMC S1 The Supplementry materials containing the fatty acid composition of the used oil, the kinetic parameters, and fitting of the kinetic rate constnat with Arrhenius equation.

Raw coal without carbonization was treated with sulfuric acid at different concentrations (70% to 95%) as a controlled sulfonation process to produce a simple and effective acidic catalyst. The synthetic catalyst (S.Coal) at the best H 2 SO4 concentration (95%) showed the best acid density (8.4 mmol/g), the best surface area (26.4 m2/g), and the best catalytic activity during the transesterification of the waste sunflower oil. The best yield (98.8%) at low-temperature conditions (40 °C) was achieved after 60 min using 3 g of the catalyst and at a methanol/oil ratio of 20:1. At high-temperature conditions (120 °C), the yield was enhanced to 99.5% after 30 min only considering the catalyst quantity and the methanol content as the same values. The S.Coal as a catalyst is of remarkable reusability for five transesterification runs either at low-temperature conditions (40 °C) or at high-temperature conditions (120 °C). Additionally, the catalyst is of higher catalytic performances than several basic and acidic catalysts demonstrate the efficiency of the sulfonation process on the raw coal without carbonization. The kinetic properties of the occurred transesterification reactions over S.coal followed the Pseudo-First order behavior and of low activation energy.

Due to the depletion of fossil fuel reserves, converting renewable and abundant biomass into fuels and chemicals becomes important [1–4]. Lignin is one of the main components of lignocellulose, mainly consisting of three aromatic units, i.e., p-coumaryl alcohol (H), coniferyl alcohol (G) and sinapyl alcohol (S), which are connected with C-O and C-C linkages. As the only renewable natural resource that contains aromatic rings, lignin is regarded as a suitable feedstock to replace fossil resources to produce aromatic chemicals [5–8] and fuels [9–11].Enzymatic hydrolysis lignin (EHL) is the by-product of bioethanol production industry, and has not been utilized effctively [12–14]. Currently, EHL as a low-grade fuel is burned to produce heat and power for the biorefining industry [15]. Compared to the kraft lignin, EHL has a high purity and low content of sulfur, and its structure is similar to the native lignin [16]. Therefore, EHL is a suitable feedstock for depolymerization reaction to produce aromatic chemicals [17].Catalytic solvolysis of lignin into aromatic chemicals has been widely investigated, some milestone works were reported. Barta et al. [18] found that supercritical methanol served both as the solvent and the hydrogen-source in the conversion of organosolv lignin into substituted cyclohexyl derivatives over a Cu-based porous metal oxide catalyst. Song et al. [19] depolymerized lignin in birch wood with Ni/C as a catalyst in methanol at 200 °C under Ar atmosphere, and obtained 54% phenolic monomers with propylguaiacol and propylsyringol as the main products. Our previous work verified that Mo-based catalysts were active for catalytic solovolysis of Kraft lignin. Ma et al. [20,21] achieved the complete conversion of Kraft lignin into small-molecule products, including C6–C10 esters, alcohols, arenes, phenols, and benzyl alcohols, with MoC1−x/AC as a catalyst ethanol at 280 °C. After that they reported that MoN2, Mo/Al2O3 and MoC1−x/CuMgAlOx showed a similar activity for the Kraft lignin depolymerization and similar products. However, the complex products are usually obtained from lignin depolymerization, posing a challenge to the subsequent separation and purification process [22]. Recently, Wu et al. [23] reported that MoS2 showed a high selectivity for conversion of guaiacol to 2-(tert-butyl)-3-methylphenol (TBC), which has been used as an antioxidant in the polymer industry. Herein, MoS2 is employed in the depolymerization of EHL. The influences of reaction parameter are examined, and the mechanism of EHL depolymerization are discussed based on the results of FT-IR and monomers and dimers conversion. Furthermore, the active species and the deactivation of catalyst are also studied.EHL was purchased from Shandong Longlive Biotechnology Co., Ltd. The raw materials are dried at 60 °C for 12 h before use. Analytical-grade chemicals and solvents, including ethanol, methanol, and isopropanol, were purchased from Tianjin Guangfu Technology Development Co., Ltd. Ammonium molybdate, sulfur, hydrazine hydrate and the model compounds were purchased from Aladdin Co., Ltd.MoS2 sample was prepared with a hydrothermal method. Ammonium molybdate (1.53 g) and elemental sulfur (0.5 g) were dissolved in 60 mL distilled water and 8 mL of hydrazine hydrate was added drop by drop to the solution. This solution was then transferred into a 200 mL Teflon–lined stainless autoclave and heated at 150 °C for 24 h. The resulting black precipitate was separated and washed with water and absolute ethanol and dried under vacuum at 60 °C for 12 h.The X-ray diffraction (XRD) patterns of fresh catalysts were recorded at room temperature using a Rigaku D/max 2500 v/pc instrument with Cu Kα radiation, operated at 40 kV and 40 mA at a scanning rate of 10 °/min in the 2θ range of 10 – 90°. The morphology and structure of samples were observed with a scanning electron microscope (SEM, S-4800, Hitachi) and a transmission electron microscope (TEM, JEM-2100, JEOL). The X-ray photoelectron spectra (XPS) for both fresh and used catalysts were recorded with a PHI 1600 ESCA system spectrometer. The X-ray source was Mg Kα (1253.6 eV), and the binding energy was calibrated using C1s at 284.6 eV as the standard. The Raman spectra were obtained on a Renishaw instrument (532 nm).EHL depolymerization reactions were carried out in a 300 mL autoclave reactor (Kemi Co. Ltd, 250 mL, made of Hastelloy). In a typical run, 1.0 g EHL or a model compound, such as guaiacol, 70 mL methanol and 0.5 g MoS2 were charged into the autoclave reactor. The reactor was sealed and purged with pure nitrogen for 5 times, then heated to the desired temperature and pressurized with hydrogen and stirred at 600 rpm. After completion of the reaction, the reactor was rapidly cooled to room temperature, and the liquid products and the catalyst were subsequently separated by filtration.In the reusability tests, the used MoS2 catalyst was washed with 10 mL methanol and then was dried at 60 °C in vacuum for 1 h before the next run.All liquid products were analyzed and quantified with an Agilent 6890/5973 N GC–MS and a 6890 N gas chromatography equipped with a flame ionization detector (FID) and a 30 m HP–5MS capillary column. The injection temperature of GC and GC-MS was maintained at 280 °C. The oven temperature increased from an initial temperature of 45 °C to a final temperature of 250 °C at a rate of 10 °C/min and kept at 250 °C for 7 min. The selectivity of TBC was calculated as follows: Selectivity of TBC % = weight of the TBC weiht of the overall product × 100 % FT-IR spectra of EHL and liquid products were collected in the transmission mode on a Nexus spectrometer (Thermo Nicolet Co.). The spectrum was obtained after 32 scans and recorded in the region 4000–400 cm−1 with a resolution of 4 cm−1.The XRD patterns of the fresh and used MoS2 are depicted in Fig. 1. In the pattern of fresh and used MoS2, the peaks of (100) and (105) planes of MoS2 are weak, suggesting that the synthesized MoS2 has amorphous structure. Fig. 2(a–c) displays the SEM images of the fresh and used MoS2. Fresh MoS2 has a fluffy flower-like structure, while the used MoS2 became agglomerated. The TEM image (Fig. 2(d)) confirms the MoS2 phase in the fresh MoS2 catalyst. The lattice spacing marked on the micrograph for the highlighted domain is 0.27 nm and corresponds to the (103) plane of MoS2.The XPS spectra were also measured for the fresh and used MoS2. As shown in Fig. 3(a), Three oxidation states including Mo4+, Mo5+and Mo6+ were detected in the fresh MoS2 samples. Two peaks located at 228 and 232.2 eV were attributed to Mo3d5/2 and Mo3d3/2 of Mo4+ species. The two peaks at 229.6 and 233 eV were attributed to Mo5+, and the other two peaks located at 232.8 and 235.9 eV were assigned to Mo6+ [24,25]. Besides, the signal of S 2 s was observed at 227 eV. Fig. 5(b) depicts the XPS spectra of S 2p energy region of the MoS2 catalysts. Four peaks located at 161.9, 161.2, 163.3 and 164.3 eV were attributed to S2− 2p3/2, S2− 2p1/2, S 2 2 − 2p3/2 and S 2 2 − 2p1/2, respectively [26]. the Mo5+ (229.6 and 233 eV) and S 2 2 − (~164–165 eV) species was assigned to the MoOxSy phase [27,28]. After reaction, the Mo5+ was gradually transformed to either Mo6+ or Mo4+ partially, and its surface ratio decreased sharply from 41.6% to 2.2%. Meanwhile, the proportion of S 2 2 − decreased from 27% to 23.9% and the ratio of S2− increased from 73% to 76.1%, respectively. Fig. 4 shows the Raman spectra of the fresh and the used MoS2 catalysts. Typically, the peaks at 308.2 and 403.5 cm−1 of the fresh MoS2 are ascribed to the in-plane E 2 g 1 mode and out-of-plane A1 g mode of the MoS2, respectively. In Raman spectra of the used MoS2, two peaks located at 1383.7 and 1581.6 cm−1 are observed, which are attributed to the D (defected) and the G (graphitized) bands of carbon, [29] confirming the formation of char. Fig. 5(a) shows the total ion chromatogram (TIC) of the liquid products obtained from EHL depolymerization over a MoS2 sample at 280 °C for 6 h in methanol. 23-Types of aromatic monomers are identified, with the total yield of 124.1 mg/g, and TBC is the main products with the selectivity of 40.3%. Without MoS2 (Fig. 5(b)), only 62.6 mg/g of monomers are formed, and the main products are eugenol, guaiacol, 4-ethylguaiacol, phenol and 4-ethylphenol.As shown in Table 1, The effect of different solvents, including methanol, ethanol, and isopropanol, were examined in EHL depolymerization on MoS2 sample. Among the solvents examined, ethanol gives the highest monomer yield, but monomers obtained in ethanol are quite complex. 2,4,6-tri-tert-butylphenol is main monomers formed in ethanol, and its selectivity is only 15.4%. In isopropanol, total yield is only 107.5 mg/g, the maximum selectivity of the monomers (2,5-diisopropylphenol) is only 10%.The effect of reaction time is shown in Fig. 6(a). The total monomer yield obtained at 2 h is only 75.1 mg/g. It rapidly increases from 2 to 3 h, and generally increases from 2 to 8 h, reaching 129.8 mg/g at 8 h. The selectively of TBC obviously increases from 2 to 3 h, and generally increases during 3–6 h, reaching 40.3%, but it is not obviously changed during 6–8 h. Moreover, when the reaction time is 1 h, a few G-type products (guaiacol, 2-methoxy-4-methylphenol, 4-ethyl-2-methoxyphenol etc.) appeared. When the reaction is prolonged to 3 h, these G-type products disappeared and large amounts of alkylphenolsare formed.The effect of reaction temperature is presented in Fig. 6(b). As reaction temperatures increases from 260 °C to 290 °C, the total monomer yield increases from 89.3 to 154.1 mg/g. However, the selectivity of TBC increases from 16.6% at 260 °C to 40.3% at 280 °C and then decreases to 36% at 290 °C. Furthermore, the total yields at 260 °C and 270 °C are similar, which are 89.3 and 92.7 mg/g EHL, respectively, but the selectivity of TBC at 270 °C is 35.2%, much higher than that at 260 °C (16.6%). Fig. 6(c) shows the effect of initial hydrogen pressure at 280 °C. The total yield has a positive correlation with the hydrogen pressure, while the selectivity of TBC shows a volcanic relationship with hydrogen pressure. Char is formed when initial hydrogen pressure is 0 and 1 MPa H2, and no char is formed at 2 and 3 MPa H2. From 0–3 MPa H2, the total monomer yield dramatically increases from 69.4 to 139.8 mg/g. The selectivity of TBC reaches a maximum of 42.4% at 1 MPa of hydrogen and decreases to 38.2% at 0 MPa H2 and 29.1% at 3 MPa H2, respectively.The FT-IR spectra of EHL and the liquid products obtained from without catalyst and over MoS2 are presented in Fig. 7, and the corresponding band assignments are summarized in Table 2 [30–32]. Band 1 (3417 cm −1) is ascribed to the O-H stretching vibration. Bands 2 (2848–2960 cm−1) corresponds to C-H stretching of CH3 and CH2. Band 3 (1456 cm−1) is ascribed to C-H deformations asymmetric in CH3 and CH2 and band 4 (1380 cm−1) is ascribed to aliphatic C-H stretch in CH3. Band 5 (1334 cm−1)，band 6 (1265 cm−1), band 8 (1120 cm−1) and band 9 (839 cm−1) are related to syringyl ring and guaiacyl ring. Band 7 (1166 cm−1) is ascribed to C-O stretch in ester group.Compared to the spectrum of EHL. the band of O-H (band 1) is obviously strengthened after reaction without MoS2, but this band is significantly weakened when MoS2 is added. Nevertheless, the bands of C-H in CH3 and CH2 (bands 2, 3 and 4) become stronger after the reaction without MoS2 and are further strengthened after the reaction with MoS2. The bands related to syringyl ring and guaiacyl ring (band 5, 6, 8 and 9) are also weakened after reaction without MoS2 and nearly disappear when MoS2 is added. In addition, the band of C-O stretch in ester group (band 7) shows the same trend as the band of syringyl ring and guaiacyl ring.The MoS2 catalyzed conversion of several lignin monomers were examined under the same conditions in the lignin depolymerization ( Table 3). Guaiacol, 4-methylguaiacol and catechol are completely converted, and TBC is also the main products. Moreover, product formed in conversion of guaiacol, 4-methylguaiacol and catechol is nearly the same as that formed in EHL depolymerization ( Fig. 8). Nevertheless, only 64.0% of phthalic ether is converted with fresh MoS2, (entry 6 in Table 3), while its product distribution is nearly the same as with EHL. When 4-Ethylphenol is the feedstock, no TBC was observed. Compared to fresh catalyst, the used catalyst shows similar activity for conversion of guaiacol into TBC.A series of model dimers were used to examine the activity of MoS2 for cleavage of different C-C and C-O bonds. The results are shown in Fig. 9. 43.3% of 4,4′-oxydiphenol is converted to alkylphenols with the cleavage of 4-O-5 linkage, but 56.7% of 4,4′-oxydiphenol is only alkylated without the cleavage of 4-O-5 linkage. In the conversion of 4-(benzyloxy)phenol, α-O-4 linkage is completely cleaved, yielding 50.5% toluene and 49.5% alkylated terephthalate. The conversion of 4,4′-methylenediphenol yields 65.8% of alkylphenols, verifying that MoS2 has a activity for cleaving C-C bonds.Without a catalyst, 62.6 mg/g monomers are produced, and the band of O-H stretching vibration is strengthened, indicating that EHL is partly depolymerizated even without a catalyst. It has been proved that alcohols such as methanol, ethanol, and isopropanol acted as the nucleophilic reagent and cleave ether linkages, depolymerizing lignin into fragments and monomers [6]. MoS2 has a high activity for cleavage of α-O-4 bond, as α-O-4 bonds in 4-(benzyloxy) phenol are completely cleaved with MoS2. Among all C–O linkages in lignin, the 4-O-5 bond between two phenyl groups has the highest dissociation energy [33]. MoS2 catalyst partly cleaves 4-O-5 bond in 4,4′-oxydiphenol without aromatic ring hydrogenation. EHL contains a high amount of inter-unit C-C bonds, including both native C-C linkages and new C-C linkages formed via condensation of reactive intermediates during delignification [34]. The formation of alkylphenols (65.8%) from 4,4′-methylenediphenol verifies that MoS2 enables the cleavage of C-C bond.After reaction without catalyst, the bands of C-H stretching of methyl and methylene groups are strengthened, compared to those in EHL, indicating that alkylation occurs even without a catalyst. After reaction with MoS2, The bands related to CH3 and CH2 are all significantly strengthened, indicating that the addition of MoS2 significantly promotes alkylation reaction. MoS2 also shows a high activity for demethoxylation/demethylation reaction, as the bands related to syringyl and guaiacyl ring also disappear. Song et al. [35,36] also demonstrated that Mo has the ability to dehydroxylate. In addition, the weakness of the band of O-H stretching vibration in EHL depolymeriation with MoS2 and the formation of alkylphenol in conversion of monomers indicate that MoS2 has the activity for dehydroxylation.In EHL depolymerization, monomers with methoxy are detected at 1 h, but disappear at 3 h, and alkylphenols appear at 3 h. Moreover, the products obtained from EHL depolymerization is nearly the same as with those obtained from guaiacol and methyl-guaiacol conversions, but is different with those obtained from 4-ethylphenol conversion. Meanwhile phenol is not observed in guaiacol conversion. Therefore, guaiacol and its derivant are the intermediates for alkylphenols, instead of phenols. The products obtained in catechol conversion is nearly the same as those obtained from guaiacol conversion and EHL depolymerization, indicating that guaiacols and its derivant first undergo demethylation and then undergo alkylation. Previous work [23] indicated that ortho-methylcatechol is the main intermediates for TBC. Phenolic hydroxyl group next to the methyl group on ortho-methylcatechol may be dehydroxylated firstly, and then alkylation reaction. The electron donating effect of OH and CH3 group in ortho-methylcatechol are main reason for the alkylation activation in the ortho-hydroxyl structure. Cui et al. [37] reported that higher alkylphenols (like tert-butylphenol) more likely to form via consecutive substitution of lower alkylphenols (like o-cresol) with methyl, ethyl or isopropyl groups supplied by solvent medium. The possible pathway of TBC formed from fragmentated lignin are proposed as in Scheme 1.The reaction pathways are proposed as in Scheme 2. EHL is firstly depolymerizated with methanol, forming lignin fragments and monomers, such as G-type monomers (guaiacol and 4-ethyl-2-methoxyphenol), S-type monomers (Syringol) and H-type monomers (phenol and 4-ethylphenol). Without catalyst, active monomers and intermediates are prone to undergo repolymerization reaction, forming a large amount of coke. MoS2 depolymerizated lignin fragments through cleavage of C-O and C-C linkages, and also stabilize active phenolic monomers through dehydroxylation, demethylation and alkylation reaction.MoOxSy species is identified to exist in the MoS2 according to the XPS analysis. In our previous works with MoS2 as catalyst [23], we confirmed MoOxSy as the active phase for the alkylation of guaiacol. After EHL depolymerization, the proportion of Mo5+ and S 2 2 − decreased from 41.6% and 27–2.2% and 23.9%, indicating that MoOxSy is converted to MoS2. In addition, the catalyst surface aggregates, and the wrinkles on the surface disappear after the reaction (Fig. 2(a,b)). Moreover, graphitic carbon was formed on the used MoS2 catalyst according to the Raman results. In summary, the loss of active phase (MoOxSy), the aggregation of catalyst surface and the formation of graphitic carbon are probably the reasons for the deactivation of MoS2.In summary, A simple one-pot method for preparing 2-(tert-butyl)-3-methylphenol (TBC) with high selectivity from EHL is reported. The highest TBC selectivity of 40.3 wt% and the total monomer yield of 124.1 mg/g lignin are achieved in methanol over MoS2 at 280 °C for 6 h under 2 MPa hydrogen pressure. The selectivity of TBC shows a volcanic relationship with hydrogen pressure and reaction temperature. The result of FT-IR and the conversion of monomers indicate that MoS2 has the high activity for demethylation, dehydroxylation and alkylation. According to the results of dimer conversions, the C-C and C-O linkages in EHL are cleaved on a MoS2 sample. The effect of time on product distribution and monomer conversion proves the pathways for alkylphenol production from EHL. The main active species for selective conversion of EHL to TBC is likely MoOxSy composed of Mo5+ and S 2 2 − . Yiming Ma: Methodology, Investigation and Writing – original draft. Hong Chen: Supervision, Conceptualization, Writing – review & editing. Yongdan Li: Supervision, Conceptualization, Writing – review & editing. Kai Wu: Software, Visualization. Qingfeng Liu: Software, Visualization. Yushuai Sang: Formal analysis, Software and Validation.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This work has received funding from the European Union’s Horizon 2020 Research and Innovation program, (BUILDING A LOW-CARBON, CLIMATE RESILIENT FUTURE: SECURE, CLEAN AND EFFICIENT ENERGY) under Grant Agreement No 101006744. The content presented in this document represents the views of the authors, and the European Commission has no liability in respect of the content.

Low selectivity and complex product distribution are the main challenges for the utilization of lignin. Herein, the selective production of 2-(tert-butyl)-3-methylphenol (TBC), an antioxidant in the polymer industry, from depolymerization of enzymatic hydrolysis lignin (EHL) on a hydrothermally synthesized MoS2 catalyst is studied. The total aromatic monomer yield is 124.1 mg/g EHL and the selectivity of TBC is up to 40.3 wt% in methanol at 280 °C under 2 MPa H2 for 6 h. The FT-IR analysis of products reveals that MoS2 has a high activity for demethylation, dehydroxylation and alkylation, and the dimer conversions reveal that C-O and C-C bonds in EHL are broken with MoS2. The guaiacol and its derivants are identified as the intermediate for formation of TBC in EHL depolymerization according to the effect of time on product distribution and monomer conversion.

No data was used for the research described in the article.Montmorillonite with chemical formula (Na,Ca)0.3(Al,Mg)2Si4O10(OH)2n(H2O) is a very soft phyllosilicate group of minerals was first described in 1847 for an occurrence in Montmorillon in the department of Vienne, France, more than 50 years before the discovery of bentonite in the US [1]. Its structure is well known nearly 7 decades [2]. Several classification systems were proposed at the same time but analysis of smectites is still difficult. MMT as a member of the smectite group, is a 2:1 clay, with a central octahedral sheet of alumina sandwiched by two tetrahedral sheets of silica [3]. Dehydroxylation of clay disregarded in classifications until now, so the structure of the octahedral sheet in smectites remained unconsidered. Although, the cis- and trans-vacant character of the dioctahedral smectites had been known for a long time but a manageable proof was lacking. Determination of the octahedral structure of the sheet for illites is possible by X-ray diffraction but not for smectites because of their turbostratic disorder. Octahedral sheet structure affected on dehydroxylation temperature of all dioctahedral 2:1 clay minerals. Dehydroxylation temperature is 550 and 700 °C for trans-vacant minerals and cis-vacant varieties, respectively. Mixed types with two dehydroxylation peaks also exist. Thus, simultaneous thermal analysis (STA) can be used for determination of the cis- or trans-vacant character of MMTs [4] (Table 1 ).In 1961, Grim and Kulbicki were classified MMTs based on phase transformations and recrystallization products of the H+-exchanged at high temperatures. Based on thermal behavior, they defined two Wyoming and Cheto-type differing primarily in the distribution of the calculated layer charge. Grim and Kulbicki neglected the dehydroxylation although considered layer charge and octahedral cation population and distribution. A low layer charge and low contents of Mg2+ substituting for Al3+, characterize the Wyoming-type in contrast to the Cheto-type, which has high a content of Mg2+ and a higher layer charge. There are also mixtures of Cheto- and Wyoming-type. These two types differ in phase transformations above 1000 °C. The Wyoming-type transforms into cristobalite and mullite whereas the Cheto-type tranformed to β-quartz, β-cristobalite and cordierite at high temperatures [5].In 1969, a classification developed by Schultz based on the amount and location of charge and the proportion of tetrahedral charge [6]. Dehydroxylation temperature and the amount of hydroxyl groups also measured but only one temperature peak was recorded, even if there were two peaks. Schultz defined seven types of MMTs and beidellites: 1) ideal MMTs Wyoming-type, 2) Chambers-type (which corresponds to the mixture of Cheto-and Wyoming-type of Grim and Kulbicki), 3) Tatilla-type, 4) Otaytype (which corresponds to the Cheto-type of Grim and Kulbicki), 5) ideal beidellite, 6) non-ideal beidellite and 7) non-ideal MMT. Ideal and non-ideal types dehydroxylate at about 700 and 550 °C, respectively. Wyoming-types display a low layer charge and only beidellite has a dominant tetrahedral charge. Differentiate between MMT and beidellite have been determined via using Greene-Kelly test [7].The ranges of composition for the different types gave by Brigatti [8] and Poppi and Brigatti [9] based on the Schultz' system [6]. Types of the dioctahedral MMT series were characterized based on crystallochemical data especially octahedral and tetrahedral distributions. Their classification turned special interest to the content of iron in the octahedral layer and 8 solid solution ranges were classified for smectites: 1) Wyoming, 2) Tatilla, 3) Otay, and 4) Chambers-type, 5) non-ideal MMT, 6) nontronite, 7) beidellite and 8) Fe-rich beidellite. Fe-rich MMT and beidellite correspond to non-ideal MMT and beidellite. The iron content in the octahedral sheet of MMT and beidellite is less than 15 % of the cations in the octahedral sheet and for non-ideal or Fe-rich MMTs and beidellites 15–30 %.Classification system was modified by new methods. In 1971, Lagaly and Weiss gave a new insight into the cation density and charge distribution of layer silicates through intercalation with alkylammonium [10]. Structural formula calculations should be performed according to Köster which means the measured layer charge has to be involved in the calculation of the composition [11]. Tsipursky [12], Drits and Muller [4] and Drits et al. [13] explained that the thermal behavior of dioctahedral 2:1 clay minerals is depended to the structure of the octahedral sheet, directly. These two aspects are incorporated in the new classification system.There are samples in common classification cannot be classified as any type proposed in the literature. Samples originating from other places than Wyoming, Otay, Tatilla, etc. are difficult to characterize. Although demand is increasing for industrial applications but the names MMT/beidellite or even their trivial names Wyoming-type, etc. don’t bear information of the minerals characteristics. Even in the smectite group, MMTs show distinct differences in chemistry, octahedral sheet structure, Fe-content, layer charge and location of charge. To describe these differences well defined adjectives are used. The adjective, that gives information on the chemistry of the mineral and is not considered to be part of the name [14,15]. It may precise the name and is not connected to it which makes variations possible [16]. It should be avoided to use the adjectives as hyphenated chemical prefix.In oil drilling industry, MMT used as a component of drilling mud, making the mud slurry viscous. It is kept the drill cool and removed the drilled solids [17–21]. As a component of foundry sand and as a desiccant, it is also removed moisture from air and gases [18]. In drought-prone soils, the clay also used as a soil additive to hold soil water. It is used in the construction of earthen dams and levees to prevent the leakage of fluids [22–26]. Swelling property of this clay makes MMT-bentonite be useful also as a protective liner for landfills and as an annular seal or plug for water wells [27]. Due to its adsorbent and clumping properties, Na-MMT is also used as the base of some cat litter products [28,29]. MMT has also been used in cosmetics [27,30,31]. In a fine powder form, MMT can also be used as a coagulant in ponds [32]. As it added into water, making the water “clouded”, attracts minute particles and then settles to the bottom. MMT is an effective absorbance for heavy metals but to date, its effect on human health is not known [33]. It's assumed that heavy metal adsorption is only applicable when the clay has direct contact to it. Hence, it will not help when ingested because almost doesn't pass through the intestinal mucous membranes, certainly. MMT has been used to treat contact dermatitis for external use [34]. Because the clay may provide some resistance to environmental toxins, it is added as an anti-caking agent to some animal's foods [35]. MMT clays have been extensively used in catalytic processes [36]. For over 60 years, MMT clays have been used as cracking catalysts [37–42]. Other acid-based catalysts use acid-treated MMT clays [43]. Other uses include use in papermaking to minimize deposit formation [44,45] and as a retention and drainage aid component [46].As mentioned before, MMT is a phyllosilicate mineral with nanolayered structure consists of stacked layers [47]. Thickness of layers is about 1 nm. Each layer is composed of one O-Al(Mg)-O octahedral sheet (about 100 nm × 100 nm, in width and length) sandwiched by two O-Si-O tetrahedral sheets [48]. The layer is positively charged due to the isomorphous substitution, so cations are existed in the interlayered space of MMT. Van der Waals and electrostatic forces held neighboring layers together to form the primary particles of clay [49]. Secondary micrometer-scale to millimeter-scale particles are formed through aggregation of primary particles (Fig. 1 ) [47].IR spectra of MMT recorded by Danková et al. presented in Fig. 2 [51]. As can be seen in this figure, an absorption band exist about 3626 cm−1 attributed to the stretching vibrations of structural OH groups in MMT [52]. The bands observed at 916 and 840 cm−1 related to the Al-Al-OH and Al-Mg-OH bending vibrations, respectively [53,54]. A complex band at 1040 cm−1 corresponds to the stretching vibrations of Si–O groups [53,54], whereas the Al-O-Si and Si-O-Si bending vibrations recorded at 523 and 470 cm−1 [55]. The band at 625 cm−1 are assigned to the out of plane vibrations of Al-O and Si-O [56]. A broad band at the range of 3420–3450 cm−1 correspond to the H2O-stretching vibrations. The shoulder at about 3330 cm−1 is an overtone of the bending vibration of water at 1635 cm−1 [57].SEM micrograph of MMT is shown a dense aggregate formed through condensation of the sheet structure-leaf-like crystals (Fig. 3 ) [51]. The layered structure of MMT is clear in this micrograph. The surface of clay hasn't homogenous dispersion. In addition, there are pores with different sizes distributed, randomly [58].In TEM image of natural sample of MMT from the Tagansoye deposit, reported by Krupskaya et al. in 2017, there is a significant amount of small and thin nano-sized particles among the laminar MMT particles with a size of 1–2 µm, covered the specimen and produced grey background in micrographs (Fig. 4 a) [59]. As can be seen in Fig. 4b which recorded by Alamri et al. in 2021, the clay has a porous-like surface and a nest-like form [60].Nitrogen adsorption–desorption isotherms and Barrett-Joyner-Halender (BJH) pore size distribution of MMT, obtained by Alamri et al. in 2021, are shown in Fig. 5 [60]. Surface area, pore volume, and particle size of MMT are 258.108 m2/g, 0.423 cm3/g, 8.092 nm, respectively.Alamri et al. were also prepared XPS spectrum of MMT, are shown in Fig. 6 [60]. The spectrum indicates that Mg, O, C, Ca, Si, and Al existed on the surface of MMT.In 2018, M. Ahmadzadeh et al. prepared the EDX spectrum of MMT clay (Fig. 7 ) [61]. In this spectrum, three sharp peaks are observed which are related to Al, Si and O elements. Several weak peaks are also observed which belong to Mg Fe, Na and K.XRD pattern of the MMT is shown in Fig. 8 and the crystallographic parameters are evaluated by measuring the (001) and (080) peaks. This pattern reported by Fil et al. in 2014 [50]. The peaks marked as MMT are indicate 2:1 swelling clay and confirm the characteristics of the MMT type clay and other peaks have been attributed to impurities corresponding to quartz. A diffraction peak of the (001) plane at 2θ = 19.733 corresponds to its basal spacing of 4.99 Å. The (080) reflection at 2θ = 68.823 also indicates that MMT has a dioctahedral structure [62,63].Fil et al. are also used from X-ray fluorescence (XRF) method to identify the major minerals and chemical compounds present in the MMT. Their results summarized in Table 2 [50].They also gave the pH profiles of clay as a function of time in a 1.5 wt% suspension at natural, acidic and basic conditions (Fig. 9 ) [50]. They were shown that when distilled water (pH 5.45) is added to MMT, pH raised to 8.15 in 45 min and to 7.7 after 75 min and then remained almost constant upon reaching to the equilibrium pH of 7.7. Increasing of pH in the first 45 minunes can be ascribed to the rapid adsorption of H+ ions in water onto the negatively charged MMT surface and as potential determining ions (pdi) in the electrical double layer (EDL) to provide electroneutrality. In addition, the H+ ions exchanged with some of the cations in the MMT lattice leading to the consumption of H+ ions.All of the above data confirm the structure of MMT clay.Hydrophilicity of MMT often causes the agglomeration of the nanoclay in the polymer matrix so it is not compatible to most of polymer matrixes [64–67]. Modification of the MMT surface, is the most important method to achieve homogenous dispersion of clay platelets in polymers. Nanoplatelets incorporated and distributed in polymeric matrixes, homogeneously. The organic cations as a modifier decreased the surface free energy of silicate layers and improve their compatibility with hydrophobic polymers [68–78]. Ion exchange is a method can be applied for MMT modification using cationic surfactants based on its cationic exchange capacity (CEC) [79–81]. For example, cationic smectites such as nontronite, laponite and MMT are the most common clays are modified through replacement of their exchangeable ions (Na+, Ca2+) with positively charged organic or biological molecules [82]. Organomodification involves amino groups which results in organic/inorganic hybrids with specific selectivity and reactivity [82,83]. Functionalization generate selectivity in catalysts via spatial constraints induction [84]. Generally, tremendous improvements in the wide range of physical and engineering properties of nanoclays have been observed, in recent years, [85–88].MMT was first discovered in 1847, in France [1]. Many researchers turned their attention to this clay due to its special properties, results small size of the particles and their unique crystal structures. These properties include cation catalytic abilities, exchange capabilities, swelling behavior, plastic behavior when wet and low permeability caused the clay be more applicable in many industries and processes [89]. Several studies have been also reported on the antibacterial properties of MMTs [90–100].To date, many reports published in the field of MMTs and many researchers prepared review papers about different aspects of these materials. We have referred to several of this reports in the following.“Polypropylene/MMT nanocomposites, synthetic routes and materials properties” have been evaluated in a review published by Manias et al. in 2001. According to this review, the nanocomposite is formed either by using functionalized polypropylenes and common organo-montmorillonites, or by using neat/ unmodified polypropylene and a semi-fluorinated organic modification for the silicates. Hybrids can be formed by solventless melt-intercalation or extrusion. The resulting polymer/inorganic structures are characterized by a coexistence of intercalated and exfoliated MMT layers. Tensile characteristics, higher heat deflection temperature, retained optical clarity, high barrier properties, better scratch resistance, and increased flame retardancy improved by small additions of these nanoscale inorganic fillers [101].In 2007, Leszczynska et al. were also reviewed polymer/MMT nanocomposites with improved thermal properties and thermal stability of MMT nanocomposites based on different polymeric matrixes, with the aim to describe the basic changes in thermal behavior of different polymeric matrixes upon addition of MMT. They also gave a brief description of analysis of volatile and condensed products of degradation and the kinetics of the process decomposition in inert and oxidative environment [102].Leszczynska et al. prepared a review about “factors influencing thermal stability and mechanisms of thermal stability improvement on polymer/MMT nanocomposites” in 2007. This work presents a detailed examination of factors influencing thermal stability, the role of chemical constitution of organic modifier, composition and structure of nanocomposites and mechanisms of improvement of thermal stability in polymer/MMT nanocomposites [103].In 2008, a review article published by Bhattacharyya and Gupta. The title was: “Adsorption of a few heavy metals on natural and modified kaolinite and montmorillonite: A review”. This article is a unique collection of vital information about the feasibility of using two important and common clay minerals, kaolinite and montmorillonite, as scavengers for removal of toxic heavy metals such as As, Cd, Cr, Co, Cu, Fe, Pb, Mn, Ni, Zn in their ionic forms from aqueous medium. The authors tried to incorporate how and why clays can be effectively used as a liner in water treatment plants [33].In 2009, Pagacz and Pielichowski provided a review article by this title: “preparation and characterization of PVC/montmorillonite nanocomposites—a review”. In this review, preparation and characterization of poly (vinyl chloride)/montmorillonite (PVC/MMT) nanocomposites are being presented. Flammability of PVC/MMT nanomaterials are also described. Finally, a future outlook is given [104].“Montmorillonite supported metal nanoparticles: an update on syntheses and applications” is the title of a review article of Varadwaj and Parida published in 2013 to cover numerous aspects of material syntheses and various fields of applications in which these materials show their significant efficacies. They concluded this review with a positive view for the future expansion of this field by the joint efforts of researchers from various scientific and industrial areas [105].An overview of the catalytic utility of MMT-K10 as solid acid, support for complex or metal nanoparticles in unimolecular and bimolecular reactions have been presented by Kumar et al. in 2014. The main part of this review is organized according to the role of clay in various organic reactions and an emphasis is given in highlighting the greenness of the processes. A fair comparison is also provided between clay catalysts with respect to other homogeneous or heterogeneous catalysts. Finally, the authors summarized their views in future trends and developments [106].Adsorbents based on MMT for contaminant removal from water are reviewed in a paper, published in 2016 by Zhu et al. The aim of this review article is to help the readers in choosing proper and developing novel clay mineral based adsorbents for target contaminants, and on the other hand can give a proper example to systematically show the various mechanisms for the uptake of contaminants on adsorbents. The mechanisms for uptake contaminants on adsorbents, various adsorbents based on MMT and uptake of contaminants by them, comparison of the adsorbents and disposal and reutilization of the spent adsorbents have been reviewed in this article [107].In 2018, a review paper published in the field of polymer nanocomposites based on silylated-montmorillonite by Bee et al. This paper focus on silylation, to coat silane coupling agents on the clay which is a type of covalent organic functionalization approach. Emphasis is placed on possible factors such as the silane configuration, reaction conditions and the nature of the solvent system, affected the degree of surface modification during silylation. In this review, the effect of impregnation of silylated-fillers in various macromolecular matrices is summarized and compiled based on recent collected literatures and their processing, morphologies, properties and future prospects are specifically detailed and discussed. The authors believe that this review provides a comprehensive overview on the effect of silylated MMT on the structure and properties of certain selected polymeric matrixes including polyamide, vinyl polymer, biodegradable based polymer, elastomeric rubber matrixes and epoxy resin [108].“Exfoliation of montmorillonite and related properties of clay/polymer nanocomposites” is the title of a review paper published in 2019 by Zhu et al. The literature survey suggests that future work should place emphases on developing green and effective exfoliation methods, deepening understanding of exfoliation mechanisms and the interfacial interactions between the inorganic MMT nanolayers and organic monomers/polymers. It is suggested that future research assembling exfoliated MMT nanolayers with functional polymeric molecules or other nano-scale building blocks to produce functional hierarchical nanomaterials with practical applications [49].In 2019, Dlaminia et al. were published a review is titled “Critical review of montmorillonite/polymer mixed-matrix filtration membranes: possibilities and challenges”. They were reviewed the articles focused on clay/polymer (CP) based mixed-matrix membranes (MMM) for water treatment. Fabrication and the structure of clay/polymer nanoparticles (CPNs), CPN membranes for water filtration, and inconvenience of layered platelets to mass transport, are the subjects explained in this paper. They conclude that it is possible to achieve significant improvements in water flux without compromising solute rejection [109].In order to increase the activity of MMT in different applications, the surface of MMT has been modified and functionalized by different methods [110–120]. In recent years, a great attention has been attracted to MMTs as the heterogeneous catalyst due to their unique properties. These solid catalysts have advantages such as high yield and simple isolation of product, simple and clean work up and reusability of the catalyst [121,122]. In this review article, we have provided an overview of using the MMTs as catalysts. We have been also evaluated the structure of some modified and functionalized MMTs and explained their catalytic application in the organic syntheses. In the following of our review research, several reports in the field of organic synthesis by non-modified MMT have been presented.Dintzner et al. generated 2,2-dimethylbenzopyran derivatives (3) via a one-pot condensation of substituted phenols (1) with phenyl bromide (2), catalyzed by MMT-K10 [123]. They reported that phenol could be directly condensed with phenyl bromide while in previously work which was originally observed by Dauben et al. in 1990, the major product was o-phenyl phenol (4) with minor amounts of p-phenyl phenol (5) and benzopyran (3) also generated under optimal conditions (Scheme 1 ) [124]. A detailed study of the intramolecular clay-catalyzed [1,3] shift reaction of 3-methyl-2- butenyl phenyl ether (6) also presented by Dintzner et al. in 2004 [125].In 1997, Li et al. used the MMT-K10 clay as a remarkable acetylation catalyst for acetylation of primary and secondary alcohols, thiols, amines and phenols with acetic anhydride in excellent yield. Mild conditions, high yield, easy separation and inexpensive and environmentally friendly catalyst are some advantages reported for these reactions [126].Bhaskar and Loganathan have developed an efficient, convenient and environment-friendly method for the acetylation of sugars (7) using the inexpensive MMT-K10 as the heterogeneous catalyst, in 1998 (Scheme 2 ) [127]. The authors believe that MMT-K10 as an inexpensive solid acid is shown to be an efficient catalyst for the per-O-acetylation of several mono-, di- and trisaccharides. The pyranose forms (8) accounted for 75–100 % of the acetylatedproducts.In 2002, Yadav et al. reported that aryl amines (9) react smoothly with cyclic enol ethers (10) on the surface of MMT-KSF under mild reaction conditions to afford the corresponding pyrano- and furano[3,2-c]- quinolines (11 and 12) in high yields with high diastereoselectivity (Scheme 3 ) [128]. The authors described the notable features of this procedure which are greater selectivity, mild reaction conditions, cleaner reaction profiles, high yields of products and ready availability of the reagents at low cost.In 2006, different bismaleimides (16) and bisphthalimides (17) were synthesized by Habibi and Marvi through the condensation reaction of maleic (13) and phthalic (14) anhydrides with different diamines (15) on MMT-K10 clay as catalyst under microwave irradiations and solvent-free conditions (Scheme 4 ) [129]. Solvent-free reaction conditions, simple experimental and product isolation procedures, easy recovery and reuse of the natural clays, cleaner reaction profiles and availability of the reagents at low cost, high yields of products and enhanced rates are the notable features of this procedure.An efficient green protocol for the preparation of amidoalkyl naphthols (22), employing a three-component one-pot condensation reaction of 2-naphthol (18), aromatic aldehydes (19), amides (20) or urea (21) in the presence of MMT-K10 clay under solvent free conditions reported by Kantevari et al. in 2007 (Scheme 5 ) [130]. Recovery and reusability of catalyst, short reaction time, excellent yields and simple workup are the advantages of this method.In 2014, Rocchi et al. developed a solvent-free, inexpensive and fast microwave-assisted method for cross aldol condensations of aromatic aldehydes (19) and ketones (23) for synthesis of aryl and heteroaryl trans-chalcones (24) under microwave irradiation and solvent-free conditions in the presence of MMT-KSF (Scheme 6 ) [131]. They explained that in comparison to their previously reported methods, this protocol constitutes a user- and environment-friendly alternative that proceeds normally in good to excellent yields [132].In 2020, Iriti et al. developed a fast, cheap, simple and environmentally sustainable method for the synthesis of 1,2-bisubstituted benzimidazoles (26) and 2-substituted benzimidazoles (27) from ortophenylnediamine (25) and aldehyde derivatives (19) catalyzed by MMT-K10 under microwave assistance (Scheme 7 ) [133]. The reactions were carried out in a short reaction time under solvent-free conditions by using an inexpensive and environmentally friendly heterogeneous catalyst. The authors shown that the reaction process is applicable in the industrial fields. They also compared this procedure to their previous work [134] and found that the proposed method does not require a previous treatment for the preparation of deep eutectic solvents (DESs) as eco-friendly and sustainable solvent and catalytic systems.In addition to the above reports, there are many other reports related to the use of modified MMTs in the organic synthesis, which shows the high potential of this clay as the catalyst. In the continuation of this review article, some of these reports are mentioned.In 2000, Pai et al. were designed the reaction of benzylation of arenes (28 and 29) in the presence of different types of Fe-K10/MMT as catalyst (Scheme 8 ). Fe3+-K10, K10-FeOO, K10-FeOA and K10-FeAA were synthesized. Each catalyst was activated at 120, 280 and 550 °C for a period of 5 h. For example, K10-FeOO was activated at 120, 280 and 550 °C to obtain K10-FeOO120, K10-FeOO280 and K10-FeOO550 catalysts, respectively. K10-FeOO120 was the best. It was found that in the reaction monobenzylated product (30) was formed as the main product (<93 %) and the three isomers of dibenzylated product (31) were produced as byproducts [135].They prepared Fe3+-K10 catalyst by the reported procedure [136]. For K10-FeOO preparation, they dissolved FeCl3 in dry acetonitrile and added MMT-K10. They stirred resulting slurry at room temperature for 5 h and then filtered and washed the clay with acetonitrile and then with benzene. If the clay washed with deionised water instead of acetonitrile and benzene, K10-FeOA and K10-FeAA are prepared depending on the reaction conditions [135].Sulfonic acid functionalized ordered nanoporous Na-MMT (SANM) was easily prepared by the reaction of Na-MMT with chlorosulfonic acid (Scheme 9 ) by Shirini et al. in 2012 [137]. The modified catalyst used for methoxymethylation reaction of alcohols (32) with formaldehyde dimethyl acetal (33, FDMA) in chloroform under reflux conditions in good to excellent yields (Scheme 10 ). Short reaction times, heterogeneous nature of reaction conditions, use of relatively small amounts of FDMA, ease of preparation, stability of the reagent, recyclability, and easy workup procedure are important features of the reported method.In 2014, a multi-functionalized catalyst has been synthesized by Varadwaj et al. through supporting 3-aminopropyltriethoxysilane (35) and mercaptopropyl trimethoxysilane (36) on the surface of K10-MMT which possesses the ability to act as either base or acid. At first, they prepared SO3H@K10-MMT by adding MPTMS into K10-MMT. Then, APTES was added to the SO3H@K10-MMT and the prepared SO3H-APTES@K10-MMT was obtained (Scheme 11 ) [138].The catalytic activity of the prepared SO3H-APTES@K10-MMT was evaluated as a heterogeneous catalyst for one-pot deacetalization–nitroaldol (Henry reaction) giving a 99.2 % product (38) from benzaldehydedimethylacetal (37), yield in just 2 h (Scheme 12 ). The authors reported that this material has also been shown outstanding capacity for the heavy metal cations adsorption and can be utilized as a potential candidate for the remediation of contaminated water. This material is also potent enough to carry out the Henry reaction without any significant loss of its activity [138].In 2017, Safari and Ahmadzadeh reported that an equimolar amounts of carbonyl compound (39 and 40), hydrazine hydrate (41), β-keto ester (42) and malononitrile (43), in the presence of zwitterionic sulfamic acid functionalized MMT nanoclay MMT-ZSA nanoclay at 90 °C under solvent free conditions formed pyrano[2,3-c]pyrazoles (44 and 45) (Scheme 13 ) in 84–95 % yields. Short reaction times, heterogeneous reaction conditions, a much mild procedure, a wide range of functional group tolerance, high reaction rates, absence of any tedious workup or purification, avoid of hazardous reagents/solvents and reusability of the catalyst are some advantages of this work. [139].For preparation of MMT-ZSA, at first, the MMT-NH2 nanoclay was prepared by silane condensation [140]. Then, for the synthesis of sulfonated MMT, chlorosulfunic acid added to MMT-NH2 (Scheme 14 ) [139].In 2017, Zarnegar et al. used NH2-MMT nanoclayas as an eco-friendly, nontoxic, inexpensive, and chemically highly stable nanocatalyst for the synthesis of azine (47) and 2-aminothiazole (50) derivatives in excellent yields at room temperature. (Scheme 15 and 16 ) [141]. Simplicity of performance, easier work-up procedure, short reaction times and high yields of the products are some advantages features that authors have been mentioned for this report.In order to preparation of NH2-MMT, the grafting of MMT with organic moieties containing amine was performed with APTES via silanization procedure (Scheme 17 ) [140,141]. NH2-MMT have been also used as nanocatalyst in a variety of chemical reactions and as a good support for heterogeneous catalytic processes, such as Ullmann coupling reaction [142], synthesis of heterocyclic compounds [143], Henry reaction [138], CS coupling reaction [140], Knoevenagel reaction [144], carbonylative sonogashira reaction [145] and synthesis of isoxazoles [146].In 2017, Pham et al. introduced an efficient and green synthesis of 4H-pyran derivatives (53) via one-pot, three-component condensation of aromatic aldehyde (19), ethyl acetoacetate (51) (or 5,5-dimethylcyclohexane-1,3-dione (52)) and malononitrile (43) under ultrasound irradiation in the presence of K2CO3 supported on acidic MMT at 50 °C in 50 % EtOH:H2O as solvent (Scheme 18 and 19 ) [147]. K2CO3 supported on the surface of acid treated MMT was changed to KHCO3 due to ions H+ presenting in the inter lamellar space react with K2CO3 during this procedure.In 2018, Ahmadzadeh et al. were synthesized isoxazole derivatives (55) in 87–96 % yields via one-pot multicomponent cyclocondensation of hydroxylamine hydrochloride (54), ethylacetoacetate (51) and benzaldehyde derivatives (19) in water under ultrasound irradiations in the presence of sn-MMT-K10 as catalyst (Scheme 20 ). This reaction is significant due to low-cost and eco-friendliness catalyst, rapid completion of the reactions, avoidance of using organic solvents, excellent yield and mild conditions [61].They prepared sn-MMT-K10 by ion exchange between SnCl2 according to the reported procedure in the literature [148].In 2019, Kancherla et al. used boric acid supported on MMT (H3BO3/MMT-K10) as an efficient reusable and eco-friendly heterogeneous nanocatalyst for the synthesis of 2,3-dihydroquinazolin-4(1H)-one derivatives (57) via condensation of anthranilamide (56) and benzaldehyde (19) (Scheme 21 ) [149]. This method is simple, requires cheaper reagents, the products are easy to separate from reaction mixture and the catalyst is reusable at least three times with negligible loss of activity. The authors believe that These catalysts can be used for many acid-catalyzed organic transformations wherein mild acidity is required.In 2021, another heterogeneous catalyst was also synthesized by Ahmadzadeh et al. named copper(II) anchored on amine-functionalized MMT (MMT-[(CH2)3-NHCHPy]-Cu(II)) [150]. In order to synthesize this catalyst, they first added MMT K10 to 3-aminopropyltrimethoxysilane (35) according to their previously reported method [146] to prepare MMT-(CH2)3-NH2. Then, MMT-[(CH2)3-NHCHPy] synthesized by adding 2-pyridine carboxaldehyde (58) and finally, MMT-[(CH2)3-NHCHPy]-Cu(II) prepared by adding copper acetate (Scheme 22 ).This modified MMT is behaved as a highly efficient catalytic system for the four-component condensation reaction of hydrazine hydrate (41), β-ketoester (42), malononitrile (43) and terephthalaldehyde (59) toward the synthesis of multisubstituted bispyrano[2,3-c]pyrazole derivatives (60). In aqueous media (H2O-EtOH) under reflux conditions (Scheme 23 ) [150]. The authors reported that the catalyst is environmentally friendly, has a simple synthesis and workup procedure and high synthesis yield, and has a short reaction time which has been recycled and reused five times for the reaction without losing its activity.We conclude this review with a brief explanation of acid treated and cation exchanged MMTs. These are high potential catalysts for organic reactions [151]. Oligomerization of alkenes [152–154], dehydration of alcohols [155–157], addition reactions [158–164], alkane isomerization and rearrangements [165,166], Friedel crafts reaction [167–172], miscellaneous (Diels alder, Suzuki and Heck) reactions [173–175], aromatic alkylation [176,177], Michael reaction [178–183] and Sakurai–Hosomi reaction [184–186] are examples of these reactions.Before the modification or functionalization of the clay structure, it is acted through the sites which have Lewis and Bronsted activity. Depending on the function, the modified clay act as an acid (Scheme 24 ) or base (Scheme 25 ) via the functions.Clays such as kaolinite [187–190], smectite [191], illite [192], chlorite [193], vermiculite [194], talc [195] and pyrophyllite [196] have been used as a catalyst in organic reactions. These clays act similar to each other. The reactions are used of the acidic nature of acid-treated or cation-exchanged clay minerals. Both Lewis and Bronsted activity are common. Free acid (in acid activated clay minerals) or dissociation of interlayer water molecules coordinated to polarizing interlayer cations caused the Bronsted activity [197]. M OH n m + + B → M O H 2 n - 1 O H m - 1 + + B H + Where B is water or an organic species in interlayer space [198].For many years, MMTs have been attracted a great attention in various fields specially as the heterogeneous catalyst. Remarkable advantages such as simple and clean work up, high efficiency, simple isolation of product and reusability of the MMTs, have been made them unique for different applications. In this paper, the structural features and various methods for modification of MMT reviewed. This review proves that is helpful for further research work on the catalytic activity of MMTs for synthesis of organic materials in different industries. Nahid Yaghmaeiyan: Data curation, Writing – original draft, Visualization, Investigation, Software, Writing – review & editing. Mahdi Mirzaei: . Reza Delghavi: .The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.The authors are grateful to University of Kashan for supporting this work by Grant No. 159189/40.

Microscopic crystals with soft phyllosilicate group of minerals which formed through precipitation of water solution, are called montmorillonite (MMT). It is concentrated and transformed by natural weathering in environment caves and left aluminosilicates which were contained in the bedrock. By the adding water, montmorillonite swells and expanded considerably more than other clays. The amount of expansion is depended on the type of exchangeable cation contained in the sample. The presence of sodium as the predominant exchangeable cation, is increased the swelling several times rather original volume. Hence, Na-MMT used as the major constituent in nonexplosive agents for splitting rock in natural stone quarries. Advantageous properties of montmorillonite made it appropriate for many applications such as use in oil drilling industry as a component of drilling mud, soil additive, component of foundry sand, desiccant to remove moisture from air and gases, catalyst and various medicinal and pharmacological applications. This review article consists the various synthetic methods for preparation of catalysts based on MMT for organic syntheses and assessing their catalytic activities.

The growing scarcity of water in the world today forces researchers to investigate more deeply various water conservation schemes and explore new water purification technologies. There is also a need to find solutions for the increasing presence of microbial and chemical pollutants in water (Brillas et al., 2009; Shannon et al., 2008). The scientific community has demonstrated the suitability of advanced oxidation processes (AOPs) for degrading contaminants present in wastewater. In AOPs, generally, the pollutants undergo mineralisation into different inorganic compounds such as salts, CO2, and water or they will be converted to readily degradable small organic molecules if treatment time is optimized (Oller et al., 2011; Comninellis et al., 2008; Pérez et al., 2006). The typical chemical feature that connects the AOPs is the formation of the hydroxyl radicals (•OH) (Malato et al., 2009). A broader definition of AOPs also includes the techniques that involve oxidants such as SO4 •- and Cl• (Sun et al., 2012; Tan et al., 2011). The major areas in which research is undertaken on the photocatalytic degradation of contaminants in water are the Fenton process (Clarizia et al., 2017; Rahim Pouran et al., 2015; Herney-Ramirez et al., 2010; Pera-Titus et al., 2004; Babuponnusami and Muthukumar, 2014) and TiO2 photocatalysis (Pelaez et al., 2012; Kumar and Devi, 2011; Sin et al., 2012; Zaleska, 2008; Mccullagh et al., 2007; Fagan et al., 2016; Ola and Maroto-Valer, 2015).The history of Fenton reaction began in 1894 when H.J.H. Fenton performed a reaction with iron ions and oxidizing agents. He observed a higher oxidative capacity of the mixture in comparison to its components (Koppenol, 1993). Even though the Fenton reaction was initially formulated for Fe(II) and H2O2 many redox-active metals such as Cu, Mn, and Ni also display Fenton-like reactions (Masarwa et al., 1988; Goldstein et al., 1993).The general mechanism of the Fenton process can be represented as follows. (1) Fe(II)+H2O2→Fe(III)+•OH+OH− k =40–80 M-1 s-1 (2) Fe(III)+H2O2→Fe(II)+HO2 •+H+ k =0·001–0·01 M-1 s-1 (3) Organic matter+•OH→degradation products Hydrogen peroxide reacts with Fe(II) to generate •OH. However, it is a challenging task to recover the Fe(II) (Eq. 2) because of the inherently slow Fe(III) to Fe(II) reduction kinetics (Wang, 2008). Studies report that a Fenton reaction proceeds through a non-radical iron(IV)-oxo (FeIVO)2+ species, but its mechanism awaits a proper experimental validation (Gonzalez-Olmos et al., 2011; Shen et al., 1992; Chen, 2019). Also, various studies illustrate the use of electrical energy (Nidheesh and Gandhimathi, 2012; Li et al., 2018; Plakas et al., 2016) and light energy (Feng et al., 2006) in speeding up the regeneration of Fe(II) in the Fenton reaction (Zhang et al., 2007; Rubio et al., 2013; Lin et al., 2014).Since sunlight can accelerate the Fenton process, it is explored as a cheap alternative. Here, [Fe(OH)]2+ is the crucial photoactive Fe(III) complex under solar light and is predominantly present at low pH of 2.8. In the presence of light irradiation, [Fe(OH)]2+ species get reduced to Fe(II), and that in turn generates further •OH and enhances the contaminant degradation. (4) [Fe(OH)]2++hv→Fe(II)+•OH The most desirable pH for the homogeneous photo-Fenton reaction is reported to be 2.8 (Clarizia et al., 2017). Even though photo-Fenton reactions are highly efficient in oxidising different contaminants, a pre and post-treatment of water is required to perform the reaction at pH 2.8. Also, when the pH of the solution is increased to neutral, that leads to iron sludge precipitation as iron hydroxides (Giannakis et al., 2016). The major pitfalls associated with the homogeneous photo-Fenton reaction is its narrow pH range and the need to remove iron sludge after the reaction; both these add up to the cost of water treatment. However, the heterogeneous photo-Fenton reaction can perform the water treatment at around neutral pH (Soon and Hameed, 2011). So advanced methods, or materials are warranted for performing the reaction at circum-neutral pH (O’Dowd and Pillai, 2020). Therefore, the development of low cost, efficient, visible-light responsive materials for performing the Fenton reaction at around neutral pH is an active area of research in AOPs (Lv et al., 2010; Hou et al., 2013). There are different techniques employed for achieving this target. In the current review, various advanced materials developed for Fenton and photo-Fenton processes are discussed. The readers are redirected to other reviews for a detailed understanding of the reactor design, usage of chelates etc (O’Dowd and Pillai, 2020; Ganiyu et al., 2018; Wang et al., 2016; Babuponnusami and Muthukumar, 2014; Clarizia et al., 2017). Illumination of semiconductors such as iron oxides, TiO2, ZnO etc. with light having energy equal or higher than the bandgap of the material leads to the formation of electrons and holes (Lee and Park, 2013). Those photo-induced electrons (excited from the valence band to the conduction band) transfer to an acceptor molecule and the molecule undergoes reduction. At the valence band, the generated hole (electron vacancy) receives an electron from a molecule which is adsorbed to the system, and that molecule gets oxidised. In the O2 atmosphere, generally O2 acts as an acceptor molecule and generates the superoxide anion (O2 • -). Also, the adsorbed hydroxyl groups (OH−) capture the holes to produce hydroxyl radicals (•OH). Similarly, many organic moieties will get oxidised to other smaller compounds. Various reactive oxygen species (ROS) have different capacity for oxidation and selectivity. The •OH and O2 • - are the two dominant reactive oxygen species involved in the Fenton reaction (Cai et al., 2016; He et al., 2016; Wang et al., 2020c). The •OH with a half-life of 10-9 s and high reduction potential (+2.80 V vs SCE, •OH/H2O; under acidic conditions) is the most reactive oxygen species involved (Feng et al., 2018; He et al., 2016). Since •OH are short-lived, it is generally produced in-situ by the illumination of UV light on H2O2 or O3 (Cho et al., 2010; Hodges et al., 2018). It is also possible to generate the H2O2 through the photo electrocatalytic mechanism, and that in-situ generated H2O2 can take part in the Fenton reaction and produce •OH. The •OH, which is the most active ROS, also has the capacity to disrupt the cell wall of microorganisms and can perform disinfection of water. One of the hindrances associated with •OH based disinfection is the scavenging of •OH by natural organic matter (NOM) present in the wastewater, which may diminish the efficiency of wastewater disinfection (Brame et al., 2014).At the initial stages of the development of photo-Fenton reaction, it was thought to be impractical to acidify the wastewater to perform the disinfection and removal of contaminants (Giannakis et al., 2016). As time progressed, researchers came up with the photo-Fenton methods of performing the disinfection of Escherichia coli (E. coli) at around neutral pH (Ruales-Lonfat et al., 2014; Rincon and Pulgarin, 2006; Spuhler et al., 2010; Rodriguez-Chueca et al., 2014). The mechanism of disinfection of microorganisms through a heterogeneous catalyst can occur through two pathways. At first, the normal semiconductor action resulting in the formation of electron-hole pairs and then to the creation of •OH. The •OH formation can also occur via the photo-Fenton action of H2O2 and Fe(II). A simplified summary of the mechanism of disinfection is given in Fig. 1. The complexity of the cell wall of microorganisms can be related to their capacity to get inactivated by photocatalytic action. As the complexity of cell wall increases, it needs either harsh conditions or longer exposure time for the complete disinfection. The resistance level of different classes of bacteria and viruses towards various disinfectants are compared in Fig. 2. Due to the unique barrier properties of the outer membrane of gram-negative bacteria, it is observed to be more tolerant to disinfectants compared to the gram-positive bacteria. In case of viruses, enveloped viruses (e.g. SARS-CoV-2) seem to be more susceptible to disinfectants than the non-enveloped viruses (Chu et al., 2019). Enveloped viruses consist of three building blocks; genetic material (DNA, RNA), protein capsid, and lipid bilayer. Non-enveloped viruses lack the outer lipid bilayer membrane (Holland Cheng et al., 1995). In general, disinfectants act on the lipid bilayer membrane of the enveloped viruses and deactivate the viruses (Chu et al., 2019). •OH generated by solar photo-Fenton processes are capable of inactivating viruses by photo-oxidation of capsid protein (Giannakis et al., 2016).Iron-based materials are usually treated as superior heterogeneous Fenton catalysts because of their low cost, negligible toxicity levels, high catalytic activity and easy methods for recovery (Pereira et al., 2012; Nidheesh, 2015; Fu et al., 2014; Garrido-Ramírez et al., 2010; Rahim Pouran et al., 2014). A heterogeneous Fenton system can generate the •OH by two methods. Either it could be the true heterogeneous catalytic mechanism or the homogeneous Fenton reaction occurring because of the leached iron from the solid catalyst (He et al., 2016). In 1998 Lin and Gurol (1998) proposed the widely accepted mechanism of heterogeneous catalytic decomposition of H2O2 by studying the reactions of H2O2 on the solid iron oxide catalyst (goethite). (5) ≡ FeIII-OH+H2O2↔(H2O2)s (6) (H2O2)s↔(≡FeII•O2H)+H2O (7) (≡ FeII•O2H)→≡ FeII+HO2 • (8) ≡FeII+H2O2→≡ FeIII-OH+•OH In the mechanism, the symbol ≡ FeIII represents the iron present on the surface. Here the interaction of H2O2 at the goethite surface (≡FeIII-OH) forms the complex (H2O2)s (Eq. 5). Then a ligand to metal charge transfer leads to the formation of a transition state complex (≡ FeII •O2H) (Eq. 6). Subsequently, the complex dissociates and forms hydroperoxyl radical (Eq. 7), and later •OH is generated in the presence of ≡FeII and H2O2. (Eq. 8). The mechanism depicts the recycling of Fe (III) and Fe (II) on the surface, so here goethite is treated as a heterogeneous catalyst.Apart from the pure heterogeneous Fenton process, the iron leached out from the solid catalyst enhances the reaction rate by homogeneous Fenton pathway (Ramirez et al., 2007; Wang et al., 2010; Hartmann et al., 2010). Zeng and Lemley (2009) reported the leaching of iron from amberlyst-15 ion-exchange resin while studying the kinetic modelling of degradation of the herbicide 4,6-dinitro-o-cresol (DNOC). Also, a faster rate of degradation of DNOC was observed during the addition of hydrochloric acid owing to the higher amounts of the leached ferrous ion at lower pH values. In another study, FeOx supported on CuFe2O4 and TiO2 was used as model systems for understanding the role of leached iron species in the heterogeneous Fenton reaction. This study pointed out that the methods such as gravimetry, X-ray fluorescence and energy dispersive X-ray analysis are not sensitive enough to account for the low metal ion leaching from the heterogeneous Fenton catalyst (Kuan et al., 2015). So, they have monitored the 4-chlorophenol (4-CP) degradation using inductively coupled plasma optical emission spectroscopy (ICP-OES) and UV–vis spectroscopy under continuous pH monitoring. Time-dependent leaching of metal ions was observed with the pH variations, and even µM/sub-ppm concentrations of dissolved metal ions were responsible for the increase in degradation rate of 4-CP in the heterogeneous Fenton system.Iron oxides are generally considered to be biodegradable, non-toxic and environmentally friendly (Nidheesh, 2015; Pouran et al., 2014; Ruales-lonfat et al., 2015a; Xu et al., 2012). Usually, the physical properties of synthesised materials are dependent on their specific surface area, particle size, morphology etc. and these properties vary greatly based on their synthesis strategies. Some of the popular methods adopted for the synthesis of iron-based materials include solvothermal procedure, hydrothermal procedure, thermal decomposition, microemulsion process and co-precipitation method (Nidheesh, 2015). Until now, sixteen pure faces of oxides, hydroxides and oxy-hydroxides are reported in the literature (Giannakis et al., 2016; Usman et al., 2018).Iron oxides have the potential to act as photo-catalysts because of their semiconducting properties. The possible semiconducting mechanism of iron oxides can be detailed as follows (Cai et al., 2016; Ruales-Lonfat et al., 2015a). (9) Iron Oxide+hν→Iron Oxide (h++e−) (10) e-(cb)+O2→O2 • − (11) h+(vb)+H2O→H++•OH (12) h+(vb)+OH−→•OH (13) H2O2+ e−→•OH + OH− (14) Fe(III)+ e−→ Fe(II) Upon light irradiation, the heterogeneous Fenton reaction gets enhanced by the production of Fe(II) from the reduction of Fe(III) to Fe(II). (15) Fe(III)+hν+OH−→Fe(II)+•OH Also on the particle surface, •OH is generated by the heterogeneous Fenton reaction between Fe(II) and H2O2 (Eq. 1). Later the •OH reacts with organic matter leading to their degradation (Eq. 3).In the iron oxide systems, the ferrous ion is part of the crystal system of oxides. This feature enhances the stability of the catalyst towards the splitting of H2O2, and thus the leaching of ferrous ions from the catalyst is reduced. Magnetite (Munoz et al., 2015; Zubir et al., 2015; Du et al., 2017; Nguyen et al., 2017; Nidheesh et al., 2014; Costa et al., 2006), ferrihydrite (Zhu et al., 2018; Xu et al., 2017; Xu et al., 2016b; Zhang et al., 2014; Zhu et al., 2018), hematite (Pradhan et al., 2013; Patra et al., 2016; Jaramillo-Paez et al., 2017; Chen et al., 2016; Huang et al., 2016), goethite (Xu et al., 2016c; Wang et al., 2017; Hou et al., 2017; Krumina et al., 2017; Jin et al., 2017; Qian et al., 2018), schwertmannite (Duan et al., 2016; Yang et al., 2016; Wang et al., 2013), lepidocrocite (He et al., 2017; Sheydaei et al., 2014), and maghemite (Wang et al., 2008; Ma et al., 2018) are some of the classes of iron minerals utilised as Fenton-catalysts. The recent developments regarding these heterogeneous Fenton catalysts are discussed in the upcoming sections.Ferrihydrite (Fh) is a naturally occurring iron oxyhydroxide mineral used as a Fenton catalyst because of its large specific surface area (Liu et al., 2010; Zhang et al., 2014). In some of the recent studies Ag/AgBr/ferrihydrite (Ag/AgBr/Fh) and Ag/AgCl/ferrihydrite (Ag/AgCl/Fh) was established as heterogeneous Fenton catalysts (Zhu et al., 2018a, 2018b). AgBr is a semiconductor which absorbs light in the visible region (bandgap of 2.6 eV). Ag nanoparticles absorb visible light because of their surface plasmon resonance (SPR) effect. Here, by introducing Ag/AgBr/Fh hybrid system, the study demonstrates the direct injection of electrons from the Ag/AgBr to the ferrihydrite (Zhu et al., 2018b). On the ferrihydrite surface, electrons help in the regeneration of Fe(II) which leads to an enhancement in catalytic degradation of bisphenol A (BPA, is an endocrine disruptor and one of the emerging contaminants of concern present in drinking water (Rubin, 2011). ). Since the electrons from catalyst are directly performing the recycling of Fe(II), the reaction is also efficient in terms of the amount of H2O2 consumed. XPS analysis was employed to determine the chemical state of different elements present in the catalyst such as Fe, O, Ag and Br (Zhu et al., 2018b). The peak at 711.72 eV in the Fe 2p spectrum was attributed to the Fe(III) coordinated to the oxygen on Fh. In Ag 3d spectra two peaks at 368.5 and 374.5 eV values represent the silver in the zero-oxidation state, and the peaks at 367.9 and 373.9 eV were associated with the Ag+ in AgBr (Zhu et al., 2018b). In a similar study, Ag/AgCl/Fh hybrid catalyst also demonstrated BPA degradation (Zhu et al., 2018a). Ag/AgCl/Fh was synthesised by an impregnation-precipitation strategy followed by a photo-reduction under UV light. The rate of degradation of BPA by six percentage Ag/AgCl/Fh (6% weight ratio of Ag added to Fh) was measured to be 0.0506 min-1 which is about five times the rate of pure Fh (k = 0.0099 min-1).Another strategy to enhance the rate of regeneration of Fe (II) is to introduce electrons from semiconductors to the heterogeneous Fenton catalyst. In another study, a BiVO4/ferrihydrite (BiVO4/Fh) system was synthesised to understand the decolourisation efficiency of acid red-18 at near-neutral pH (Xu et al., 2017). EPR spectrum showed that the introduction of BiVO4 to the ferrihydrite enhanced the generation of •OH. XPS studies and 1,10-phenanthroline spectrophotometric method concluded the increase in the concentration of Fe(II) on the surface of the BiVO4/Fh. Furthermore, enhanced H2O2 consumption was observed for BiVO4/Fh system compared to the pure ferrihydrite. This was rationalised by the Fe(II) regeneration on the surface, by the photogenerated electrons from BiVO4. Similarly, a 15% doped composite of cerium oxide (CeO2) and Fh showed 98.7% degradation of tetracycline antibiotic (Huang et al., 2020). Here, the mechanism details the critical role of Ce4+/ Ce3+ cycle in helping the regeneration of Fe(II). Furthermore, 7%TiO2/Fh nanohybrid depicted the efficient removal of cefotaxime antibiotic under UV light (Jiang et al., 2019).In a recent report, a composite material of oxidised multi walled carbon nanotubes (CNTs) and ferrihydrite (CNTs/Fh) was prepared and evaluated in the degradation of BPA (Zhu et al., 2020). A 3% CNTs/Fh system depicted seven times higher efficiency compared to simple Fh in degrading the pollutant. Cyclic voltammetry (CV) studies revealed a 14 mV lowering of half-wave potential (E1/2) of CNTs/Fh (0.827 V vs. RHE) compared to Fh (0.841 V vs. RHE) revealing the fast reduction of Fe (III) (thermodynamic aspect). Also, the effective transfer of electrons from H2O2 to Fh was considered as the dynamic aspect of the increase in the rate of Fenton reaction. The possible mechanism deduced from DFT calculations, and CV characterisation is summarised in Fig. 3.Transition metal-doped iron oxides with a spinel structure are normally named as ferrites. They have a general formula of MxFe3−xO4 (M is a bivalent transition metal ion) with a face-centred cubic lattice formed by oxide ions. Among the different iron-based materials tested as heterogeneous photo-Fenton catalysts, the ferrites are of particular interest because of their narrow bandgap (1.9 eV, for ZnFe2O4) and high stability (Sharma et al., 2015; Hou et al., 2013; Valdés-Solís et al., 2007; Wang et al., 2014). Ferrites are preferred as heterogeneous Fenton catalysts because of their easiness in recovery and reuse owing to their magnetic properties (Laurent et al., 2008; Polshettiwar et al., 2011; Sharma and Singhal, 2015). Ferrites are chemically stable (Yang et al., 2013), and because of their narrow bandgap, they are also active catalysts under visible light (Wang et al., 2011). Sharma and Singhal (2015) demonstrated the synthesis of magnetic nano-spinel having formula MFe2O4 (M=Cu, Zn, Ni and Co) using a sol-gel method. Among all the four ferrites, CuFe2O4 was found to be best (k = 0.228 min-1) for the degradation of azo dye RB5, which was attributed to the coupling between Fe3+/Fe2+ and Cu2+/Cu+ redox pairs leading to the efficient production of more •OH radicals. Similar studies also reported the use of copper ferrites for gallic acid removal (Fontecha-Cámara et al., 2016), degradation of sulfonamide antibiotic (Gao et al., 2018), and antibacterial therapy (Liu et al., 2019). Fontecha-Cámara et al. (2016) studied three commercially available iron oxides; copper ferrite, magnetite and ilmenite (FeTiO3) for the removal of gallic acid and the highest catalytic activity was displayed by copper ferrites. Cu2+ occupied the octahedral site of the copper ferrite spinel, and the collective effect of iron and copper ions significantly improved the rate of Fenton reaction by generating more •OH radicals. In another study, ZnFe2O4 was synthesised from precursors such as Fe(NO3)3 and Zn(NO3)3 through a hydrothermal treatment procedure and depicted the visible light degradation of orange II (Cai et al., 2016). Experiments performed with various radical scavengers such as tert-butanol, sodium oxalate and iso-propanol showed that •OH generated on the surface was the key species responsible for the degradation. A generalised scheme of the mechanism of generation of •OH and the regeneration of Fe(II) on the surface of ferrites is summarised in Fig. 4. The stability of the catalyst was understood by performing multiple runs with the recycled catalyst. After the first cycle, the reaction rate constant was observed to be 0.0468 min-1. Even after five cycles, the catalyst showed a similar reaction rate constant (k = 0.0483 min-1). Also, the amount of H2O2 decomposed by the catalyst during the five cycles was equivalent. Invariably all these indicate the reusability of the catalyst. Xiang et al. (2020) prepared ZnFe2O4 nanoparticles having yolk-shell structure and evaluated in the degradation of tetracycline under visible light. The yolk-shell structured ZnFe2O4 nanoparticles had a higher specific surface area and presented better visible light absorption capacity in comparison to the spherical ZnFe2O4 nanoparticles. The higher visible light absorption was correlated to the possibility of multi-scatterings of light in the inner yolk-shell structure of ZnFe2O4. Later, Hermosilla et al. (2020) reported an environmentally friendly synthesis of manganese ferrites (Mn‒Fe2O4) via routes such as sol-gel, combustion and reverse microemulsion. Bio-recalcitrant compounds such as ciprofloxacin (a fluoroquinolone antibiotic) (Davis et al., 1996) and carbamazepine (anti-depressive drug) (Ballenger and Post, 1980) were successfully degraded by the photo-Fenton action of visible light active Mn‒Fe2O4. The magnetisation is one of the crucial characteristics of a material for its separation and reusability. α-Fe2O3 (hematite) is reported as a weak ferromagnet, and its content has a trivial contribution to the magnetisation of the material (Raming et al., 2002). Here, the magnetisation value of the sol-gel synthesised Mn‒Fe2O4 came to be 41.0 emu g-1 and the lowest value was reported for Mn‒Fe2O4 synthesised by reverse microemulsion route (3.7 emu g-1). The relative content of α-Fe2O3 was lowest in the sol-gel Mn‒Fe2O4, and it was associated with its higher values of magnetisation.Some of the recent studies report the use of magnetite (Fe3O4) or its composites for disinfection of E. coli in water (Tong et al., 2020; Feng et al., 2019; Arshad et al., 2019). A composite material of Fe3O4 and flower-like MoS2 (Fe3O4/MoS2) effectively inactivated E. coli up to six log scale within 30 min (Tong et al., 2020). The Fe3O4/MoS2 was active at a broad pH from 3.5 to 9.5, and the catalyst could be separated magnetically owing to its saturation magnetization value of 40.6 emu g-1. Similarly, a graphene composite of Fe3O4 was successful in inhibiting the growth of Pseudomonas aeruginosa and S. aureus (Tong et al., 2020). Wang and co-workers developed a therapeutic approach by combining the copper ferrite antibacterial therapy with photothermal therapy (PTT) (Liu et al., 2019). The hydrothermally prepared haemoglobin functionalized copper ferrite nanoparticle (Hb-CFNPs), effectively generated the •OH and initiated the cell membrane disruption. Further shining 808 nm laser light (near-Infrared, NIR) increased the cell membrane permeability by hyperthermia and resulted in leakage of bacterial contents. In-vitro experiments revealed the broad-spectrum antibacterial activity over the E. coli (100% removal), and S. aureus (96.4% removal) bacteria and the therapeutic method showed significant results in the S. aureus infected abscess treatment. The coupling between Fe3+/Fe2+ and Cu2+/Cu+ redox pairs catalysed the production of •OH and promoted the oxidative damage of bacterial cells. A brief illustration of the synthetic strategy and the therapeutic application of Hb-CFNPs is outlined in Fig. 5.There are various reports of the application of iron oxide family of materials for wastewater treatment and microbial inactivation (Nieto-Juarez and Kohn, 2013; Pecson et al., 2012; Xu et al., 2012). In 2015, Ruales-Lonfat et al. (2015b) studied the microbial inactivation efficiency of four commercially available iron oxides. Hematite, goethite and wustite used O2 as electron acceptor and performed the photocatalytic activity even in the absence of H2O2. However, the magnetite was only active in the presence of H2O2. It is important to note that no bacterial growth was observed after the photo-Fenton treatment. These results are very significant because excluding H2O2 from the reaction decreases the cost of the water treatment to a great extent. The concentration of the catalyst used was 0.6 mg/L Fe3+, and iron concentration similar to this scale is usually observed in natural water sources. This seems to be a useful strategy in a large-scale application for bacterial inactivation.Size and morphology of the nanomaterials have a significant correlation with the physical and chemical characteristics exhibited by them (Mai et al., 2005; Kundu and Jayachandran, 2013; Xie et al., 2013). Hematite (α-Fe2O3) having morphology such as microtubes, nanorods, nanorings are reported in the literature (Xiong et al., 2011; Vayssieres et al., 2005; Hu et al., 2007). Xiao et al. (2018) studied the morphological evolution of hematite by adjusting the hydrothermal reaction time. After 6 h of hydrothermal treatment, the nanoparticles attained a spherical morphology, and upon further heating, elliptical, olive-like and burger-like morphologies were observed at 12, 18, 24 h, respectively. Burger-like α-Fe2O3 was found to be better in the removal of acid red G in comparison to other morphologies of α-Fe2O3, and a 98% degradation efficiency was observed under visible light within 90 min.Recently, a composite material of schwertmannite/graphene oxide (SCH/GO) was synthesised through an oxidation-co-precipitation route and demonstrated removal of tetracycline antibiotic under visible light. The photo-Fenton catalytic tests were performed in real wastewater matrices such as raw food wastewater and biogas fluid of anaerobically digested food. The SCH/GO nanocomposite was efficient in selectively degrading the tetracycline in the presence of a comparable concentration of moieties such as chlorides, sulfates, phosphates and nitrates. The SCH/GO system showed fifteen times higher rate constant of tetracycline degradation compared to the SCH, because of its improved optical absorption property and separation of electron-hole pairs (Ma et al., 2020).Similar to the Fenton reaction in the presence of H2O2, Fe(II/III)-oxalate system also reported having the superior capacity to degrade organic pollutants (Wei et al., 2013; Liu et al., 2012; Lan et al., 2008). A FeWO4 nanosheet material synthesised by a hydrothermal method showed facet dependent surface Fenton chemistry in the presence of oxalic acid (Li et al., 2019). Density functional theory (DFT) studies concluded that the {001} facets were efficient in producing reactive oxygen species in comparison to {010} facets. Also, DFT analysis confirmed that •OH generated on the {001} facets diffused faster to the solution and kept the {001} facets vacant for the continuous activation of oxalic acid molecules into radicals.Apart from directly employing various iron minerals as heterogeneous Fenton catalysts, iron minerals can be incorporated into numerous supporting materials like zeolites (Soon and Hameed, 2011; Nidheesh, 2015; Hartmann et al., 2010), metal-organic frameworks (MOFs) (Liu et al., 2017; Cheng et al., 2018), clays (Garrido-ramírez et al., 2010; Navalon et al., 2010), graphene oxide (GO) (Nidheesh, 2017; Wang et al., 2019), silica (Gan and Li, 2013; Zhong et al., 2011) etc. Some of the desirable properties needed for the supporting materials for holding the iron-based catalysts for photo-Fenton reaction can be their ability to perform the reaction for multiple cycles and the lesser leaching of the Fe ions. Moreover, they need to be durable against highly reactive radicals. In this section, recent reports on various supporting materials for heterogeneous Fenton processes and their desirable properties are discussed.The incorporation of iron into clays can be performed by pillaring (Guimarães et al., 2019; Tabet et al., 2006), impregnation (Herney-Ramirez et al., 2008; Hassan and Hameed, 2011) etc. Generally, inorganic supporting materials provide thermal stability, resistance to organic solvents and high mechanical strength (Cheng et al., 2006).Clay materials are abundantly present in the earth crust, but they may not be used as excellent Fenton catalysts because of their low iron content in them. Even though the layered clay materials have a large surface area, their interlamellar space is mostly inaccessible because of higher electrostatic interaction present between the layers (Garrido-ramírez et al., 2010). These shortcomings are circumvented by pillaring of clays. Among the different clay materials, pillared clays are of special interest because of their catalytic and adsorption properties. Through the pillaring process, (stacking and then connecting the 2D layers) various large-sized poly oxo-hydroxy metal cations are incorporated into the structure of clays by replacing the smaller ions (Aznarez et al., 2015; Nogueira et al., 2011). This process makes the interlamellar space accessible for the reactants and leads to a significant increase in the porosity and surface area of the material. Pillaring process also exposes some of the catalytic sites, and additional catalytic sites are added in case iron compounds make the pillars (Navalon et al., 2010). Since the crystal structures and other characteristic properties of the pure clay materials are well-defined, the difference in catalytic activity mainly arises from the pillars incorporated (Baloyi et al., 2018). Further, the Fe(III) species can be considered as immobilised in the interlayer spacing of pillared clays. So, the ion species is stable against the differences in solution pH, and that results in limited leaching of iron (Herney-Ramirez et al., 2010). There are various reports of the use of pillared clays for the degradation of dyes (Li et al., 2015; Ayari et al., 2019), pharmacologically important compounds (Hurtado et al., 2019; Khankhasaeva et al., 2017), and phenolic compounds (Hadjltaief et al., 2015; Catrinescu et al., 2012). However, the applications of pillared clay-based systems in more complex matrices, especially those with heavy organic load, are rare in literature. Recently, the photo-Fenton activity of Al-Fe smectite pillared clay has been demonstrated for the treatment of winery wastewater with high amounts of recalcitrant polyphenolic compounds (Guimaraes et al., 2019). The catalyst was prepared by intercalating poly-hydroxy aluminium (Al3(OH)4 +5) and (Fe3(OH)4 +5) species between the layers of natural smectite. The photo-Fenton studies performed under UV-C light radiation resulted in a 75.2% percentage total organic carbon (TOC) removal of the winery wastewater. In another study, Xu et al. (2016b) used hydroxy-iron montmorillonite (Fe/Mt) as a host material, and BiVO4 semiconductor was loaded into the interlayers of Fe/Mt. An 8%BiVO4/Fe/Mt composite demonstrated an 85.2% TOC removal of acid red-18 under visible light irradiation. The remarkable •OH generation capacity of the system was associated with the synergistic effect between BiVO4 and Fe/Mt and the photo-induced injection of electrons from BiVO4 to Fe(III) ions.Similar to the examples of the addition of plasmonic systems with the ferrites, Ag/AgCl was impregnated onto sepiolite clay which was modified with hydroxy iron (Ag/AgCl/Fe‐S) (Liu et al., 2017). This catalyst exhibited excellent activity in degrading BPA. Electrochemical impedance spectroscopy (EIS) was performed to understand the charge transfer resistance (CTR) and ease of separation of electron-hole pairs in the catalyst. In a three-electrode electrochemical system, 0.1 mol/L KCl was used as an electrolytic solution, and a glassy carbon electrode modified with prepared catalysts was employed as a working electrode. Several reports suggest that a lower charge transfer resistance can be correlated with facile separation of electrons and holes (Ganiyu et al., 2018). Here among the three catalytic systems studied, Ag/AgCl/Fe‐S was reported with the lover CTR, and that corroborate the enhanced degradation of the BPA.Layered double hydroxides (LDHs) are a class of clay-based materials having brucite-like sheet structures made of metal hydroxides (Gursky et al., 2006; Shao et al., 2013). Their intercalated anions/cations can be easily exchanged by cation exchange to alter their properties (Zhang et al., 2014; Zhang et al., 2012). The strong electrostatic interaction observed between the layers and interlayer anions provides a well-oriented structure for the LDHs (Zhang et al., 2020). This ordered layered structure endows the LDHs with plenty of sites for the interaction of pollutants and H2O2 (Jack et al., 2015). The general hydrophilic nature provided by the hydroxyl groups promotes distinct interaction of hydrophilic contaminants and H2O2 with the active catalytic sites on the surface (Yang et al., 2020). Bai et al. (2017) synthesised a Co/Fe LDH through a co-precipitation strategy and demonstrated the Fenton-like removal of nitrobenzene. The mechanism of the process was studied by an •OH scavenger, and •OH was identified to be the key radical involved. In another study, a Fe‒Ni LDH was reused for the synthesis of a magnetic catalyst (Ni3Fe/Fe3O4). Here, Fe‒Ni LDH was treated with the orange II dye and heated under nitrogen atmosphere to obtain the novel catalyst (Ni3Fe/Fe3O4). The thermo-magnetic curves depicted the superior magnetic properties of the synthesised catalyst. In the past decade, there have been various reports of using copper-containing LDHs for the mineralisation of phenol, a major waste generated in the petrochemical industry (Zhang et al., 2010a, 2010b). But in those scenarios, the degree of mineralization of phenol was comparatively low, and the system utilized a higher dosage of H2O2. Even though many of the studies discuss the synergistic effect of copper with other metals present, yet studies concerning the in-depth understanding of phenol degradation mechanism remains unaddressed (Zhou et al., 2011). In a similar perspective, Wang et al. (2018) prepared a series of CuNiFe LDHs by varying the Cu/Ni ratios. The specific band observed in the Raman spectra at 460 cm-1, and 533 cm-1 corresponds to the lattice vibration in LDHs. For the samples where the Cu/Ni ratio is higher than 0.5, a particular band is observed at 294 cm-1, attributed to the Cu(OH)2. The intensity of this band increases with the increase in Cu/Ni ratio. Experiments revealed that the catalytic activity (phenol mineralisation) increased upon decreasing the Cu/Ni ratio. It is remarkable to mention that, when the concentration of H2O2 was kept near the theoretical value (M Hydrogen peroxide /M phenol =14) they observed mineralisation of 90% phenol. At lower Cu/Ni ratios, electron transferred from Ni2+ to Cu2+ and facilitated the generation of Cu+ species. Here Cu+ reacted with H2O2 in a Fenton like mechanism and produced the •OH. The mechanism is summarised in Fig. 6.The large-scale applications of these materials vary depending upon the local availability of the particular clay materials. Currently, many of the Fenton related studies using clay materials concentrate on the degradation and mineralisation of dyes. Extensive research is needed to develop new catalysts with disinfection properties and the capacity for removal of antibiotics.Perovskites are a class of ABX3 compounds in which the X anion is mainly O2- (Ferri and Forni, 1998; Zhu and Thomas, 2009). Perovskite compounds have a cubic geometry with A cation surrounded by 12 X anions, and B cation surrounded by 6 X anions (Smith et al., 2019; Quan et al., 2019). These compounds received their generic nomenclature from the mineral perovskite (CaTiO3). In the last decade, ABO3 perovskite family of oxides such as EuFeO3 (Ju et al., 2011) , LaFeO3 (Nie et al., 2015), BiFeO3 (Rusevova et al., 2014; Luo et al., 2010) garnered a great deal of attention as heterogeneous photo-Fenton catalysts for the degradation of various organic pollutants. A nano-BiFeO3 perovskite catalytic-system was demonstrated to have the degradation capacity of BPA. They have studied the capping action of various organic ligands such as oxalic acid (OA), formic acid (FA), glycine (Gly), nitriloacetic acid (NTA), and ethylenediaminetetraacetic acid (EDTA) on the nano-BiFeO3 catalyst. Studies show that the EDTA_BiFeO3 system was accelerating the BPA degradation and the efficiency of OA_BiFeO3 system was lower than the bare BiFeO3 catalyst (Wang et al., 2011). To further understand the system, density functional theory (DFT) studies were carried out on the OA_BiFeO3 and EDTA_BiFeO3 models. DFT studies gave the insight that unique hydrogen bonding interaction observed in the EDTA_BiFeO3 catalyst was responsible for the weakening of the O-O bond and the generation of •OH ( Fig. 7).In a recent effort, Cu-substituted LaFeO3 perovskite was used for the degradation of BPA (Pan et al., 2020). Using a citric acid complexation method (Zhao et al., 2016), by altering the copper doping ratio they demonstrated the synthesis of novel Cu-substituted LaFeO3 catalysts. The generation of oxygen vacancy in the Cu-substituted LaFeO3 played a critical role in redistributing charge on the surface of the catalyst, and that helped the efficient decomposition of H2O2. The XRD phase evolution studies depicted that at calcination temperature of 700 ºC, the XRD spectrum became narrower and sharper. Also, the LaFeO3 perovskite structure of LaCuxFe1−xO3-δ solid solution was retained when the x values changed from 0.1 to 0.5. The peaks corresponding to Miller indices (121) and (240) got widened and the observed peak-offset was attributed to the lattice contraction of the crystal. Theoretical calculations showed an approximate 0.05-angstrom decrease in the Fe‒O, and La‒O bond lengths after copper substitution and that also portrays the volume contraction of the LaCuxFe1−xO3-δ unit cell. These parameters undoubtedly suggested that copper got incorporated into the LaFeO3 perovskite replacing Fe in the structure (Pan et al., 2020).In a similar study, Cu-doped LaFeO3 was used as a visible-light active catalyst for the photo Fenton degradation of methyl orange (To et al., 2018). The performance of 15 mol% Cu-doped catalyst was better than that of pure LaFeO3 catalyst. In another study, Cu-doped BiFeO3 was synthesised by a sol-gel method, and it depicted the degradation of 2-chlorophenol under visible light (Soltani and Lee, 2017). In the above cases of copper doping, along with Fe(II), Cu(I) was also acting as an active species in a Fenton like manner in splitting the H2O2 to •OH. In a different study, Chu et al. (2018) used the Ag-doped LaCaFeO3-δ (Ag-LaCaFeO3-δ) perovskite as a peroxymonosulfate (PMS) activating agent for the effective removal of bacterial pathogens. EIS studies have suggested an increase in lattice oxygen vacancies in Ag-LaCaFeO3-δ than LaCaFeO3-δ. A synergistic effect of free radicals (SO4 •- and •OH) and silver ions towards the bacterial inactivation was observed. The studies demonstrated the antimicrobial effect of the catalyst on the E. coli and S. aureus (a methicillin antibiotic-resistant bacteria). The study also displays that the silver leaching observed after 48 h of the reaction was 0.09 mg/L, which comes below the guidelines of the World Health Organization (WHO) for safe drinking water.Carbon-based materials namely carbon nanotubes (CNTs) (Yao et al., 2016; Yang et al., 2018), activated carbon (AC) (Yao et al., 2013; Navalon et al., 2011), biochar (Fang et al., 2015; Yan et al., 2017), graphene oxide (GO) (Nidheesh, 2017; Divyapriya and Nidheesh, 2020), g-C3N4 (Sudhaik et al., 2018; Hasija et al., 2019) etc. have been exploited in the heterogeneous Fenton reactions. The recent developments in using two-dimensional carbon-based materials for Fenton related applications are discussed here.Many researchers have shown that incorporating carbon materials with heterogeneous Fenton catalysts helps in quick reduction of Fe(III) to Fe(II), because of its fast single electron transfer ability. Graphene is a two-dimensional monolayer of carbon atoms with superior electron mobility, mechanical stability and electrical conductivity (Dai, 2013; Bekyarova et al., 2013). It is reported that the presence of graphene provides support to the Fenton catalyst, and it enhances the performance of the Fenton reaction (Nidheesh, 2017; Divyapriya and Nidheesh, 2020; Han et al., 2014). In a GO‒Fe3O4 Fenton catalyst, GO is considered as a sacrificial electron donor (Zubir et al., 2015). The unpaired π electrons present in the sp2 carbon domains (CC) of GO transfer electron to the iron centres of Fe3O4 and accelerate the reduction of Fe(III) to Fe(II) (Zubir et al., 2014). XPS analysis on the GO‒Fe3O4 system evidenced a continuous reduction of Fe(III) to Fe(II). Hence the strong electron transfer ability depicted by the graphene-related materials is a crucial factor that contributes to the enhanced catalytic efficiency of graphene-based material supported heterogeneous Fenton catalysts. The GO‒Fe3O4 catalyst showed a 97% removal of Acid Orange 7 (AO7) whereas Fe3O4 was only effective in removing 65% of AO7 under photo Fenton conditions. Similar recyclability of Fe(II) species is reported for CNTs supported FeS systems (Ma et al., 2015). The several chemical moieties present on the GO (carboxyl, hydroxyl, hydrophobic groups etc.) and the higher specific surface area promotes the adsorption of organic pollutants to the surface of GO and contributes to the effective removal of pollutants (Bagri et al., 2010; Suárez-Iglesias et al., 2017). The hydrogen bonding, π-π interaction, hydrophobic interaction and electrostatic interaction are the four possible interactions that cause the better adsorption of pollutants to the GO surface (Wang et al., 2019). Boruah et al. (2017) prepared a magnetically recoverable Fenton catalyst by decorating Fe3O4 nanoparticles on an amide-functionalized graphene sheet. The specific π-π and electrostatic interaction between the sp2 carbon system of graphene and organic pollutants assisted the mineralisation of various phenolic compounds under sunlight irradiation. The catalyst was stable up to ten cycles, and lower electron-hole recombination was inferred from the photoluminescence studies. In a similar study, Wan and Wang (2017) used polyol process and an impregnation method to prepare Fe3O4/Mn3O4/reduced graphene oxide hybrid material. Under the optimum conditions (H2O2 =6 mM, catalyst =0.5 g/L, pH = 3) catalyst showed a 98% degradation of sulfamethazine, one of the pharmaceutically active compound. Zheng et al. (2018) loaded Fe3O4 nanoparticles on the functionalized graphene oxide (GO) nanosheets through urushiol molecules as a linker (Fe3O4‒U-rGO). Urushiol is known for its strong coordinating ability to the metal oxides, and it bonds with various materials through its phenolic hydroxyl groups (Zheng et al., 2011, 2014). The composite catalytic material prevented the iron sludge formation and unfavourable decomposition of H2O2 to H2O and O2. Hence the Fenton catalytic performance was significantly improved and the process exhibited complete degradation of rhodamine B and methylene blue. The catalyst also exhibited excellent reuse stability up to seven cycles with minimal sludge formation. The synthesis method of the composite catalyst and its reaction pathway are summarised in Fig. 8. Graphene oxide membranes showed tremendous potential in water filtration technologies, but their low permeation flux was hindering their large-scale applications (Wang et al., 2016; Yin et al., 2016; Gao et al., 2013). Recently, a composite material of GO and metal-organic framework (MOF) exhibited six times higher permeation flux (26.3–30.6 L m-2 h-1 bar-1) with an enhanced separation efficiency, compared to GO nano-sheets. Also, a forty-minute visible light photo-Fenton action on the material removed 97.27% of BPA compound (Xie et al., 2020).The need to remove iron sludge after the wastewater treatment makes the homogeneous Fenton process non-viable (Zhu et al., 2019). Guo et al. (2017) used a low amount of graphene (0–2 wt%) to modify the ion-sludge obtained from homogeneous Fenton process to prepare a heterogeneous Fenton catalyst named as iron sludge-graphene (Fe‒G). The Fe‒G catalyst was characterised as FeOOH particles entrapped inside graphene sheet. Owing to the mesoporous structure and the increased adsorption ability of Fe‒G catalyst, it showed an improved degradation rate of metronidazole (an antibiotic) compared to the bare iron sludge.The development of graphene oxide (GO) composites for innovative disinfection technologies is an emerging research topic (Arshad et al., 2019; Moreira et al., 2018; Singh et al., 2020). Hu et al. (2018) prepared a hybrid material of reduced graphene oxide (rGO), silver nanoparticles (AgNP), and Bi2Fe4O9 (rGO-Ag/BFO) through an evaporation process. The hybrid material exhibited 100% bactericidal performance (> 6 logs) of E. coli. Generally, 3 logs of photocatalytic bacterial disinfection is achieved in 1–4 h; (Laxma Reddy et al., 2017) however in the process using the hybrid catalyst, approximately 6 logs of disinfection was achieved in 20 min. High performance of the material was associated with the synergistic effect of various mechanisms such as rGO/AgNP co-assisted photocatalysis, photo-Fenton reaction, and rGO assisted silver ion release. The rGO-Ag/BFO was also excellent in the disinfection of gram-negative P. aeruginosa and gram-positive S. aureus. Graphitic carbon nitride is a polymeric medium band-gap material (2.7 eV) with efficient photocatalytic property (Zhao et al., 2015; Cao et al., 2015; Wang et al., 2012). Its remarkable thermal and chemical stability have set the stage for preparation of various state-of-the-art nanocomposites for solving the energy storage and environmental pollution issues (Sudhaik et al., 2018; Hasija et al., 2019; Wu et al., 2013; Liu et al., 2016). The Fe doped g-C3N4 modified with mesoporous carbon was effective in removing Acid Red 73 dye for a wide pH ranging from 4 to 10 (Ma et al., 2017). The XPS analysis identified the Fe‒N species formed on the N-rich C3N4 as the active sites for the Fenton reaction. Cyclic voltammetry (CV) experiments verified that the mesoporous carbon could accelerate the Fe (III) to Fe (II) cycle. Similarly, Hu et al. (2019) doped the g-C3N4 with different ratios of Fe(III) and the obtained Fe‒g‒C3N4 catalyst was successful in removing phenol, BPA and 2,4-dicholorophenol. The study reported that 5% doping of Fe in g-C3N4 was the optimum iron concentration for phenol removal, and the catalyst was efficient in degrading the components of a complex wastewater system such as cooking wastewater. Fe(III) forms strong σ and π bonds with the triazine ring skeleton of the g-C3N4, and that helps the material to act as an efficient heterogeneous Fenton catalyst upon light irradiation.Many of the studies that deal with Fenton chemistry and disinfection only take care of inactivating bacteria in the system; but genetic material could stay active in the medium (Tong et al., 2020; Thakur et al., 2020). Through the horizontal gene transfer process, it is possible to transfer the antibiotic resistance gene (ARG) from one bacterium to another bacterium which does not possess the antibiotic resistance (Koonin et al., 2001; Thomas and Nielsen, 2005). Therefore, methodologies need to be developed to eliminate ARGs from the wastewater to prevent the rapid growth of antibiotic-resistant bacteria (ARB). A ternary nanocomposite system prepared by Saha et al. (2020) was successful in inactivating the commercially available plasmids pUC18 and pBR322 containing the ampicillin resistance gene (ampR). The synthesis of ternary nanocomposite of reduced graphene oxide (rGO), iron oxide and g-C3N4 was carried out by in-situ mixing of all the precursor chemicals. The fragmentation route of the plasmids was confirmed by agarose gel electrophoresis studies performed after treating the system with ternary catalyst, H2O2 and visible light. Initially, the plasmids were at a supercoiled fashion, and upon light irradiation over the system, plasmids transferred to a relaxed and single-stranded form. After an exposure time of around 30 min, the plasmids disintegrated into smaller fragments. The various phenomena that lead to the inactivation of plasmids involve photocatalytic activity by iron oxide and g-C3N4, photo-Fenton activity, relaxation of charge carriers by rGO etc. and are summarised in Fig. 9 (Adapted from reference (Saha et al., 2020)).Recent advancements show that incorporating goethite or hematite with the g-C3N4 can result in a heterogeneous photo-Fenton catalyst, capable of degrading tetracycline antibiotic under visible light (Wang et al., 2020b; Zhao et al., 2020). The combination of hematite (α-Fe2O3) and g-C3N4 resulted in a solid-state Z-scheme type catalyst (Wang et al., 2020b). In a similar perspective, a co-calcination approach of melamine (Hughes, 1941) (a cyclic compound of formula C3H6N6) with Fe-based metal-organic framework (MIL-53(Fe)), resulted in the formation of Z-scheme heterostructure catalyst α-Fe2O3 @g-C3N4 (Guo et al., 2019). The photoluminescence emission spectra suggested the enhanced separation ability of photo-generated electron-hole pairs. The higher number of electrons participated in the regeneration of Fe(II) boosted the production of •OH and resulted in higher degradation of tetracycline antibiotic. Many-a-times, the large quantity of commercial H2O2 needed in the Fenton reaction considerably increases the operating cost of the reaction and hinders its industrial-scale applications (Comninellis et al., 2008). The g-C3N4 having a negative conduction band (CB) potential compared to the reduction potential of O2/H2O2, can transfer two electrons to the oxygen and result in an in-situ production of H2O2 (Kofuji et al., 2016; Moon et al., 2017). It is significant to note that, in the α-Fe2O3/g-C3N4 system, H2O2 was produced on the g-C3N4 and the in-situ produced H2O2 was decomposed to •OH on the α-Fe2O3 surface. Also, the hole in the VB of hematite was capable of oxidising OH- to •OH. These characteristics make the α-Fe2O3/g-C3N4 a promising candidate for wastewater purification applications (Wang et al., 2020b).Zeolites are framework aluminosilicate structures composed of linked MO4 tetrahedra (M= Si4+, Al3+) (Suib, 1993; Armbruster and Gunter, 2001; Weckhuysen and Yu, 2015). An array of zeolites are available with specific pore sizes. So they find distinct applications in separating mixtures of molecules based on size and they are also called as molecular sieves (Kita et al., 1997; Jia et al., 1994; Primo and Garcia, 2014). One of the striking features associated with zeolites is their selectivity towards the guest molecules compared to other high surface area materials such as activated carbon and silica gel (Lesthaeghe et al., 2007; Martinez-Macias et al., 2015). Fe-containing zeolites are widely studied because of their structural uniformity and high catalytic activity in removing different contaminants (Aleksic et al., 2010; Gonzalez-Olmos et al., 2009; Hartmann et al., 2010; Navalon et al., 2010). Different studies show the enhancement in the catalytic degradation rate by shining UV light on Fe-zeolite systems (Kasiri et al., 2008; Kusic et al., 2006; Noorjahan et al., 2005). Gonzalez-Olmos et al. (2012) reported the mineralisation of phenol and imidacloprid (an insecticide) using iron-containing zeolites, Fe-ZSM5 and Fe-beta. Studies performed at near-neutral pH demonstrated that Fe-ZSM5 catalyst was efficient in producing •OH compared to Fe-Beta. They also performed these experiments in a pilot-scale under the solar light using a compound parabolic collector (CPC) (Gonzalez-Olmos et al., 2012). In a similar perspective, Fe-ZSM5 catalyst was prepared by an ion-exchange method and used for removal of diclofenac, an anti-inflammatory drug. Characterisation by scanning electron microscopy (SEM) and inductively coupled plasma mass spectrometry (ICP-MS) gave insights about the morphology and composition of the catalyst. After two hours of treating diclofenac under optimal conditions ([H2O2] = 50 mM, [Fe]FeZSM5 = 2.0 mM, UV-A), they reported low toxicity and biodegradability. Catalyst also gave a similar performance in consecutive runs (Perisic et al., 2016). In another study, Fe3O4 nanoparticles were deposited on zeolite-Y through a wet impregnation method (Yang et al., 2019). A 9% iron-loaded zeolite degraded 90% phenol at neutral pH within two hours. The possible adsorption of phenol on the catalyst and the splitting of H2O2 to generate •OH is schematically represented in Fig. 10.The ideas of improving the efficiency of Fenton catalyst are well explored in literature, but the strategies for improving the H2O2 utilization are relatively ignored. Since the Fenton reaction is applied for treating wastewater with massive loads of contaminants, it needs the external addition of large amounts of H2O2. Optimization of H2O2 amount could make the Fenton reaction considerably cheaper. Here the idea is to reduce the self-decomposition of H2O2 to H2O and O2 and to prevent the reaction of excess H2O2 with •OH. In a recent study, Wang et al. (2020a) prepared a Fe3O4-zeolite-cyclodextrin catalyst (F-Z-C), which acted as a nano-reactor capable of controlling the local reactant concentration. Cyclodextrins are known for forming inclusion complexes with the guest molecules through van der Waals forces, chemical bonds etc (Saenger, 1980 ). In this process, the dispersed contaminants (e.g. methylene blue) adsorb on the F-Z-C catalyst which result in increased contaminant concentration over a local region. Then the •OH generated on the catalyst-water interface reacts with the adsorbed contaminants and degrades them. Also, the F-Z-C catalyst can store excess H2O2 and release it once necessary. This ‘storage- release’ effect prevents the self-decomposition of H2O2 and improves the H2O2 utilization efficiency.Metal-organic frameworks (MOFs) are a class of supramolecular assemblies formed from the interaction of various metal ions and organic ligands (Deria et al., 2014; O’Keeffe and Yaghi, 2012; Howarth et al., 2016). The high porosity, excellent surface area and capacity to act as nano-reactors attracted wide attention on MOFs for various applications such as gas storage, liquid-phase separations, catalysis, drug delivery etc (Li et al., 2011; Wang et al., 2015; Kuppler et al., 2009; Denny et al., 2016; Wang and Astruc, 2020; Horcajada et al., 2012; Li et al., 2009). In the past decade, a surge has been observed in the utilization of iron-contained MOFs for heterogeneous Fenton reactions owing to their efficient •OH generation capacity (Liu et al., 2017; Cheng et al., 2018; Sharma and Feng, 2019). Fe-MOFs have a higher tendency for the recombination of generated electron-hole pairs resulting in lower photocatalytic activity (Liang et al., 2015; Liu et al., 2018). Nowadays, preparation of composite materials with Fe-MOFs and semiconductor is devised as a successful strategy to promote the charge transfer efficiency (Chandra et al., 2016; Shen et al., 2015). MIL (Materials Institute Lavoisier) group of metal-organic frameworks are one of the most explored classes of MOFs in the field of environmental remediation (Farha and Hupp, 2010; Janiak and Vieth, 2010). Li et al. (2018) formulated a one-pot solvothermal synthesis method for the preparation of TiO2 @MOF (TiO2 @NH2-MIL88B-Fe) heterostructures (abbreviated as SU-3, where three stands for the optimal molar ratio of Ti: Fe). The TiO2 @MOF displayed enhanced photodegradation of methylene blue (MB) under visible LED light. To identify the key reactive species in the Fenton system, experiments were performed with EDTA (h+ scavenger), p-benzoquinone (O2 •- scavenger) and tert-butyl alcohol (•OH scavenger). The excess addition of tert-butyl alcohol slowed down the degradation of MB, and it indicated the vital role of •OH in the Fenton reaction. Electron spin resonance (ESR) technique employed using 5,5-dimethyl-1-pyrroline N-oxide (DMPO) as a spin trapper resulted in a characteristic four peak signal (ratio 1:2:2:1) attributed do the DMPO-•OH, indicating the presence of •OH (Du et al., 2017). A plausible mechanism of the photocatalytic decomposition of MB by SU-3 is represented in Fig. 11.In another study, Fe3O4 and carbon aerogel (CA) were combined with the MIL-100(Fe) and the composite material was evaluated for the removal of tetracycline hydrochloride (TC), a contaminant of emerging concern, under UV light. The CA incorporated nanocomposite with a higher surface area (389 m2 g-1) enhanced the performance by 1.6 times compared to MIL-100(Fe)@ Fe3O4. Also combining the system with CA resulted in improvement of the water stability of the MOF (Rasheed et al., 2018). In a similar study, Wu et al. (2020) prepared a series of Fe-MOFs and studied the role of Fe-oxo clusters in the framework for the degradation of TC. Fe-oxo clusters were regarded as the absorption antennae, and Fe-MOFs showed substantial absorption in the visible region. Among the MIL-101, MIL-53 and MIL-88B, photo-Fenton activity was highest in the MIL-101, because of its higher number of coordinatively unsaturated iron sites. Generally, the metal centres in a MOF are fully occupied by the organic ligands, and this inhibits the metal sites from activating H2O2. So, introducing coordinatively unsaturated metal sites becomes a successful strategy in activating H2O2. Tang and Wang (2018) prepared the CUS-MIL-100(Fe) having open iron centres and reported 100% degradation of sulfamethazine (a commonly used sulfonamide antibiotic). Also the Fenton experiments performed for multiple cycles, revealed excellent structural stability of the catalyst and minimal leaching of iron which is lower than the environmental standards demanded by European Union (<2 mg/L).In a recent report, distinct nano-architectures of road-like, spindle-like, and diamond-like MIL-88A-Fe were prepared and studied to examine the correlation of different exposed crystal facets towards catalytic performance (Liao et al., 2019). The shape-selective synthesis of the catalyst was achieved by simply varying the water/DMF ratio during the solvothermal synthesis, and the contribution of (100) facet decreased upon increasing the quantity of DMF. The peak area analysis of XRD pattern showed that (100) facet has a ratio of 60%, 30% and 15% in the road-MIL-88A-Fe, spindle-MIL-88A-Fe and diamond-MIL-88A-Fe, respectively. Also, DFT studies revealed the easier activation of H2O2 on the (100) crystal facet compared to (101), and the rod-MIL-88A-Fe was chosen as the best catalyst for Fenton reaction.Zero-valent iron (ZVI) with a standard reduction potential of EH 0 (Fe2+/Fe0) = −440 mV is considered as an effective reducing agent (Fu et al., 2014). ZVI can give out two electrons in the presence of H2O2 or O2 and form the Fe(II) species responsible for Fenton reaction (Eq. 16 and 17). ZVI is known for reacting differently in aerobic and anaerobic conditions. In aerobic conditions, it acts by oxidizing contaminants and in anaerobic condition by reducing contaminants (He et al., 2016). (16) Fe0+O2+2 H+→Fe(II)+H2O2 (17) Fe0+H2O2+2 H+→Fe(II)+2 H2O (18) Fe(II)+H2O2→Fe(III)+·OH+OH- Over the past decade, ZVI was demonstrated for treating different varieties of organic and inorganic contaminants such as dyes (Wang et al., 2017), phenolic compounds (Ambika et al., 2020; Minella et al., 2019), antibiotics (Zhou et al., 2019), nitroaromatic compounds (NACs) (Zarei et al., 2019), arsenic (Tucek et al., 2017), heavy metals (Li et al., 2017), chlorinated organic compounds (Ezzatahmadi et al., 2017), nitrates etc (Ezzatahmadi et al., 2017 ). The heavy metal iron content in wastewater poses significant challenges during its treatment (Demirbas, 2008; Fu and Wang, 2011). In the last decade, ZVI emerged as a catalyst for treating wastewater, having a large load of heavy metal content (Vilardi et al., 2018; O’Carroll et al., 2013). Li et al. (2017) tested conventional adsorbents and precipitants such as activated alumina, Ca(OH)2, Fe3O4, nanoTiO2 etc. for removing heavy metal ions and compared their performance with the ZVI. However, only the ZVI demonstrated simultaneous removal of multiple heavy metals present (Ni(II), Zn(II), As(V) and Cu(II)). Since the standard reduction potential of copper is more positive compared to that of ZVI, it can easily receive electrons from ZVI. Interestingly, the chemical reduction method of Cu(II) and As(V) using ZVI is less susceptible to changes by the difference in pH or introduction of chelates. For metal ions such as Zn (II) and Ni(II) having standard reduction potential more negative than that of ZVI, precipitation, adsorption and electrostatic attraction are the major methods for the metal ion removal (Li et al., 2017). The oxyanions of arsenic were primarily removed by co-precipitation with Fe(II).Zero-valent iron microspheres have shown to be efficient catalyst for UV light photo-Fenton processes with a TOC removal of 99% for phenol and 83% for oxalic acid (Blanco et al., 2016). The studies also demonstrated the degradation of 90% 1,4-dioxane by ZVI microspheres under solar irradiation for 180-minutes. The results of this study showed similar values of degradation at various pH values and indicate that iron solubilisation is not an essential step in this process and the photo-Fenton reaction is taking place on the surface of the catalyst (Barndok et al., 2016). In a similar study, nano-ZVI was evaluated for the removal of ciprofloxacin antibiotic. With a ratio of 5:1 for nZVI: H2O2, a 99.3% removal of ciprofloxacin was achieved in 120 min at neutral pH. The experiment performed under UV light demonstrated a 100% degradation of ciprofloxacin within 25 min (Mondal et al., 2018). Kakavandi et al. (2019) reported the use of nZVI supported on kaolinite (a layered silicate mineral) for the removal of acid black 1 (AB1) dye. The reaction was performed at pH 2.0 with 0.3 g/L of catalyst, and even after four cycles the catalyst was efficient in removing 72% of the dye, indicating catalyst durability and potential for reuse. In a recent study, Jiang et al. (2020) investigated the role of formic acid in the pathway for degradation of prechlorinated organic contaminants. One of the major concerns about chlorinated contaminants is their resistance to degradation via an oxidative pathway. In the study, formic acid acted as a scavenger of •OH and generated carbon dioxide radical (CO2 •-). Carbon dioxide radical is known for transferring one of its electrons to chlorinated contaminants and performing the degradation by a reductive pathway. They studied the role of formic acid in the generation of CO2 •- and confirmed the presence of carbon dioxide radical by electron paramagnetic resonance (EPR) analysis. The generalised mechanism for oxidative and reductive routes of degradation followed by ZVI is summarised in Fig. 12.The first report of inactivation of two waterborne viruses ΦX174 and MS-2 using ZVI came in 2005 (You et al., 2005; Hossain et al., 2014). Later Lee et al. (2018) demonstrated that nano zero-valent iron (nZVI) could act as a potent bactericide under anaerobic conditions. Under aerobic conditions, a more significant amount of nZVI was required for inactivation, possibly due to the surface corrosion and oxidation of nZVI by oxygen. Another study in 2009 by Diao and Yao (2009) reported the inactivation of Pseudomonas fluorescens (gram-negative bacteria), Bacillus subtilis var. niger (gram-positive bacteria) and Aspergillus versicolor (fungus) using nZVI. A sulfidated micro zero-valent iron (S-mZVI) was studied for the removal of antibiotic-resistant E. coli bacteria and antibiotic-resistant gene (ARG) TetB. The S-mZVI was prepared in a planetary ball mill by mixing sulfur and mZVI in a molar ratio of 20 (Fe/S). SO4 •– and •OH radicals generated was attributed to the effective removal of ARG (Zhang et al., 2020). Similarly, a Fe/Ni nanoparticle system was synthesised by a liquid-phase reduction method using NaBH4. Here the bimetallic system displayed superior activity compared to the ZVI in the removal of f2 bacteriophage. An optimum ratio of 5:1 (Fe/Ni) showed the highest catalytic performance and both metals existed in the nanoparticle as Fe0 and Ni0 (Cheng et al., 2019).Chlorination is one of the popular methods employed for disinfection of drinking water, but it produces carcinogenic disinfection byproducts (DBPs) like chloroform, chloroacetic acid etc (Hrudey, 2009; Hua and Reckhow, 2007). Therefore, ZVI is of interest since it will not result in the production of any DBPs. Two of the limiting factors that prevent the practical application of ZVI include poor dispersibility and its low disinfection efficiency. Because of the lower disinfection efficiency, a higher dosage of ZVI is required, and that adds to the cost of the process. The electrostatic and magnetic interactions between ZVI particles lead to its aggregation and poor dispersibility. Also, under aerobic conditions, the iron oxide layers formed over the ZVI decelerates the electron transfer from the ZVI core to the exterior (Fu et al., 2014; Sun et al., 2019). Recently Sun et al. (2019) prepared amorphous zerovalent iron microspheres (A-mZVI) and crystalline nanoscale zerovalent iron (C-nZVI) for studying the efficiency of removal of E. coli under aerobic conditions. C-nZVI produced •OH on the surface of a catalyst but A-mZVI generated •OH by the iron dissolution and oxygen activation in the solution. SEM and TEM images of the E. coli treated with C-nZVI depicted the spherical nanoparticles on the surface of bacteria. But the sample treated with A-mZVI showed a thick covering of interconnected flakes on the surface of E. coli bacteria. In the A-mZVI, a fast dissolution of Fe2+ was observed, and later it deposited on the E. coli as iron oxy-hydroxy species. A thicker adsorption of oxide layers over the bacteria resulted in better inactivation efficiency by A-mZVI. Also, the magnetization studies revealed that the corrosion products of A-mZVI are essentially non-magnetic. Therefore, the sedimentation of the reaction mixture after treatment with A-mZVI was not influenced by a magnet. The difference in the mode of action of C-nZVI and A-mZVI and their physical mode of separation are schematically represented in Fig. 13. Since this technique employs direct gravitational precipitation of the inactivated bacteria covered by iron oxides, it can also prevent the release of ARG into the water (Sun et al., 2019).ZVI is an abundantly available, non-toxic, and comparably low-cost material that has also shown applications as heterogeneous Fenton catalyst (Fu et al., 2014). It has successfully demonstrated the removal of microorganisms, heavy metals, and contaminants of emerging concern from drinking water, and it also functions without formation of toxic disinfection by-products (DBP) (Giannakis et al., 2016; Sun et al., 2019; Du et al., 2020). Because of its versatility, it has great potential for future applications in large-scale water treatment plants. However, it is a challenging task to understand the complex mechanism of action of ZVI because its mechanisms of action may involve oxidation, reduction, co-precipitation, surface adsorption etc. Its mechanism also varies according to the contaminants which it reacts with. Also, ZVI treatment may result in the formation of smaller quantities of corrosion products such as Fe(OH)2, Fe(OH)3, Fe2O2 etc. and they could be detrimental to the pipelines in water distribution channels (Fu et al., 2014). Table 1. summarises operational parameters used for performing the Fenton reaction and the catalytic activity observed for different classes of heterogeneous Fenton catalysts.The implementation of Fenton reaction for real-water applications becomes complicated due to the presence of a complex matrix of organic substances present in real water called as Natural Organic Matter (NOM) (Giannakis et al., 2016). Various studies have shown the positive impact of NOM in enhancing the efficiency of Fenton reaction (Spuhler et al., 2010; Huling et al., 2001; Georgi et al., 2007; Vione et al., 2006; Moncayo-Lasso et al., 2009). Fe(III) can form complexes with the NOM (Fe3+-NOM), and this complex is less prone to precipitation and depict higher absorption in the UV–visible range (Voelker et al., 1997; Walte and Morel, 1984). Georgi et al. (2007) reported that the presence of humic acid (a type of NOM) in the Fenton system (50–100 mg/L) had shifted the optimum pH of the reaction towards the neutral condition. Ruales-Lonfat et al. (2015a) compared the E. coli inactivation of hematite with the Milli-Q water and water collected from Geneva lake. The natural water was not interfering with the photocatalytic semiconducting action of hematite (hematite/light/water) and with both water samples complete E. coli inactivation was observed within 120 min. When H2O2 was introduced to the system (hematite/light/water/H2O2), natural water system showed slightly higher inactivation rate in comparison to Milli-Q water. This observation could be related to the photo Fenton action showed by iron species which got complexed and solubilised by the NOM. The formed complex enhances the reaction rate by participating in the homogeneous Fenton reaction.On the contrary, some other studies report the inhibition of Fenton process in the presence of NOM (Bogan and Trbovic, 2003; Lindsey and Tarr, 2000; Lindsey and Tarr, 2000). Fenton experiments performed to degrade polycyclic aromatic hydrocarbons in the presence of humic acid, and fulvic acid (classes of NOM) exhibited the inhibition of •OH formation (Lindsey and Tarr, 2000). In a similar study Lindsey and Tarr (2020) observed four times lower radical formation in natural water compared to pure water. Since the classes of NOM present in real water varies depending on the source of water, detailed localised studies are needed to understand the effect of NOM and its interaction with the heterogeneous Fenton catalyst in either enhancing or inhibiting the Fenton reaction process.Various parameters such as light source, dosage of catalyst etc. need to be optimised for the efficient performance of Fenton reaction. Choice of the perfect light source for the photo Fenton reaction is a critical aspect considering its economic viability. The optimum light radiation required for the photo Fenton process is in the UV region and near-visible spectrum up to 560 nm wavelength (Carra et al., 2015). Solar light is a sustainable source of energy, and in areas where the availability of sunshine is limited, artificial light sources are required. Mercury UV-lamps were the common source of light for photo Fenton reaction (Guimarães et al., 2019). But considering its low energy efficiency and possible mercury contamination, Xenon lamp-based sun simulators are widely used for photo Fenton reactions (Hu et al., 2019; Cai et al., 2016). Recent studies report the use of light emitting diodes (LEDs) as an energy-efficient light source in the heterogeneous photo Fenton reaction (Zhu et al., 2018a, 2018b). LEDs have a longer lifespan compared to Xenon lamps, and they convert less amount of energy in the form of heat (Carra et al., 2015; Matafonova and Batoev, 2018).Optimising the dosage of catalyst, H2O2 etc. is an important procedure in minimising the operating cost and achieving the highest catalytic efficiency of Fenton reaction. The amount of natural NOM, carbohydrates, proteins and inorganic species (carbonate, nitrate, sulphate, etc.) present in water are the specific parameters that define a water matrix (Lado Ribeiro et al., 2019). Photo Fenton inactivation experiments performed on E. coli. bacteria in the urban wastewater sample resulted in a 2.43 log disinfection (Rodríguez-Chueca et al., 2012). When the water matrix was changed to distilled water, a log 5.81 inactivation was observed. In a recent report, Ling et al. (2018) studied the effect of chloride and phosphate on the ZVI in a Fenton reaction. Chloride ions were shown to have accelerated the decomposition of H2O2 and enhanced the reaction rate, but the phosphate ions were observed to inhibit the H2O2 decomposition. It was assumed that the insoluble iron phosphate layer formed on the ZVI surface could have blocked the catalytic sites on ZVI and thus resulted in a decreased reaction rate. A proper investigation of the critical parameters of the water matrix can provide a rational understanding of the possible interactions of catalyst and H2O2 with organic/inorganic compounds of the water matrix and obtain insights on the potential effects on Fenton reaction rate.Reusability/recyclability of heterogeneous Fenton catalysts is an essential parameter in view of its economic implications. Multiple cycles of Photo Fenton reaction performed with ammonia modified graphene/Fe3O4 catalyst resulted in the lowering of pollutant degradation efficiency of the catalyst (Boruah et al., 2017). The first cycle showed 92.43% degradation of phenol whereas, at the 10th cycle phenol degradation reduced to 75.50%. A similar decrease in the catalytic efficiency over multiple runs is observed for almost all types of heterogeneous Fenton catalysts (Tang and Wang 2018). An environmentally friendly sustainable model of reusing/recycling the heterogeneous Fenton catalyst is a highly sought-after area of research. Recently Serrà et al. (2020) reported an interesting zero-carbon-emission circular process resulting in water remediation and energy production. Spirulina microalgae was cultivated in wastewater which was rich in iron and heavy metals. During the photosynthetic process, spirulina utilised carbon dioxide and released oxygen to the atmosphere. This microalgal biomass was fermented and produced bioethanol. The resultant biomass was dried and burnt for thermal energy generation. The iron-rich ash obtained after combustion was utilised as heterogeneous photo Fenton catalyst under sunlight for wastewater treatment. Finally, the mineralised water was reused for cultivating the next batch of microalgae and completed the cycle. The research area of zero-carbon-emission circular processes using heterogeneous Fenton catalysts is expecting significant advances in recent years. A large scale implementation of similar techniques will effectively contribute to tackling the current global issues like excessive CO2 emission and atmospheric pollution (Keijer et al., 2019).Even though most reports on Fenton catalysts are successful in showing better results of degradation of organic contaminants and transformation of inorganic contaminants, more work and insights are needed to understand the mechanistic aspects of the reactions involved. Theoretical understanding of the underlying mechanisms of the Fenton and Fenton-like reactions by density functional theory (DFT) studies can provide new insights and information in this area of research (Buda et al., 2001). DFT studies help in understanding the role of surface defects in enhancing the catalytic activity of heterogeneous Fenton reactions. Only by the in-depth understanding of the mechanism, further improvements in the catalyst performance can be achieved. Recently, static and dynamic DFT calculations performed by Hsing-Yin Chen and co-workers on the intermediates of Fenton reaction concluded that •OH is the predominant species below pH 2.2 (English for Writing Research Papers Useful Phrases, 2016). This study also reported that as the pH increases from 2.2 to 4.6, iron(IV)-oxo complex [(H2O)5FeIVO]2+ was the major complex and at pH >7.9 a deprotonated dihydroxy species, [(H2O)3FeIVO(OH)2] was the active intermediate. Nonetheless, the high-level ab initio calculations question the presence of dihydroxy species in aqueous Fenton reactions and yet there are no reports of the successful synthesis and characterisation of [(H2O)3FeIVO(OH)2] species in the literature. Likewise, it is essential to explore the electron transfer mechanism in the Fenton catalyst by quantum chemical calculations (Qin et al., 2017; Vorontsov, 2019). Later the theoretical studies should be validated with necessary experimental evidence.The lack of standardised procedures for reporting the catalytic activity is a major concern in the heterogeneous photo-Fenton process. Different groups of researchers use various ratios of concentrations of catalyst: H2O2 for performing the reactions (See the catalytic activity summarised in Table1). Therefore, a standardised procedure for comparing the efficiency of the catalyst is inevitable. In many cases, even though the authors argue about carrying out the reaction at a neutral pH, the reaction pH changes during the experiment. Usually, the pH of the reaction mixture decreases due to the formation of smaller degradation products such as oxalic acid, formic acid etc (Kuan et al., 2015). Hence, use of buffers and/or continuous monitoring of pH is necessary while performing the reaction. Another challenge that needs to be addressed is the leaching of iron species from the catalyst and deactivation of catalytic sites by the adsorption of impurities from wastewater. So, more work is needed to understand catalyst stability and longevity as well as technologies that can use the catalyst in a sustainable way.2-dimensional (2D) nanomaterials with high surface area have played a significant role in various energy and environmental applications. 2D materials such as MXene composites are less explored in the Fenton chemistry, and a great deal of attention should be paid on them to unveil their potential. Also, many of the reports on heterogeneous Fenton catalysis show higher activity in the ultraviolet region of the electromagnetic spectrum. Since the sun can be a low-cost source of energy (i.e., depending on the technology for capturing and utilizing the radiation) and because of its inexhaustible nature, researchers need to target more on the synthesis of catalysts which are visible light active and have high activity under visible light spectrum. When it comes to the broad-scale application of photo-Fenton reaction for water treatment, materials that utilize a broader spectrum of sunlight and doing so at high quantum yields will have more potential for practical applications. Nowadays, the research community is witnessing computer-aided design of various catalysts. So, researchers should come forward for applying in silico tools in designing novel catalytic materials having excellent activity for photo Fenton reaction (Poree and Schoenebeck, 2017).The toxicity of nanomaterials is itself a widely debated topic. The nanoparticles embedded in different matrices get released into the environment, and they interact with the living cells in a dynamic fashion (Hamers, 2017). Nanomaterials which are synthesised with distinct properties, may undergo chemical changes once they are released into the environment. Hence a comprehensive understanding of the interaction of nanoparticles with the living matter is necessary for the large-scale application of these catalysts and for avoiding any future health hazards. Different reports present great achievements in degrading various contaminants of emerging concern via Fenton and Fenton-like processes, but excellent outcomes in complex wastewater matrices are still rare in the literature. So, opportunities for combining heterogeneous photo Fenton process with other established wastewater treatment technologies should be explored for commercial-scale applications.Photo-Fenton reaction is proved to be a promising method for the removal of bacterial and fungal pathogens from wastewater (O’Dowd and Pillai, 2020; García-Fernández et al., 2012). Also, there are successful reports on the removal of MS2 coliphage (bacteriophage having similar properties to that of human enteric viruses) from wastewater through photo-Fenton treatment (Nieto-Juarez et al., 2010; Ortega-Gómez et al., 2015). Nieto-Juarez and Kohn (2013) investigated the fate of MS2 coliphage upon photo-Fenton treatment with four commercially available iron oxide species (hematite, goethite, magnetite and amorphous Fe(OH)3. The study reported 99.9% virus removal by all the iron species studied via a photo-Fenton process combined with physical removal such as adsorption. A recent study reports the effectiveness of homogeneous solar photo-Fenton for inactivation of hepatitis A virus (an extremely resistant non-enveloped virus) in water (Polo et al., 2018). Since the SARS-CoV-2 is an enveloped virus, which remains more susceptible to disinfectants compared to a non-enveloped virus (Chu et al., 2019), photo-Fenton reaction offers the possibility of effectively deactivating SARS-CoV-2 virus present in wastewater. Therefore, future studies should aim at identifying suitable heterogeneous Fenton catalysts for virus removal with respect to their inactivation efficiency and absorption capacity. The research area of heterogeneous photo-Fenton has a lot of room for development, and it has remarkable potential in addressing pressing challenges in industrial-scale water treatment.Photo-Fenton treatment of wastewater is an evolving technology for removing organic, inorganic, and microbial contaminants from water. In comparison to the homogeneous Fenton reaction, heterogeneous catalysts have displayed great potential for commercial applications in view of their wide pH range of application, low sludge formation, and reusability. Along with the recent literature reports on the advances in materials for heterogeneous Fenton reaction, schematic illustrations in this review provide a basic understanding of the electron transfer mechanisms and the formation of reactive oxygen species in Fenton reactions. The excellent magnetic properties of materials such as ferrites, magnetite, and their composites have enabled significant developments, due to easy separation and reusability of such materials after wastewater treatment. The incorporation of plasmonic materials with iron minerals to broaden their visible light absorption capacity is highlighted as a growing area of research. Various supporting materials used in the heterogeneous Fenton catalytic systems have a specific role in altering the catalytic activity of the system. Incorporation of semiconductor nanoparticles and carbon-based two-dimensional materials with iron minerals to speed-up the electron transfer and Fe(II) regeneration has shown promise and is expected to receive more attention in the coming years. The higher specific surface area provided by graphene-related supporting materials leads to enhanced adsorption of pollutants on the Fenton catalytic surface. Pillaring process of clay materials exposes the active site of the catalyst and the specific hydrophilic nature of LDHs assists the interaction of hydrophilic contaminants with LDHs. Literature review on studies on the application of Fenton, Fenton-like, and photo-Fenton technologies show these processes play important role in the inactivation of pathogenic microorganisms and destruction of contaminants of emerging concern in various types of wastewaters. Along these lines, there is potential for further enhancement of process performance when introducing catalysts of good selectivity, synergistic action, and versatility for the treatment of various types of source waters/wastewaters polluted with a variety of pathogenic microorganisms, organic and inorganic contaminants of concern. Fe-contained perovskites are seeing growing interest in Fenton-like processes, especially on aspects related to innovative strategies to dope or modify perovskites materials to achieve higher catalytic activity and better process performance. New developments on the synthesis and applications of MOFs and zeolite materials, especially for water treatment, demonstrate there have been important advances in recent years and point towards further progress, especially on topics focused on tailor-designed framework structures with improved functionality and higher activity for more efficient Fenton-like applications. Zero-valent iron-based technologies have seen huge interest in both mechanistic studies and field applications for the treatment of numerous pollutants due to the environmental compatibility of iron and versatility of the process towards oxidation or reduction reactions. It is expected that more applications will be seen in the future, especially if challenges on n-ZVI dispersibility, longevity, and reactivity control are addressed. The materials which can in-situ produce the H2O2 by a two-electron transfer to oxygen and the techniques that improve the H2O2 utilisation by reducing its self-decomposition can considerably decrease the operating cost of the reaction and pave the way for commercial-scale applications of Fenton reaction. Therefore, a sea of opportunities is wide open in this area for the cost optimisation of the existing technology and to develop brand new materials with extraordinary catalytic efficiency.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Image 1 This project has received funding from the European Union's Horizon 2020 Research and Innovation Programme under grant agreement number 820718, and is jointly funded by the European Commission and the Department of Science Technology of India (DST). Dionysiou also acknowledges support from the University of Cincinnati through the Herman Schneider Professorship in the College of Engineering and Applied Sciences.

Heterogeneous Fenton catalysts are emerging as excellent materials for applications related to water purification. In this review, recent trends in the synthesis and application of heterogeneous Fenton catalysts for the abatement of organic pollutants and disinfection of microorganisms are discussed. It is noted that as the complexity of cell wall increases, the resistance level towards various disinfectants increases and it requires either harsh conditions or longer exposure time for the complete disinfection. In case of viruses, enveloped viruses (e.g. SARS-CoV-2) are found to be more susceptible to disinfectants than the non-enveloped viruses. The introduction of plasmonic materials with the Fenton catalysts broadens the visible light absorption efficiency of the hybrid material, and incorporation of semiconductor material improves the rate of regeneration of Fe(II) from Fe(III). A special emphasis is given to the use of Fenton catalysts for antibacterial applications. Composite materials of magnetite and ferrites remain a champion in this area because of their easy separation and reuse, owing to their magnetic properties. Iron minerals supported on clay materials, perovskites, carbon materials, zeolites and metal-organic frameworks (MOFs) dramatically increase the catalytic degradation rate of contaminants by providing high surface area, good mechanical stability, and improved electron transfer. Moreover, insights to the zero-valent iron and its capacity to remove a wide range of organic pollutants, heavy metals and bacterial contamination are also discussed. Real world applications and the role of natural organic matter are summarised. Parameter optimisation (e.g. light source, dosage of catalyst, concentration of H2O2 etc.), sustainable models for the reusability or recyclability of the catalyst and the theoretical understanding and mechanistic aspects of the photo-Fenton process are also explained. Additionally, this review summarises the opportunities and future directions of research in the heterogeneous Fenton catalysis.

The water electrolysis process occurs through two simultaneous half-cell reactions: the oxygen evolution reaction (OER) on the anode and the hydrogen evolution reaction (HER) on the cathode. The Alkaline OER is a 4-electron–proton transfer process that makes the reaction sluggish with high overpotential and complex reaction mechanisms [1,2]. Nickel (Ni)-based compounds including Ni-based oxides and (oxy)hydroxides are among the most efficient precious-metal-free catalysts for alkaline OER due to their desirable advantages such as enhanced reaction kinetics and structure/performance stability [3]. Relationships between metallic Ni and various O-containing surface compounds formed during anodic oxidation of polycrystalline Ni in aqueous alkaline media can be described by the Bode diagram (Fig. 1 ) [4]. Mild anodic polarization of metallic Ni results in the reversible formation of α-Ni(OH)2; moderate anodic polarization results in the irreversible conversion of α-Ni(OH)2 into β-Ni(OH)2 as well as in the direct oxidation of Ni to β-Ni(OH)2; and, this process is accompanied by the development of NiO that is sandwiched between Ni and β-Ni(OH)2 (marked as a NiO sandwich in Fig. 2 ). The purple lines and the formation of γ-NiOOH were suggested by Bode [5]. The γ-NiOOH phase is believed to be the highest-achievable Ni oxidation state [6]. It is most commonly assumed that the β-NiOOH oxidation phase is most active towards the OER [7].So far, many research efforts have focussed on improving the OER performance of Ni by the design and optimization of the catalyst structure [6,8,9].Sonoelectrochemistry is the combination of ultrasound with electrochemistry. The use of ultrasound in electrochemistry offers many advantages including [10]: a) gas bubble removal at the electrode surface; b) solution degassing; c) disruption of the Nernst diffusion layer; d) enhancement of mass transport of electroactive specious through the double layer; and, e) activation and cleaning of the electrode surface. Recently, it was reported that ultrasonication greatly enhances the electrocatalytic properties of metallic surfaces [11–16]. Our group also investigated the effect of ultrasound on Ni(poly) in alkaline media and found that the rate of the HER was greatly enhanced.In this work, we investigated the effects of ultrasound (24 kHz) on the OER on polycrystalline Ni immersed in 1.0 M aqueous KOH solution at room temperature. We applied ultrasound (i) during linear sweep voltammetry (LSV) experiments and (ii) for surface treatment of the Ni(poly) electrode for 30 min and then we conducted the LSV experiments under silent conditions (in the absence of ultrasound).All electrochemical experiments were carried out using a potentiostat/galvanostat (BioLogic-SP 150) in a three-electrode configuration. The voltammetry experiments were performed using a double-jacketed sonoelectrochemical cell. Ultrasonication was applied by a f = 24 kHz ultrasonic probe (Hielscher UP200S, 200 W @ 60% fixed amplitude, the tip Ø = 14 mm, and the tip area = 153.9 mm2 (1.5386 cm2). The ultrasonic or acoustic power (P acous) was found to be 44 ± 1.40 W by calorimetrically using the methods of Margulis et al. [17] and Contamine et al. [18]. In order to keep the temperature at T = 298 ± 1 K a refrigerated circulator (JULABO, Germany) was connected to the sonoelectrochemical cell.A polycrystalline nickel Ni(poly) disc (Ø = 5 mm) of geometric surface area (A geom) of 0.196 cm2 was used as a working electrode (WE). The WE was mechanically polished using alumina suspension (down to 0.05 μm, Buehler Micro polish) to obtain a mirror-like surface rinsed with UHP water, ultrasonicated in UHP water for ∼30 s and finally rinsed in UHP water under ultrasonic conditions. The reference electrode (RE) was a homemade reversible hydrogen electrode (RHE) [19]. All potential values in this work are reported with respect to the RHE. The counter electrode (CE) was a Ni mesh (40 mesh woven from 0.13 mm diameter wire, 99.99% metal basis, Alfa Aesar, Germany) in a rectangle shape (20.67 × 10.76 mm2). Its surface area was at least 10 times larger than that of the WE. The distance between the ultrasonic probe and the working electrode was ca. 3 cm. The experiments were carried out in N2 (g) (99.999%) saturated 1.00 M (pH = 13.7) aqueous KOH (Sigma-Aldrich, 99.99% in purity) solution prepared using ultra-high-purity water (Millipore, 18.2 MΩ cm in resistivity).The performance of Ni(poly) towards the OER in aqueous alkaline electrolytes was investigated by a series of linear sweep voltammetry (LSV) in the potential region of + 1.10 ≤ E app ≤ +1.70 V vs. RHE at the potential scan rate of ν = 0.30 mV s−1 in 1.0 M KOH aqueous solutions in the absence of ultrasound (silent conditions), during (with) ultrasound and after 30 min ultrasound.The potential values from linear sweep voltammetry (LSV) experiments were IR corrected using the following equation (1): (1) E IR - c o r r e c t e d = E app - - I R where I is the measured current and R is the electrolyte resistance, measured for each electrolyte employed. The R value was determined by electrochemical impedance spectroscopy (EIS) in the high-frequency region from the value of the real impedance (Z’ ) where the imaginary impedance (Z ’’) is zero in the Nyquist plot. The EIS experiments were carried out in the 100 kHz to 0.1 Hz frequency (f) range with a voltage perturbation of + 10 mV at an applied potential of +1.60 V vs. RHE at T = 298 K.The surface structure and morphology of the Ni(poly) electrodes before and after ultrasound treatment were studied using a scanning electron microscope (SEM) Zeiss-Ultra 55-FEG-SEM operating at 10 kV accelerating voltage.In order to study the effects of power ultrasound on the electrochemical surface area of Ni(poly), the “capacitance” and “β-NiOOH” methods were used. The “capacitance” method consists of cycling the Ni electrodes at different scan rates in a non-faradic charging process to determine the electrochemical surface area (A ecsa) [20]. A series of cyclic voltammograms (CVs) on Ni(poly) in 1.0 M KOH were generated at different scan rates (5, 10, 20, 50, 100, 200, 300, 400 mV s−1) in the potential region of + 0.80 V vs. RHE to + 0.90 V vs. RHE. The double-layer capacitance value (C dl) was obtained by plotting the charging current (I c, A) vs. scan rate (ν, V s−1) and by using equation (2): (2) Slope = C dl = Δ I c Δ v The electrochemical surface area was calculated by using the specific capacitance density (c) of 40 μF cm−2 and equation (3) [20,21]. (3) A ecsa = C dl C Fig. 2a and 2b show the CVs of the Ni(poly) electrode before and after 30 min of ultrasonication at different scan rates (5, 10, 20, 50, 100, 200, 300, and 400 mV s−1) in the potential range of + 0.80 ≤ E app ≤ +0.90 V vs. RHE where non-faradic currents occur. Fig. 2c shows plots of current vs. scan rate at a potential of + 0.85 V vs. RHE before and after 30 mins of ultrasonic exposure.The “β-NiOOH” method consisted of integrating the β-NiOOH reduction peak once steady-state polarization was reached at a high scan rate. The β-NiOOH method was carried out by running 10 CV cycles from + 0.50 ≤ E app ≤ +1.60 V vs. RHE at a scan rate of ν = 100 mV s−1 before and after 30 min US (Fig. 2d). The A ecsa values for this method were calculated using the β-NiOOH reduction peak of the 10th cycle (from 1.2 to 1.4 V vs. RHE) divided by the specific charge density of 420 μC cm−2 (equation (4)) [20]. (4) A e c s a = Q 420 where Q is the charge associated with the β-NiOOH reduction peak. The A ecsa values before and after 30 min of ultrasonication treatment for both capacitance and beta methods are summarised in Table 1 . It needs to be mentioned that the difference between the A ecsa values from the “capacitance” and the “β-NiOOH” methods is related to the basis of measurements of both methods. The capacitance method is related to conductivity and homogeneity of surface for double layer charging while the beta method is related to the faradaic reaction of nickel hydroxide to nickel oxyhydroxide transformation [20]. It can be observed from Table 1 that ultrasound does not seem to affect the electrochemical surface area of the Ni(poly) electrode, indicating that the electrochemical surface area was not significantly modified due to erosion caused by the implosion of acoustic cavitation bubbles on the electrode surface [22]. Fig. 2e and 2f show the SEM images of Ni(poly) before and after 30 min US. Before US a smooth surface is seen except the scratches due to mechanical polishing. After 30 min US some irregular pits could be observed, however, it is unclear whether these arose from the actions of inter-facial ultrasound. Such features are sometimes found widely scattered across non-sonicated surfaces (see, for example, some pits in non-sonicated electrode Fig. 2e). The pit areas in both non-sonicated and sonicated electrodes have been marked red in Fig. 2e and 2f. These pits have little influence on electrochemical measurements because there are very few of them and their contribution to total A ecsa is relatively small. Aqueous ultrasonication did not significantly roughen the electrode and the surface roughness remained almost unchanged [23].The effect of ultrasound on the oxygen evolution reaction (OER) at Ni(poly) in 1.0 M aqueous KOH solution was investigated by linear sweep voltammetry (LSV). Fig. 3 a shows the LSVs for the OER on Ni(poly) in N2 saturated 1.0 M KOH aqueous solutions at a low scan rate of ν = 0.3 mV s−1 before, during (with) and after 30 min ultrasonic treatment. It can be observed that the ultrasonic (US) treatment increases the OER activity. Fig. 3b demonstrates the Tafel plots obtained from the LSV curves in the OER region. Tafel slopes (b) at low and high overpotentials and the potential at + 10 mA cm−2 (E +10 mA cm -2) are tabulated in Table 2 . Results from Table 2 indicate that lower potential requires to reach + 10 mA cm−2 in presence of ultrasound and after ultrasonic treatment. However, even when ultrasound is “on” during the OER experiments, the lower overpotential at + 10 mA cm−2 is required when compared to after ultrasonic treatment.Ni-based materials show the Tafel slope values between 40 mV dec-1 to 130 mV dec-1. Also, it is well known that there are generally two Tafel regions for the OER, separated at ∼ 1.5 V vs. RHE in 1.0 M KOH [6,7]. According to Table 2, the Tafel slopes of 52, 55, 50 mV dec-1 at low overpotentials and 141, 90 and 130 mV dec-1 at high overpotentials were obtained for the OER on Ni(poly) before ultrasonication (US), with US and after 30 min US, respectively. The Tafel slopes are in good agreement with the literature [7,24,25]. By comparing the Tafel slopes under different US conditions reported in Table 2, it can be concluded that ultrasound does not change the Tafel slopes significantly for the OER and does not affect the mechanism of OER. It is worth mentioning that the experiments have been repeated several times and almost the same values have been obtained showing the reproducibility of the work. Fig. 3c illustrates the plot of E at + 10 mA cm−2 (E+ 10 mA cm -2) vs. different ultrasonic conditions. It can be seen in Fig. 3c that the overpotential to reach + 10 mA cm−2 decreases when US is “on” during the OER experiment. Fig. 3d shows the Nyquist representation of the impedance data of Ni(poly) before US, with US and after 30 min US at T = 298 K and E = +1.60 V vs. RHE. For all US conditions, a depressed semi-circle can be seen. Accordingly, the data were approximated by the modified Randles circuit shown in Fig. 3d, whereas the capacitance is replaced by a constant phase element. Note, for α = 1 the CPE reflects an ideal capacitance. R s correlates with the cell ohmic resistance (electrodes). R ct represents the charge transfer resistance and may also include other contributions such as the adsorption of intermediates. CPE is a constant phase element that is often associated with the capacitive charging of a rough electrode. The parameters obtained from the EIS measurement are shown in Table 3 . According to Table 3, the Ni(poly) electrode after 30 min US treatment has the lowest charge transfer resistance compared to the two other conditions. While the R s are almost constant in all US conditions. Since no significant increase in the electrochemical surface has been observed on Ni(poly) by applying US, the enhancement of OER activity of Ni(poly) after ultrasonication treatment could be due to the reaction of radicals at the electrode/electrolyte interface such as (OH•, H•, H2O2, etc) caused by collapsing cavitation bubbles. It was reported before that such radicals could react with the electrolyte species and produce a secondary sonochemical reaction [15,16,26,27].We have developed a simple in-situ method to activate Ni(poly) electrodes in 1.0 M aqueous KOH solution towards the OER by ultrasonic treatment (24 kHz, 60% amplitude, 44 W) for 30 min. It was shown that ultrasound improves Ni(poly) OER activity by reducing the overpotential needed to achieve + 10 mA cm−2 by –23 mV and charge transfer resistance from 98.5 Ω before US to 11.1 Ω after 30 min US treatment. However, the US treatment does not affect the electrochemical surface area of Ni(poly) or Tafel slope. The enhancement of OER activity of Ni(poly) could be attributed to the formation of free radicals by collapsing cavitation bubbles and the secondary sonochemical reactions at the electrode/electrolyte interface. However, understanding the exact reason and the mechanism will still need a wide range of experiments and spectroscopy measurements.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.The authors would like to thank NTNU and ENERSENSE for the 3-year financial support for FF’s doctoral studies. FF would like to thank Professors Gregory Jerkiewicz and Christophe Coutanceau for their useful advice.

The development of cost-effective and active water-splitting electrocatalysts is an essential step toward the realization of sustainable energy. Its success requires an intensive improvement in the kinetics of the anodic half-reaction of the oxygen evolution reaction (OER), which determines the overall system efficiency to a large extent. In this work, we designed a facile and one-route strategy to activate the surface of metallic nickel (Ni) for the OER in alkaline media by ultrasound (24 kHz, 44 W, 60% acoustic amplitude, ultrasonic horn). Sonoactivated Ni showed enhanced OER activity with a much lower potential at + 10 mA cm−2 of + 1.594 V vs. RHE after 30 min ultrasonic treatment compared to + 1.617 V vs. RHE before ultrasonication. In addition, lower charge transfer resistance of 11.1 Ω was observed for sonoactivated Ni as compared to 98.5 Ω for non-sonoactivated Ni. In our conditions, ultrasound did not greatly affect the electrochemical surface area (A ecsa) and Tafel slopes however, the enhancement of OER activity can be due to the formation of free OH• radicals resulting from cavitation bubbles collapsing at the electrode/electrolyte interface.

The fossil-based energy system and the dependence on fossil fuels have exacerbated climate change, resulting in an environmental crisis [1]. A change in the fuel production and consumption strategies become necessary in order to reduce the greenhouse gases and other emissions responsible for the global warming. Any alternative to reduce the dependence on petroleum should address the production of energy and chemicals from renewable feedstocks, such as biomass [2]. Biomass has the potential to decrease net emissions of carbon since the used raw materials grow removing carbon dioxide from the atmosphere by photosynthesis [3].Hydrogen production from biomass, in addition to reducing greenhouse gas emissions, would contribute to the expansion and economic viability of the biorefinery [3]. Hydrogen has been globally accepted as an environmentally friendly fuel, since huge energy is contained in the H-H bond and its combustion only releases water to the environment [4]. Moreover, hydrogen can be used in several technologies such as fuel cells and internal combustion engines or turbines [5,6]. Hydrogen can be produced effectively from biomass through a sort of processes [7–10], among which is the aqueous-phase reforming [11].Aqueous-phase reforming (APR), which can be driven at relatively mild conditions, is able to manage diluted aqueous wastes of different oxygenated hydrocarbons to obtain valuable products (either hydrogen or other value-added chemicals) [12]. APR process was first introduced in 2002 by Dumesic and co-workers [13], and since then, had attracted a considerable R&D activities. In APR, reactants remain in liquid phase, unlike steam reforming (SR), what avoids an energetically demanding vaporization-step. Furthermore, low reaction temperatures shift the Water-Gas Shift (WGS) equilibrium towards further formation of hydrogen with the consequent reduction of carbon monoxide content [14].Glycerol, a major by-product of the biodiesel production process, is one of the 12 platform molecules for biorefineries proposed by the US Department of Energy [15]. A huge surplus of glycerol has been generated in the last years, thus, its valorisation represents a challenge for the biodiesel plants profitability [16]. Aqueous-phase reforming of glycerol comprises the decomposition (Eq. 1) and the Water-Gas Shift (WGS) (Eq. 2) steps: (1) C 3 H 8 O 3 → 4 H 2 + 3 C O (2) C O + H 2 O ↔ H 2 + C O 2 The overall reaction stoichiometry for the ideal APR of glycerol is given by reaction 3: (3) C 3 H 8 O 3 + 3 H 2 O → 7 H 2 + 3 C O 2 As an immature technology, this process requires constant investigation for active and stable catalytic materials and optimization of operating conditions to improve current results to the point of being profitable for the industry [17].Numerous works have focused on noble metal based catalysts for APR, especially Pt and Re, due to their high efficiency for C–C, O–H and C–H bonds cleavage and WGS reaction [17–24]. Thanks to its higher availability and economy, Ni-based systems have also been widely studied as an alternative to those upscale metals [25–29]. Cobalt is another transition metal that has attracted attention for this type of process [30–32]. Nevertheless, leaching is a large drawback for transition metals. Both catalytic systems, based on precious and transition metals, present a certain deactivation, mainly due to hydrothermal instabilities [33]. Among the strategies considered, bimetallic catalysts upgrade glycerol conversion and gaseous products and improve stability [30,34–37]. For instance, bimetallic Pt-Co supported on multi-walled carbon nanotubes increases the glycerol reforming activity of the monometallic catalyst by 4, and the WGS activity by 32 [30].APR process is clearly impacted by operating variables as a substrate concentration, temperature and system pressure, and contact time, among others [38,39]. Several authors have optimized temperature and pressure conditions to enhance gaseous products [19,40–43], a few others have conducted research that address other parameters such as feed concentration, mass of catalyst/ reagent mass flow rate ratio, reaction time and feed flow rate [38,39,44]. The reported results, however, become contradictory since they depend on the interrelation with other variables and the reaction system [39]. Moreover, most of the literature is focused on catalyst performance with respect to either gas o liquid phase product distribution.The present work aims to investigate the effect that operating conditions exert on the product distribution during the APR of glycerol over a 0.3PtCoAl catalyst, in order to maximize the hydrogen production by APR. This catalyst, synthesized by impregnating Pt on cobalt aluminate support, has been previously tested in long-term reactions (100 h TOS), proving to be efficient for H2 production and stable [45]. Due to its promising performance, in this work this optimized catalyst formulation was used as a benchmark to search the suitable reaction conditions for hydrogen production. For this purpose, the most handled process variables, such as glycerol concentration in the feedstream, coupled temperature/pressure and contact time, were investigated. The catalytic performance was evaluated based on the most commonly applied reaction indices, and a comprehensive analysis of both gaseous and liquid products is presented. In addition, exhausted catalyst was also characterized to gain knowledge in the main deactivation causes.Bimetallic 0.3Pt/CoAl catalyst was synthetized in two steps. First, cobalt aluminate with a nominal Co/Al mole ratio of 0.625, was synthesized by coprecipitation. An aqueous solution containing appropriate amounts of Co and Al precursors (10.3 g of Co(NO3)2·6H2O and 21.2 g of Al(NO3)3·9H2O) was added dropwise to a vigorously stirred solution containing sodium carbonate while pH was adjusted to 10 with NaOH solution (2 M). The resulting slurry was aged for 24 h at room temperature, filtered, washed several times with de-ionized water and dried in an oven at 110 °C overnight. The cobalt aluminate spinel was formed by calcination at 500 °C (heating rate 5 °C/min) for 5 h in a static air atmosphere. Thereafter, Pt was impregnated (nominal loading 0.3 wt%) using aqueous solution of tetraammineplatinum(II) nitrate as precursor, in the solution/support proportion of 1.5/1 (vol./vol.). After impregnation, the sodden solid was dried in an oven at 110 °C for 17 h and finally, calcined at 350 °C (heating rate 5 °C/min) for 5 h.The bulk composition of the catalyst was evaluated by ICP-AES. The specific surface area and the main pore size were estimated by the BET and BJH methods, respectively. The measurement was performed using nitrogen at 77 K as an adsorbent gas (Tristar II 3020). Prior to the physisorption measurement, the sample was outgassed at 300 °C for 10 h in order to clean the solid surface.XRD diffraction patterns of the calcined, reduced and spent catalyst were obtained on a PANalytical X´pert PRO diffractometer (CuKα radiation, λ = 1.5406 Å, graphite monochromator)), with a step size of 0.026° (2θ) and a counting time of 2 s. The crystallite average size was calculated by Scherrer equation from the peak broadening and the identification of the crystal phases was carried out on the basis of ICDD database. 27Al Solid State NMR measurements at 104.26 MHz for 27Al were performed (9.4 T Bruker AVANCE III 400 spectrometer). Chemical shifts were referenced externally to the AlCl3 aqueous solution at 0 ppm. The spectra were acquired at a spinning frequency of 60 kHz employing a PH MASDVT400W BL 1.3 mm ultrafast probe head.The XPS analyses were performed on a SPECS spectrometer with Phoibos 150 1DDLD analyzer and a monochromatic X-ray beam Al K target (1486.7 eV). The electron energy analyzer was operated at pass energy of 30 eV and step size of 0.08 eV. The C 1 s photoelectron line (BE = 284.8 eV) was used to calibrate the binding energies of the photoelectron. The catalyst was analyzed either in calcined and reduced form. The reduction of the catalyst was carried out in-situ at 600 °C with 20% H2/Ar flow, for 1 h.Temperature programmed reduction of the fresh calcined (H2-TPR) catalysts was carried out in a Micromeritics AutoChem 2920 apparatus, equipped with a thermal conductivity detector (TCD). About 50 mg of sample was initially heated in He stream at 550 °C for 1 h (heating rate 10 °C/min). Then, sample was cooled down to room temperature into Ar flow, and switched to 5% H2/Ar flow while temperature was ramped to 950 °C at 10 °C/min, and hold for 1 h.Temperature programmed hydrogenation (TPH) was conducted on the spent catalyst is order to analyse carbonaceous deposits. Sample was first heated at 550 °C for 1 h, under a He flow, and cooled down to ambient temperature. Then, a flow of 5% H2/Ar was passed through the sample heated at 10 °C/min up to 950 °C and m/z = 15 (CH4) signal was recorded with mass spectrometer (Pffeifer Vacuum OmniStar).The amount of surface Pt and Co sites were evaluated by H2 pulse chemisorptions (5% H2/Ar, loop volume 0.5312 mL) at 40 °C (Micromeritics AutoChem 2920 equipment). Initially, catalyst surface was cleaned by passing a He flow at 500 °C. First, H2 pulse was applied on sample reduced at 250 °C (to titrate the metallic Pt). Thereafter, sample was further reduced at 600 °C, and subsequent pulse chemisorption was completed (to titrate the total metallic sites). H/Me (Me=Pt, Co) stoichiometry of 1/1 was assumed. The exposed metallic area of Pt and Co (SPt o and SCo o) was calculated assuming 0.084 nm2 and 0.0662 nm2 per Pt and Co sites, respectively. The average Pt size was calculated by formula dPt o (nm) = 6000/(ρ·SPt o) [46].The surface acid and base properties of the reduced solid were evaluated by temperature programmed desorption (TPD) of NH3 and CO2, respectively, conducted in a Micromeritics AutoChem 2920 equipment coupled to Mass Spectrometer (MKS Cirrus). Previously, sample was cleaned by passing a He flow at 550 °C for 1 h and cooled down to room temperature. Then, the solid was reduced at 600 °C in 5% H2/Ar flow (heating rate 10 °C/min), hold for 2 h and cooled down in He flow to 90 °C. Then, a series of 10% NH3/He or 5% CO2/He pulses were introduced at 90 °C. Subsequently, the reversibly adsorbed NH3 or CO2 was evacuated by flowing He for 60 min. Finally, the temperature was ramped to 950 °C at a heating rate of 5 °C/min, and the signals m/z = 17 (NH3) and 44 (CO2) were monitorized (MS Pffeifer Vacuum OmniStar). The total amount of acid and basic sites was calculated from the integration of pulse areas, whereas the strength was evaluated from the corresponding TPD curve. The model reaction of skeletal isomerization of 33DM1B (3,3-dimethyl-1-butene) was used to characterize the Brønsted acid sites. The catalyst (ca. 100 mg) was in-situ reduced, and cooled down to the reaction temperature (300 °C) under inert gas-flow. The 33DM1B partial pressure and flow rate were set at 20 kPa and 15.2 mmol/h, respectively. The obtained products were online analysed by GC (column RTx-1, Restek) coupled to a flame ionization detector. The percentage of leached metal was measured by means of ICP-MS analysis of the resulting liquid aliquot.The APR activity tests were carried out in a fixed-bed up-flow reactor (Microactivity Effi, PID Eng&Tech). The catalyst (particle size between 40 and 160 µm) was placed on a stainless steel frit, covered with a quartz wool plug, and in-situ reduced under 10% H2/He flow at 600 °C for 2 h (heating rate 5 °C/min) at atmospheric pressure. The reactor pressure was regulated by He flow. When the desired pressure was reached, the He flow was switched to bypass and the liquid feedstream pumped into the reactor (Eldex optos 5985-1LMP pump) while the temperature was raised at 5 °C/min up to the reaction temperature. From the Weisz-Prater and Mears criteria it was confirmed that both external and intraparticle mass transfer effects were negligible in our experiments (Table S1, Supporting Information).The product stream was cooled down to 5 °C in a Peltier device around gas-liquid separator. The gas stream was on-line analysed by GC (µGC Agilent, 4 parallel columns MS5A, PPQ, Al2O3-KCl). The gaseous products were quantified by external calibration. The liquid phase product stream was periodically sampled and analysed by either off-line GC-FID (Agilent 6890 N, HP-Wax bonded PEG column) or HPLC-RI (Waters 616, Hi-Plex H column). The liquid products identified were acetaldehyde (MeCHO), acetic acid (AcOH), acetone (ACTN), ethanol (EtOH), methanol (MeOH), ethylene glycol (EG), 1,2-propylene glycol (PG), hydroxyacetone (HA), propanal (EtCHO), propanoic acid (PA), 1-propanol (1-PrOH) and 2-propanol (2-PrOH). Pure reference compounds were used for quantification. The total organic carbon (TOC) was measured off-line on a Shimadzu TOC-L apparatus. The carbon balance was above 90% for all the experiments.The catalytic performance was evaluated based on parameters summarized in Table 1.The main physico-chemical properties of the 0.3Pt/CoAl catalyst are given in Table 2. Both the actual platinum loading and the Co/Al atom ratio were close to the nominal values. Regarding textural characteristics, both the calcined and reduced forms of the solid showed mesoporous nature (Fig. S1, Supporting Information) with isotherms of type IV and H1 hysteresis, both having unimodal pore size distribution. The textural properties of the catalyst barely varied upon reduction (i.e. SBET: 10.3% decrease; dpore: 13.5% increase). The former feature was due to the inherently lower surface area of the metallic Co and Pt, while the latter feature suggested that Pt was mostly deposited into the smallest pores.The 27Al NMR analysis (Fig. S2, Supporting Information) of the support Co/Al exhibited only two peaks at 6.9 and 71.8 ppm, corresponding to aluminium ions in octahedral and tetrahedral symmetry, respectively [47]. In the bare support, strongly prevails the octahedral symmetry (Aloctahedral/Altetrahedral=96/4). After Pt impregnation, a resonance peak around 33 ppm emerged, indicating the presence of penta-coordinated aluminium. This peaks represented about 8% of the total area. Concomitantly, the relative amount of octahedral aluminium decreased to 84%. These findings suggested that Pt ensembles anchored on octahedral sites.The oxidation state and concentration of surface elements of the sample were surveyed by XPS. The Co 2p spectrum of the calcined catalyst ( Fig. 1a) presented the characteristic pattern of cobalt oxide, with the Co 2p3/2 peak at 781.1 eV and a strong shake up feature at 785.1 eV. The 2p3/2-2p1/2 line separation is 15.6 eV. These remarks virtually exclude the presence of Co3+ ions. It is worth pointing out that during XPS analysis the beam emitted may partially reduce the cobalt oxide species. Reduction of the sample by hydrogen at 600 °C gives rise to additional Co 2p3/2 feature at 778.2 eV (Fig. 1b) that could be unambiguously assigned to metallic cobalt [48]. Detailed XPS spectra from Pt 4d and Al 2p levels for calcined and reduced samples are shown in Fig. 1c–d. The Pt 4d5/2 spectra exhibited binding energy (BE) values of 316.8 ± 0.3 eV for the calcined solid and shifted to 314.0 ± 0.3 eV for the reduced solid. According to literature, the latter denotes the presence of fully reduced metallic Pt at the catalyst surface [49]. The Al 2p peak was measured at 74.2 eV for both forms of the solid, either calcined and reduced, indicating that octahedral sites of Al3+ cations were dominant [50].XRD patterns of fresh and reduced solids are displayed in Fig. 2a. The calcined form of the solid showed diffraction peaks consistent with both the standard cobalt oxide (PDF 00-042-1467) and cobalt aluminate (PDF 00-044-0160) spinel structure, in agreement with the support composition. In the reduced form of the solid, additional peaks, characteristic of Co0 in both hcp and fcc phases could be observed. The measured mean crystallite size of the spinel and metallic cobalt were 6.3 and 6.9 nm, respectively (Table 2). The absence of reflections attributable to platinum phases (neither in the calcined nor the reduced forms) suggested that the size domains were below conventional XRD detection limit, and could be ascribed to the low loading and high dispersion of Pt.The H2-TPR profile of the fresh calcined solid (TPRa) is shown in Fig. 2b, and exhibits four reduction peaks. The peak assignation was done according to [45]. The low temperature peak (at 192 °C) was ascribed to the concomitant reduction of PtOX species and free surface Co3+ to Co2+ species promoted by hydrogen spillover over Pt0. The peak at 331 °C was ascribed to the reduction of Co3+ species in close interaction with the support. The intense reduction peak centered at 563 °C was assigned to Co2+ reduction to Co0. Finally, the peak at 763 °C was assigned to the reduction of cobalt ions in the cobalt aluminate (CoAl2O4) phase. In order to further investigate the temperature required for full reduction of both Pt ensembles and Co3+ species not in the aluminate spinel phase, two additional TPR experiments were done consecutively. First, TPRb, where the calcined solid was reduced up to 600 °C and hold for 1 h; subsequently, after cooling down to room temperature, sample was again reduced up to 950 °C (TPRc). The TPRb reduction profile from room temperature to 600 °C was identical to TPRa, and represented around 70% of its hydrogen uptake. The TPRc profile showed a single, broad reduction peak at 780 °C, ascribed to the reduction of the cobalt aluminate spinel. No peaks at below 625 °C were detected. Therefore, it was confirmed that both platinum species and the cobalt as segregated Co3O4 were completely reduced at 600 °C. Based on the H2-TPR results, the catalyst was reduced at 600 °C for 2 h prior to the catalytic runs.H2 pulse chemisorption (Fig. S3, Supporting Information) was carried out to titrate the metallic surfaces. As expected, the catalyst reduced at 600 °C for 1 h showed about 5 times more metallic Co surface than metallic Pt surface (2.01 m2/g vs 0.44 m2/g). These values indicated that only 1.84% of the total surface was due to metals. For platinum, the calculated dispersion was 58% with an average diameter of 2.4 nm, in agreement with the absence of XRD peaks.Ammonia and carbon dioxide TPD experiments revealed the amphoteric character of our spinel based catalyst (Table 2). Its surface was predominantly basic, as basic sites density was two-fold larger than acid sites density. In addition, the basic sites were primarily weak (88% contribution) while the acid sites were medium strength sites (86% contribution) (Fig. S4, Supporting Information). The very low activity in the 33DMB1 isomerization (Table 2) in comparison with other Lewis solids [51] indicated they are predominantly of Lewis-type.Based on the obtained liquid and gaseous products, Reynoso et al. [45] suggested a plausible reaction pathway for the glycerol APR on cobalt aluminate derived catalysts. Reaction pathway consisted on two main routes, which needed both acid and metallic sites ( Scheme 1). In outline, dehydrogenation to glyceraldehyde, preferably on metal sites, which undergoes decarbonylation to produce ethylene glycol, methanol and finally hydrogen. On the other hand, dehydration route, mainly on acid sites, first produces hydroxyacetone and, by subsequent dehydration/hydrogenation, yields C3 liquid products. Further transformation of the liquid products due to C–O bond cleavage leads to the formation of alkanes, which decreases the evolution of hydrogen. In addition, CO can be converted by WGS, increasing H2 yield, or can be hydrogenated (together with CO2) to produce methane and alkanes by Fischer-Tropsch reaction, constituting a hydrogen selectivity challenge.The influence of feedstock concentration on the catalytic reaction was explored at 260 °C and 50 bar, at WHSV of 6.8 h-1 (flowrate: 0.1 mL/min, catalyst mass: 0.9 g) and varying the glycerol concentration (5, 10 and 20 wt%). Fig. 3 shows the effect of glycerol concentration in the feedstock on APR global results after 3 h TOS. The global glycerol conversion was very high (>99%) for all the glycerol concentrations, pointing to very active catalyst for glycerol reforming. Others also reported about the promotional effect of Pt-Co catalysts in the APR reactions and attributed to the PtCo alloying [30,52].Larger differences were obtained in the carbon conversion to gas, which slightly decreased as the concentration of glycerol fed increased (e.g. 41% for lower concentration and 33% for the most concentrate feedstream). This decrease was more pronounced by increasing the glycerol content from 5% to 10%, since by increasing up to 20% the decrease was practically negligible (1.2%). This trend indicated that increasing glycerol concentration, increased the carbon content in the liquid products. Similar results were reported by others [41,53]. For more diluted feedstocks, the availability of the active sites (either metallic and acid/base) increases, thus reactions involved in the APR proceed more extensively to obtain more volatile (gas phase) compounds. Consequently, it can be deduced that feedstocks with low glycerol concentration were more advantageous for gas production (deeper degree of reforming), while more concentrated ones would be preferred for liquid production (i.e. for hydrogenolysis of glycerol by in-situ produced H2) [54,55]. It could also be observed that for the WHSV values used in this study, at glycerol concentrations of 10 wt% or above, there were not enough available active sites to further decompose intermediate molecules, thus reaching almost constant Xgas. This behavior implies that reaction order with respect to glycerol concentration decreased with glycerol concentration.The most important effect of glycerol concentration was on the hydrogen yield, which showed a significant drop from 50.6% to 26.7% when glycerol concentration increased from 5% to 20%. This tendency is in line with the results reported by others [38,41] and was consistent with the above idea, that is, the surface coverage increased with glycerol concentration (i.e. less free sites being available). Moreover, the increase of liquid products yield with glycerol concentration was at the expense of hydrogen consumption, since the yield of products of hydrogenation increased (e.g. 1,2-propylene glycol).Both selectivity to hydrogen and to alkanes showed slight decreasing trend with glycerol concentration, which was also reflected in the almost constant H2/CH4 ratio ( Table 3). High values of hydrogen selectivity (above 85%) were obtained for the three feedstream compositions. Selectivity to alkanes, above 10%, was considerably high in comparison to values reported in the literature for Pt supported on alumina (around 8%) [38], and could be due to cobalt, which is active for CO/CO2 hydrogenation reactions [56].Concerning the gas product distribution (Table 3), an increase in glycerol concentration affected both CO2 and H2 concentration in the opposite way, increasing the former and decreasing the latter. For example, passing from 5% to 20% glycerol concentration, H2 concentration decreased by 11% while CO2 concentration increased by 33%. A decreasing trend for H2 was reported by others [41], and was attributed to the slight increase in the yield of liquid products, being most of them formed though hydrogen consuming reactions. CO content increased with the glycerol concentration, especially at the highest glycerol concentration, due to the lower availability of free metallic centers for WGS. At the reactor outlet, H2/CO2 ratio decreased with glycerol concentration, from 3.7 to 2.5. In all cases, this ratio was above the theoretical (7/3). These results agreed with the decreasing trend of SCH4. The lowest hydrogen concentration in the gas product was 67%, when feeding 20 wt% glycerol/water mixture. CH4 content decreased slightly with the increase of glycerol content. Indeed, H2/CH4 ratio was not almost varied with glycerol concentration, which agree with the almost constant SH2.The H2 production rate (FH2) did not increase in proportion to the increase in glycerol concentration (see Table 3). For instance, when glycerol concentration varied from 5% to 20%, two-fold increase on FH2 was obtained (218 vs 460 µmol/(gcat·min)) when, by stoichiometry, a 4-fold increase would be expected. Opposite trend was shown by the molar flow of hydrogen per mole of converted glycerol (i.e. hydrogen selectivity ratio SRH2, which was limited to 7) which decreased from 3.95 to 1.44 passing from 5% to 20% glycerol feed. These features suggested that H2 lost in hydrogenation/hydrogenolysis reactions increased in higher proportion by increasing the glycerol concentration.Higher concentration of liquid products ( Table 4) was obtained by increasing the glycerol content in the feed stream, in agreement with the decreasing trend of Xgas. An increase in glycerol concentration produced an increment in both hydroxyacetone (HA, primary product from glycerol dehydration) and 1,2-propylene glycol (PG, product from hydroxyacetone hydrogenation) yields. As previously reported for cobalt aluminates catalysts [57], hydrogenation reaction seems to occur more rapidly than dehydration. PG dehydration on acid sites can lead to the formation of acetone or propanal, depending of the primary or secondary hydroxyl elimination by dehydration [58]. Further hydrogenation of both intermediates produce 2-propanol and 1-propanol, respectively. This same route can also form propanoic acid, the main liquid product. Other authors also reported this product in the liquid stream of glycerol APR [39,59]. Among the liquid products, those whose yield was most affected corresponded to propanoic acid, which decreased at expense of the increase of hydroxyacetone and 1,2-propylene glycol, production of the later consuming hydrogen. The ratio of products from dehydrogenation route to products from dehydration route ( Fig. 4) presented a maximum at 10 wt% glycerol. Therefore, it could be concluded that this glycerol concentration provides a balance between this two reaction routes. However, the yield of dehydration-route products had a more pronounced rise with glycerol concentration than those obtained via dehydrogenation-route.Experiments to determine the effect of coupled temperature and pressure variables on the catalyst APR performance were performed varying temperature in the 220–260 °C range, while pressure was established to ensure a liquid-phase reaction mixture (1.8–4.2 bar above the bubble point of the feedstock). This means that isolated effect of pressure was not analysed, but that of the coupled temperature and pressure. The following coupled temperature and pressure pairs were used (°C/bar): 220/25; 235/35; 245/40 and 260/50. The reaction conditions were 10 wt% glycerol concentration and WHSV= 6.8 h-1 (0.1 mL/min of glycerol, 0.9 g of catalyst). The obtained results are shown in Fig. 5.Conversion of glycerol reached almost 100% except for the mildest conditions (Xgly=89%), the later indicating the endothermic characteristics of the reforming reaction [60]. The lowest carbon conversion to gas (21%) was achieved at the mildest operation condition. More severe conditions enhanced carbon conversion to gas and, consequently, decreased the yield of liquid products. Despite same Xgly trend, carbon conversion to gas exhibited a continuous increase with temperature-pressure, reaching a maximum of 43.2% at the most severe conditions (260 °C/50 bar). High temperatures promoted the reforming of glycerol and the intermediate liquids, by promotion of C-C and C-O bonds cleavage to obtain C-containing more volatile compounds [38].The effect of couple temperature and pressure variables strongly affected hydrogen yield, which increased with temperature, in parallel with Xgas. For instance, from 220 °C/25 bar to 260 °C/50 bar an overall increment of 126% was attained by YH2, as due to the endothermic nature of glycerol reforming [60], which favored the glycerol decomposition. The gas products include hydrogen and C-containing compounds (see Table 5). Produced hydrogen could be further reacted giving alkanes in the gas phase and intermediate liquid compounds. Hydrogen lost in alkane formation (SH2) was computed for each run, and the obtained trend is depicted in Fig. 5. For the mildest operation conditions, where 11% of glycerol was unreacted, selectivity to hydrogen was 88%. For the rest of conditions, where glycerol conversion was almost complete, SH2 increased with temperature reaching a maximum of 90% at the most severe conditions. At full glycerol conversion (higher T/P conditions), APR proceeded more extensively, through the reforming of intermediate liquid products, what allowed to obtain more hydrogen. The observed decreasing yield of methane (limited by thermodynamics) indicated less hydrogen consumption, what explained the increasing hydrogen selectivity trend. Similar features for SH2 were obtained by others [19]. Regarding alkane selectivity, it moderately increased with the operation temperature, being methane the most representative of them.Regarding the C-atom amount of the produced alkanes, the majority corresponded to methane, which accounted for 63% at the lowest temperature and 71% at the highest, once again suggesting that C-C scission reaction were promoted by temperature [61]. Table 5 summarizes the gas product composition. Hydrogen was, by far, the most abundant product, with around 70% mole percentage, independent of the reaction conditions. Thanks to the increasing trend in conversion to gas, H2 production rate increased with temperature/pressure. CO2 was the main carbon-containing product, followed by methane (3.5–5.2% range), which was the most abundant alkane. Traces of ethane, ethylene, propane, and butane were also detected (compiled as C2 +). Alkanes were formed by either the subsequent reaction hydrogenation of CO/CO2 and Fischer-Tropsch reactions [62]. The formation of C4 + compounds suggested that Pt-Co catalysts had some activity in C-C coupling reaction, in addition to their recognized great activity in WGS reaction. The latter could be confirmed from the very low CO content in the gaseous product (< 0.2%) for all the temperatures studied.The H2/CO2 ratio was 2.8 and 2.9 in the mildest and the most severe conditions, respectively. These values slightly exceed the theoretical value of 7/3 for glycerol APR, which indicated that glycerol was partially reformed to intermediate species that can readily undergo dehydrogenation reactions while keeping carbon atoms. The large yield of propanoic acid agreed these results.Depending on the applied T/P conditions, around 57–79% of the carbon contained in glycerol came out in the liquid product. As the glycerol conversion reached almost 100% (except at 220 °C/25 bar, with 89%), the production of intermediate oxygenated liquids was considerable. Indeed, the spatial velocity (6.8 h-1) was insufficient for a deep reforming of glycerol molecules. The identified liquid products ( Fig. 6) comprised acids (acetic acid, propanoic acid), ketones (acetone, hydroxyacetone), aldehydes (acetaldehyde, propanal), C3 alcohols (1,2-propylene glycol, 1-propanol, 2-propanol) and C1-C2 alcohols (ethylene glycol, ethanol, methanol). Other peaks detected by chromatography, which accounted less than 5% of all area) could not be identified. The wide variety in the liquid fraction pointed out the complexity of the glycerol APR reaction network and the strong influence of coupled temperature/pressure variable. It must be said that 1,3-propanediol was not obtained in the liquid. Formation of 1,2-propylene glycol and 1,3-propylene glycol is competitive, their selectivity depends on which intermediate, hydroxyacetone or 3-hydroxypropanal, is preferentially produced. The former intermediate is produced by Lewis acid sites [63] while the later requires Brønsted acid sites [64]. The dominant Lewis characteristics of the catalyst (Table 2) explained the absence of 1,3-propylene glycol.Most of the liquid products contained a three-carbon chain. On the other hand, the (C1 + C2)/C3 compounds yields ratio in the liquid stream indicated monotonous increase with the operation temperature (insert in Fig. 6) confirming that temperature promoted C-C cleavage. Most of the liquid products incremented its yield with reaction temperature, being 1,2-propylene glycol, hydroxyacetone and ethylene glycol the exceptions. The former two products resulted from the direct dehydration of glycerol (hydroxyacetone) and its subsequent hydrogenation (1,2-propylene glycol). These results indicated that an increase in reaction temperature favored the dehydrogenation pathway and explained the improvement in hydrogen yield at higher temperatures.The effect of contact time was studied in terms of WHSV (higher WHSV, shorter contact time), varying the flowrate of the feedstream from 0.02 to 0.5 mL/min over 1.8 g of 0.3Pt/CoAl catalyst. The experiments were performed at 260 °C/50 bar with a 10 wt% glycerol in the feedstream. Fig. 7 shows glycerol conversion, carbon conversion to gas, hydrogen yield and selectivity to hydrogen and methane.The effect of WHSV was very noticeable in all the parameters represented, except in Xgly, which remained close to 100% for a wide range of WHSV, only declining to 97% for the highest WHSV studied (17 h-1). These results indicated again the pronounced activity of this catalyst to glycerol decomposition, even working at very high WHSV. Nevertheless, carbon conversion to gas was highly sensitive to contact time: as WHSV increased, carbon conversion to gas decreased. Augmenting feed flowrate from 0.02 to 0.1 mL/min (WHSV = 0.68 and 3.4 h-1, respectively) resulted in a 50% drop in the carbon conversion to gas. Further increase in WHSV resulted in a less severe decay in Xgas. Operation at high WHSV values (i.e. short contact time) hindered consecutive reforming reactions of the intermediate liquid products, thus resulting in less gaseous compounds. Similar trend was reported in the literature [19,41,59]. Interestingly, selectivity to hydrogen increased with WHSV, i.e., the shorter the contact time, less hydrogen was lost in gas phase products. Similar trend was reported by others [30] and was ascribed to a lowered rate of alkanes production. Though the space velocity employed was calculated on the liquid flowrate basis, the gases (H2, CO, CO2) flowed with the liquid stream. Therefore, the SH2 trend suggested that CO/CO2 hydrogenation (producing hydrogen loss in the gas phase) were lessened by the short contact times [65]. Analogous to Xgas, a decrease in hydrogen yield and selectivity to methane could be observed with WHSV, with a concomitant increase on CO and CO2. This was consistent with the increase on hydrogen selectivity. At lower WHSV, the contact time between the intermediate liquids, gases and catalyst was higher, thus enhancing the hydrogen consumption reactions (such as hydrogenation CO/CO2 and hydrogenolysis of the substrate and liquid intermediates) which decreased the hydrogen yield. Regarding SRH2, it decreased with WHSV, passing from 3.89 to 1.27 molH2/molGlyc-converted when WHSV increased from 0.68 to 17 h-1.The CO/H2 molar ratio in the gas stream increased with WHSV ( Fig. 8), i.e. long contact time favored WGS reaction, thus this reaction occurs to a lesser extent at high WHSV. The H2/CO2 ratio indicates the competition between C–C and C–O scission [66]. This value was slightly higher than the theoretical one (7/3) and practically remained above 2.7 regardless of the WHSV used, which indicated that 0.3Pt/CoAl catalyst had a high capacity for C-C bonds breakage prior to C-O bond breakage.The liquid products yields at different WHSV are provided in Fig. 9. As seen before, low WHSV presented an exceptionally high carbon conversion to gas, and therefore, the yield to liquid products was insignificant. Largely, the yield of the liquid products increased with WHSV, in accordance with Xgas decrease. At higher WHSV, a wide variety of liquid products could be distinguished. Among the products obtained, a noticeably high yield of hydroxyacetone and 1,2-propylene glycol could be noted at the shortest contact time (yields of 37.6% and 41.5%, respectively). The yields of both compounds increased in similar way with WHSV, which suggests that at such short contact times, dehydration and hydrogenation reactions predominate at the same rate to the detriment of C-C cleavage. This lower activity for C-C bond scission caused a drop in the yield of C2 products from 13.4% at 6.8 h-1 to 5% at 17 h-1. Yields of propanoic acid, ethanol, acetone and both 1-propanol and 2-propanol reached their maxima at 3.4 and 6.8 h-1, respectively, then decreasing dramatically as WHSV achieved 17.0 h-1. It is evident that at short contact times less fragmentation of the initial molecule takes place. Contrariwise, for the longest contact time (0.68 h-1), it is only possible to detect compounds resulting from several reaction stages such as propanols and acetone.After each reaction at different T/P conditions, the spent catalysts were analysed by analysed by N2 adsorption-desorption isotherms, XRD, ICP-MS and TPH analyses. The nitrogen isotherms and BHJ pore size distribution are shown in Fig. S5, Supporting Information. They still showed type IV isotherms with H1 hysteresis. The most notable differences with respect to the reduced form is that hysteresis loops were narrower in the spent catalysts, though they retained mesoporosity. In the spent catalysts, new generated small pores of around 4 nm contributed to the total pore volume, suggesting the presence of new phases not observed in the fresh reduced form (e.g. gibbsite). As can be seen in Table 6, the specific surface area of the spent catalyst increased when the reaction was performed at the mildest condition (18% increase at 220 °C/25 bar) while decreased for the other conditions tested (14% decrease at the most severe conditions). At lower temperatures, a greater decrease in the average pore diameter was also observed (decreased by 36% at the mildest conditions), which had a tendency to increase with the reaction temperature without reaching the value of its reduced state. XRD diffractograms of the spent catalysts (Fig. S6, Supporting Information) showed peaks from cobalt spinel (either cobalt oxide and cobalt aluminate) and metallic Co, which remained in the spent catalysts. The later suggested that bulk cobalt remained in metallic form. As in the fresh catalysts, no peaks from Pt were visible, indicating that Pt remained highly dispersed. New, sharp peak emerged for the spent catalysts at about 18º (2θ), which could be ascribed to the gibbsite phase (PDF 033-0018). The intensity of the gibbsite peak decreased with the operation temperature/pressure. The same trend observed in the textural properties indicated that gibbsite is surely related to this textural trend.Metal leaching in the liquid product was also investigated by ICP-MS. The results showed that the reaction conditions influenced cobalt leaching. Although the overall results confirmed low metal leaching, cobalt was found to be more leachable at low reaction temperatures (cobalt leaching was 2.30% after reaction at 220 °C vs 0.74% at 235 °C). These apparently inconsistent results agreed with results obtained with catalyst 0.625CoAl after 30 h TOS at 235 °C and 260 °C [57]. A comprehensible explanation could be found in the re-deposition of hydroxylated alumina. At higher temperatures, pH of the reaction medium increased, thus greater leaching was expected. However, in hot water the solubility of the inorganic oxide materials is low, which facilitates the re-deposition of leached alumina, decorating the cobalt and therefore protecting it from leaching [67]. This was in agreement with the gibbsite phase detected by XRD. As for the other metals, both platinum and aluminium had a low percentage of leaching (below 4·10-3% and 0.02%, respectively), understanding that the platinum conferred the stability of the catalyst during reaction. The production of stable catalysts against leaching is a challenge in biomass transformation processes. The most researched strategies to stabilize the supported metal nanoparticles focus on overcoat using techniques such as Atomic Layer Deposition (ALD) [68] or the embedment into support structure via strong metal–support interaction (SMSI) [69–71].The quantification of the carbonaceous deposits was measured by TPH (Temperature-Programmed Hydrogenation). The results obtained by TPH did not show any relation to the reaction condition. The samples with the highest amount of carbonaceous deposits correspond to those for the reactions at 235 and 245 °C. Surprisingly, the catalysts after the reaction at 260 °C had a content of carbonaceous material very similar to that used in the reaction at 220 °C (about 3 times less than at the other conditions). It is worth highlighting the much lower (three to four orders of magnitude) coke deposits in glycerol APR as compared to glycerol steam reforming [72,73]. This was due to the ability of hot compressed water to dissolve carbonaceous deposits [67].0.3Pt/CoAl catalyst, synthesized by impregnation of Pt over cobalt aluminate (nominal Co/Al = 0.625) support, was characterized and tested for glycerol aqueous-phase reforming under various reaction conditions. Specifically, the glycerol concentration in the feedstream, the coupled temperature/pressure variable and the space velocity (in terms of WHSV) process variables were studied. At the conditions studied (260 °C/50 bar), the glycerol conversion did not show significant variation when glycerol concentration was increased from 5 to 20 wt% nor when the space velocity was increased from 0.68 to 6.8 h-1. Only at 220 °C/25 bar (WHSV = 6.8 h-1, 10 wt% glycerol/water) did the glycerol conversion drop below 90%. The highest carbon conversion to gas was achieved at a lower glycerol concentration, at highest temperature/pressure, and at lengthy contact time. As expected, the conditions where the highest conversion to gas was achieved were the ideal ones to obtain higher hydrogen yield.Increasing the glycerol concentration in the feedstream from 5% to 20% (260 °C/50 bar, 10 wt% glycerol/water, WHSV = 6.8 h-1) also showed an increase in the yield of the liquid products formed through the dehydration/hydrogenation of glycerol such as hydroxyacetone and propylene glycol. Conversely, by increasing temperature/pressure from 220 °C/25 bar to 260 °C/50 bar (10 wt% glycerol/water, WHSV = 6.8 h-1) 1,2-propylene glycol yield decreased while ethanol yield increased. As well, higher hydrogen yield was achieved at a higher reaction temperature. On the topic of post-reaction characterization, temperature/pressure conditions undoubtedly affected cobalt leaching. However, further investigation is needed to clearly establish leaching mechanism, which will help to overcome this challenge.Increasing the feed flowrate, and consequently the WHSV, did not change the composition of the outflow gases. Nonetheless, due to the shorter contact time, the production of liquids increased, especially the liquids obtained by the direct reaction of glycerol (hydroxyacetone and 1,2-propylene glycol). A. J. Reynoso: Investigation and Writing – original draft. J.L. Ayastuy: Funding acquisition, Conceptualization, Writing – review & editing. U. Iriarte-Velasco: Formal analysis, Writing – review & editing. M.A. Gutiérrez-Ortiz: Resources, Funding acquisition and Supervision.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This research was supported by grant PID2019-106692EB-I00 funded by MCIN/ AEI/10.13039/501100011033. The authors thank for technical support provided by SGIker of UPV/EHU and European funding (ERDF and ESF).Supplementary data associated with this article can be found in the online version at doi:10.1016/j.jece.2022.107402. Table S1 Supplementary material .

This study examined the influence of process variables (glycerol concentration in feed, coupled temperature/pressure and space velocity) in the catalytic performance in the APR of glycerol over 0.3Pt/CoAl catalyst in a continuous fixed-bed reactor in order to maximize the production of H2. The effect of glycerol concentration in the feed was studied from 5 to 20 wt%, the coupled temperature/pressure varied from 225 °C/25 bar to 260 °C/50 bar and the spatial velocity was changed from 0.68 to 17 h-1. Our results reflected that H2 production was favored at higher reaction temperature/pressure (3.62 vs. 2.49 molH2/molGly-converted, at the most severe and mild conditions, respectively), lower WHSV (3.89 vs. 1.27 molH2/molGly-converted, at the lowest and highest space velocity, respectively) and more diluted feedstocks (3.95 vs. 1.44 molH2/molGly-converted, at the most diluted and concentrated freestreams, respectively). A threshold value at 10 wt% glycerol was found for the ratio of dehydrogenation to dehydration liquid products. The post-reaction catalyst was also characterized by several techniques, showing that Co leaching was the major drawback, especially at the mildest operation conditions, while carbonaceous deposits are negligible.

Because of the growing demand for energy, hydrogen production has evolved as a fossil fuel substitute due to its high heat and lack of environmentally concerning exhaust gases [1–5]. At present, as the application of HER, electrochemical catalysis and photoinitiated water splitting have opened up new avenues to obtain clean hydrogen [6–8]. Traditional electrochemical catalytic methods require noble metal catalysts with high catalytic performance, which are precious and rare. In the development of electrochemical catalysts for the HER, low-cost and earth-abundant catalysts with high catalytic efficiency have emerged to promote hydrogen evolution under water splitting conditions. Transition metal nitrides [9], phosphides [10], carbides [11] and dichalcogenides [12] have been considered promising replacements for traditional platinum-like catalysts, which exhibited a low free energy of H∗ (ΔGH∗) and excellent conductivity.The HER mechanism in an acidic solution can be summarized in two steps: the Volmer reaction ( H 3 + O + e − + C a t . = H a d s + H 2 O ) and the Heyrovsky reaction ( H ads + H 3 + O + e − = H 2 + H 2 O ) [13]. The adsorption/desorption of H∗ on the catalytic surface determines the process of hydrogen evolution. Therefore, the lower the ΔGH∗ is, the higher the catalytic efficiency. Traditional noble catalysts for electrochemical water splitting with a low ΔGH∗ have been widely reported [14–16]. The Pt-like d-band states can facilitate the adsorption/desorption of H∗ to reduce the ΔGH∗. To date, heteroatoms (such as P, S and N) also exhibit a similar activity for the adsorption/desorption of H∗ [17–19]. For example, heteroatoms can induce a change in charge at transition metal atoms on the catalyst surface to form Pt-like d-band states. Xia Long et al. reported a novel catalyst of iron-nickel sulfide with outstanding activity and stability in an acidic solution [20]. The overpotential (OP) was 105 mV at 10 mA cm−2 with a Tafel slope (Ts) of 40 mV dec−1. They investigated the catalytic mechanism with DFT calculations, which indicated that the S atom induced a change in charge of the Fe and Ni atoms on the catalyst surface to reduce the ΔGH∗. The Pt-like d-band states contributed to the lower energy barrier for H+ adsorption and facilitated the HER.Heretofore, P has been widely studied due to its lone pair electrons [21–23]. These paired electrons in the 3p and hollow 3d orbitals can affect H∗ adsorption, thus benefiting the HER [24–26]. Recently, an increasing number of transition metal phosphides have been reported, such as Fe2P, CuxP, CoxP and NixP [22–26]. The traditional and widespread preparation method for phosphides is through high-temperature phosphorization with NaH2PO2 or NH4H2PO2 in a N2 atmosphere [27]. However, phosphides easily aggregate in the formation process, which decreases the activity of catalysts for the HER. To avoid this problem, many studies about different morphologies have been reported: Ni2P microspheres, Ni2P nanoparticles and porous carbon matrix loaded with Ni2P. Dan Ma et al. investigated Ni2P using carbon-based substrates for HER activity [28]. N-doped reduced graphene oxide (N-RGO) was adopted as the substrate for Ni2P nanoparticles and the above construction showed enhanced HER performance. The electron density of Ni was modulated by P and the doped N, and this modulation was beneficial for H∗ adsorption and a low ΔGH∗. The N-RGO substrates alleviated the aggregation of Ni2P and provided a large surface area for active site exposure.Currently, MoS2 has been the most promising candidate for HER catalysts with a similar ΔGH∗ of H∗ adsorption to Pt [29,30]. The d orbitals of Mo are easily induced by the s and p orbitals of adjacent heteroatoms to expand and form Pt-like d orbitals, thus improving HER activity. However, because of poor conductivity and few active sites, pure MoS2 exhibits undesirable catalytic performance. To solve this issue, many efforts have been made to perfect a nanostructure or load other heteroatoms, such as MoS2 with a 2D-layered structure, amorphous MoS2 and heteroatom-doped MoS2 [31–33], which shows excellent HER activity in acidic solutions but demonstrates poor activity in alkaline or neutral solutions. Therefore, developing catalyst with outstanding catalytic performance in alkaline or neutral solutions is a research hotspot. The doping of single Ni atoms can enhance the catalytic performance for the HER in an alkaline or neutral solution. Qi Wang et al. reported a novel catalyst of MoS2 decorated with single atoms of Ni, and this catalyst exhibited a low overpotential and Ts in solutions with a wide pH range [34]. The single Ni atoms were introduced into the MoS2 S-edge and H-sites of the basal plane, which reduced the ΔGH∗ of H∗ adsorption and extended the pH range that can catalyze HER. Because of its excellent conductivity and plentiful defects, RGO has been widely reported as a substrate to enhance the catalytic performance of MoS2. Lin et al. have prepared RGO loaded with Ni-doped MoS2 composite [35]. They found that the doped Ni atoms could facilitate the formation of H∗ and accelerate the adsorption and desorption of H∗ on the catalyst surface. In principle, the doping of Ni atoms into MoS2 could increase the HER activity at different pH values.Recently, heterogeneous interfaces between multicomponent catalysts have been proposed [35,36]. Because of the synergistic effect between the different components and heterogeneous interfaces, the charge distribution is changed, and active sites are formed, which improves catalytic performance. In this work, a novel Ni2P/MoS2 cocatalyst was prepared by using porous N-doped carbon as the substrate (Ni2P/MoS2-CC). The synergistic effect between Ni2P and MoS2 was investigated through theoretical calculations and experimental verification. Raman and XPS spectra proved the presence of synergistic effect between Ni2P and MoS2. The results showed that the cocatalysts had an outstanding catalytic performance for the HER not only in acidic solutions but also in alkaline solutions. The onset potential (OP) values were 280, 350 and 40 mV in acidic, phosphate-buffered saline and alkaline solutions, respectively.Details about the reagents used in this work are shown in the Supporting Information (SI).The detailed preparation route is presented in Scheme 1 . Typically, pectin (1.5 g), nickel nitrate (0.87 g, 3 mmol), ammonium hypophosphite (1.0 g, 12 mmol) and melamine (2.5 g, 20 mmol) were placed into 70 mL of deionized water. Then, the solution was placed into a 150 mL hydrothermal reactor. The temperature was set to 150 °C and kept for 12 h. The resulting material was dried at 105 °C, and then it was carbonized under a N2 atmosphere at 900 °C to obtain Ni2P-loaded N-doped carbon substrates. The aforementioned Ni2P-loaded N-doped carbon substrates (0.1 g) and ammonium tetrathiomolybdate (0.01 g) were added into a 25 mL hydrothermal reactor with 15 mL deionized water, which was then kept at 150 °C for 20 h. The obtained materials were named Ni2P/MoS2-CC catalysts. The CC catalysts, Ni2P-CC catalysts and MoS2-CC catalysts were prepared using the same method without the addition of nickel nitrate, ammonium hypophosphite and ammonium tetrathiomolybdate, respectively.The DFT calculation details, structural details and computational methods are all described in the SI.The details about the preparation of the working electrode are presented in the SI. The used electrolytes were 0.5 mol L−1 H2SO4, 1.0 mol L−1 PBS and 1.0 mol L−1 KOH. The voltages were referenced to a reversible hydrogen electrode (RHE), E (vs. RHE) = E (vs. SCE) + 0.224 V. The cyclic voltammetry (CV) curves were obtained from a potential range of 0.0–0.1 V (vs. RHE) at scan rates of 40, 60, 80, 100, and 140 mV s−1. The electrochemical impedance spectroscopy (EIS) performance was tested from a frequency of 0.01–105 Hz with potentials of 100, 125, 135, 140 and 150 mV and amplitude of 5 mV. A modified Randles equivalent circuit was adopted to fit the Nyquist plots and obtain the electrolyte resistance (Rs) and charge transfer resistance (Rct). The detailed methods for the above tests are presented in the SI.The morphologies and microstructures of the as-prepared materials were observed by scanning electron microscopy (SEM), high resolution transmission electron microscopy (HRTEM) and a surface-area analyzer. The crystal structures of the prepared composites were investigated by X-ray diffractometry (XRD) and Raman spectroscopy. The surface components were obtained from X-ray photoelectron spectroscopy (XPS). The detailed methods for the above characterization procedures are presented in the SI.In order to investigate the morphology of Ni2P/MoS2-CC, scanning electron microscopy (SEM) images were obtained, and the results are presented in Fig. 1 . It can be clearly seen that the Ni2P/MoS2 nanoparticles were uniformly dispersed on the N-doped carbon substrate surface. Compared with Ni2P-CC (Fig. 1a) and MoS2-CC (Fig. 1b), Ni2P/MoS2-CC (Fig. 1c and 1d) exhibited more scalloped edges to provide more loading sited. As shown in Fig. 1c and 1d, a 3D-interconnected structure could be observed, which provided plentiful defects for active sites. The flower-like structure could reduce the agglomeration of MoS2, which improved catalytic performance of the prepared composite. The EDX images (Fig. 1e–1k) corresponded to SEM results indicated the presence of C, N, O, Ni, Mo, P and S, which were uniformly distributed on the carbon substrate surface.As shown in Fig. 2 a and b, uniformly distributed Ni2P/MoS2 nanoparticles were observed on the substrate surface, and the porous structure could be clearly seen. The d spacing 0.22 nm corresponded to the Ni2P (111) crystal lattice, respectively [37]. In addition, the interface between Ni2P and MoS2 could be clearly seen from Fig. 2b, and Ni2P particles were coated with MoS2 sheets. This specific structure provided a synergistic effect that improved HER performance. Moreover, the lattice distance of MoS2 cannot be observed due to its amorphous structure [38,39].The crystal structure of the as-prepared catalysts were investigated by X-ray diffractometry (XRD), and the results are shown in Fig. 3 a. Obviously, the peak at 26° can be attributed to the graphite carbon diffraction (002), indicating that the carbon substrates possessed excellent conductivity for electron transport [37]. The peaks of Ni2P could be easily observed at 2θ values of 41°, 45°, 47°, 54°, 55°, 66° and 75°, which corresponded to (111), (201), (210), (300), (211), (310) and (400) of Ni2P (PDF 01-074-1385). The corresponding peaks of Ni2P and carbon could be seen in CC, Ni2P-CC, MoS2-CC, Ni2P/MoS2 and Ni2P/MoS2-CC composites. In addition, a neglectable shift was observed, indicating that the each other have no effect on the crystal phase. This result was in accordance with the aforementioned analysis. However, the basic peak patterns of MoS2 could not be found, which illustrated that the as-prepared MoS2 had an amorphous structure.Raman spectra were obtained to determine the degree of graphitization and the presence of MoS2, and the results were presented in Fig. 3b and 3c. From Fig. 3b, it can be clearly seen that the characteristic peaks of graphite carbon (G line, 1572 cm−1) and amorphous carbon (D line, 1344 cm−1), and the value of ID/IG indicates the degree of graphitization. The G band in the Raman spectrum of carbon materials is assigned to the stretching bond of sp2 -hybridized carbon. Meanwhile, the D band is attributed to the disorder induced by structural defects and impurities. It could be observed that Ni2P/MoS2-CC exhibited the lowest degree of graphitization (ID/IG = 1.24), which indicated the presence of rich structural defects for the electron transport. The peaks of MoS2 could be clearly observed at 373 and 400 cm−1, which were assigned to E1 2g and A1g of the Mo–S phonon mode, respectively [35]. This result further suggested that Ni2P/MoS2-CC composite was successfully prepared. A shift in the Raman spectra of Ni2P/MoS2-CC could be clearly observed compared with that of MoS2-CC, and this shift was attributed to the presence of the interface between Ni2P and MoS2. The Ni2P and N-doped carbon structure affected the number of layers along the z orientation and improved the uniform distribution of MoS2 [35]. All of the above results enhanced catalytic performance for the HER. Fig. 3d presented the FTIR spectroscopy of Ni2P/MoS2-CC, Ni2P and MoS2. The characteristic peaks of Ni2P and MoS2 could be observed in the Ni2P/MoS2-CC curve, which indicated that Ni2P and MoS2 successfully loaded on the N-doped carbon matrix surface. The presence of C–N (1520 cm−1) and CO (2400 cm−1) were apparent in Ni2P/MoS2-CC, implying the presence of nitrogen and oxygen sources.X-ray photoelectron spectroscopy (XPS) was used to investigate the electronic states of the elements on the surface of Ni2P/MoS2-CC. As shown in Fig. 4 , the peaks appeared at 286, 532, 401, 130, 855, 172 and 233 eV can be attributed to C, O, N, P, Ni, S and Mo elements, respectively [27,35,40]. The peaks at 284.7 and 286.0 eV were assigned to C–C or CC and C–O in Fig. 4b, respectively. As shown in Fig. 4c, the O 1s spectra were divided into two peaks of 531.0 and 532.4 eV, which were attributed to adsorbed oxygen and hydroxyl oxygen, respectively. In addition, the peaks in Fig. 4d were attributed to pyridinic-N (398.7 eV), pyrrolic-N (400.9 eV), oxidized-N (402.8 eV), Ni–N (396.2 eV) and Mo 3p (395.0 eV) [41]. The N dopants that were introduced into the carbon substrates and Ni2P/MoS2 could serve as electron acceptors to improve the catalytic performance for the HER. Three peaks at 129.5, 130.2 and 133.2 eV corresponded to 2p3/2, 2p1/2 and P–O bonding were observed in the P 2p spectra, respectively. The Ni 2p spectra were divided into six peaks: 853.2 eV (2p3/2), 868.4 eV (2p1/2), 872.3 eV (Ni-Ox), 856.7 eV (Ni-Ox), 860.8 eV (satellite) and 879.9 eV (satellite). The presence of Ni-Ox bonding was attributed to the oxidation of surface Ni atoms. A clear negative shift of 0.8 eV could be observed and it was caused by the strong interaction from the MoS2 charge density effect. The peaks located at 162.7, 161.3 and 169.0 eV corresponded to S 2p1/2, S 2p3/2 and SO4 2−, respectively. The Mo 3d spectra were fitted to Mo4+ 3d3/2 (228.9 eV), Mo4+ 3d5/2 (232.6 eV), Mo6+ (232.6 eV) and S 2s (226.1 eV) in Fig. 4h. As shown in Fig. 4g and 4h, a positive shift to a high binding energy could be observed. It is worth noting that the electronic interaction from the adjacent Ni2P could affect the charge distribution of Mo and S. This result demonstrated that the electronic interaction between Ni2P and MoS2 could facilitate the electron transport from MoS2 to Ni2P, which enhanced the activity of catalysts for the HER.The catalytic performance for the HER was studied by a three-electrode system in 0.5 mol L−1 H2SO4 solution, 1.0 mol L−1 PBS solution and 1.0 mol L−1 KOH solution, respectively. The corresponding result curves of the as-prepared catalysts are presented in Fig. 5 . The onset potential (Eonset), Cdl and OP values at a current of 10 mA cm−2 are listed in Table 1 . Additionally, commercial Pt/C was also tested as the reference in this work. The commercial Pt/C exhibited an outstanding performance with a negligible Eonset, and the OP at a current of 10 mA cm−2 was 10 mV in the 1.0 mol L−1 KOH electrolyte (Fig. 5a). As shown in Fig. 5a, the pure N-doped carbon substrates showed negligible activity for the HER. Compared with Ni2P-CC, MoS2-CC composite and Ni2P/MoS2, the prepared Ni2P/MoS2-CC composite exhibited a better performance with a lower OP of 170 mV (vs. RHE) at a current of 10 mA cm−2 (Table 1), which can be attributed to the synergistic effect between Ni2P and MoS2 at their interface and the N-doped carbon matrix. As previously reported, the use of only MoS2 revealed a good catalytic performance for the HER [42,43]. Clearly, the addition of Ni2P enhanced the activity of catalysts in alkaline and neutral solutions. It is worth noting that the rate-determining step in 1.0 mol L−1 KOH was the Volmer reaction ( H 2 O + e − → H ads + OH − ) . For Ni2P, the Ni atom could play a hydroxide-acceptor role, and the P atom could play a proton-acceptor role, which was attributed to the enhancement of the unfilled d-orbital in Niδ+ caused by MoS2. The stronger the unfilled d-orbital in Niδ+ is, the higher the OH− adsorption ability. As presented in Fig. 5b, the as-obtained catalyst also exhibited outstanding catalytic activity for the HER in acidic and neutral solutions with Eonset values of 280 and 310 mV, respectively. The aforementioned results indicated that the as-prepared catalyst had a wide pH range for the catalysis of the HER. Compared with previous reports (Table 2 ), the obtained Ni2P/MoS2-CC composite exhibited a better catalytic performance, indicating that this method opened a novel avenue to prepare catalysts for the HER [44–51].To investigate the reaction kinetics for the HER process, the Ts was obtained from the equation η = b l o g ( j ) + a (where b represents the Ts), and the plots are presented in Fig. 5c and 5d. The rate-determining step of the Volmer reaction in 1.0 mol L−1 KOH and 0.5 mol L−1 H2SO4 was the OH− production ( H 2 O + e − + Cat. → H ads + OH − ) and the H+ adsorption ( H 3 + O + e − + Cat. → H ads + H 2 O ) , respectively. As shown in Fig. 5c and 5d and Table 1, Ni2P/MoS2-CC possessed lower Ts than the others, which indicated that Ni2P/MoS2-CC had favorable kinetics for H2 evolution. The synergistic effect between Ni2P and MoS2 at their interface introduced the charge density distribution of the d-orbital, which facilitated OH− adsorption. It should be noted that Ts values of 95 and 75 mV dec−1 in alkaline and acidic solutions were lower than 120 mV dec−1, which suggested that the reaction mechanism was the Volmer-Heyrovsky mechanism [17,22].In order to investigate the intrinsic catalytic performance of Ni2P/MoS2-CC composite for HER, the kinetic current density for hydrogen evolution/oxidation reactions (HER/HOR) was fitted with simplified Butler–Volmer equation: j k = j 0 ( e α F R T η − e 1 − α R T η ) where j 0 represents the exchange current density of intrinsic activity, α represents the transfer coefficient regarding to the symmetry of the HER/HOR. F, R, and T are Faraday's constant (96,485 C mol−1), the universal gas constant (8.314 J mol−1 K−1) and the temperature (around 293 K), respectively. η represents the applied overpotential (V). It could be clearly seen from Fig. 5f that the fitted curves of Ni2P/MoS2-CC (0.76 mA cm−2) showed a larger exchange current density than Pt/C (0.34 mA cm−2), indicating that Ni2P/MoS2-CC had a faster HER kinetics. These results can be attributed to the H∗ adsorption/desorption energy on the active sites [52].To estimate the intrinsic catalytic activity, the electrochemical active surface areas (ECSA) of the catalysts were obtained through the calculation of the double-layer capacitance (Cdl) according to the cyclic voltammetry (CV) curves (shown in SI) at different scan rates, and the results were shown in Fig. 5e. As presented in Fig. 5e, the prepared Ni2P/MoS2-CC composite had a superior Cdl, up to 19 mF cm−2, which was caused by the synergistic effect between Ni2P and MoS2 at their interface and the substrate nanostructure. Additionally, Ni2P/MoS2-CC composite possessed an outstanding specific surface area of 390 m2 g−1 with a pore size distribution of 4 nm (shown in the SI), which was beneficial for the transport of electrons and Hads. As shown in Fig. 5f, 5g and 5h, Ni2P/MoS2-CC exhibited a higher TOF values and exchange current density, which also indicates that Ni2P/MoS2-CC composite has higher intrinsic activity [53].Electrochemical impedance spectroscopy (EIS) was conducted to study the kinetics of the HER, and the corresponding results are presented in Fig. 6 . The values of electrochemical impedance are listed in Table 3 . As shown in Fig. 6a and 6b and Table 3, the catalysts had small solution impedance. It can be observed that Ni2P/MoS2-CC composite exhibited a smaller semicircle in Fig. 6a, which suggested a lower Rct value (87.2 Ω) and indicated that this catalyst possessed a higher activity for the HER than the others. The Rct value of Ni2P/MoS2-CC composite (87.2 Ω) was smaller than CC (591.6 Ω), Ni2P-CC (192.1 Ω) and MoS2-CC (490.2 Ω), indicating that the prepared Ni2P/MoS2-CC composite had high interfacial charge transfer efficiency and dynamic velocity. This result was attributed to the porous structure and the synergistic effect between Ni2P and MoS2 at their interface, which accelerated the transport of electrons and Hads [54,55]. The conductivity of the catalyst was improved by the addition of the N-doped carbon substrate and Ni2P. Compared with the other potentials, the Rct exhibited a clear change. As shown in Fig. 6b, the semicircle gradually decreased as the overpotential decreased, indicating that a high overpotential was beneficial for the HER [56,57]. The experimental data were well fitted with the Randles equivalent circuit (Fig. 6c). The Tafel slope calculated from the electrochemical impedance was 93 mV dec−1, which was similar to the value obtained from the LSV curves. This result reflected outstanding electrode kinetics of the as-prepared catalyst.To describe the durability of Ni2P/MoS2-CC, the LSV curves before and after 10,000 cycles and the voltage-time response were obtained, and the results are shown in Fig. 7 . Meanwhile, the structure of Ni2P/MoS2-CC composite after the stability test was also investigated. It can be observed that a negligible change occurred in Fig. 7a and 7b, which indicated that the catalyst maintained good stability after a long-term experimental test. The characteristic peaks of Ni2P were observed in the XRD patterns, as shown in Fig. 7c, which corresponded to the standard Ni2P. It should be noted that the peaks of MoS2 located at 373 and 400 cm−1 appeared in the Raman spectra, indicating that MoS2 was definitely present on the substrate. There was no change in the structure after analyzing the SEM images, and the porous structure was effectively retained. The aforementioned analysis proved that the catalyst had outstanding durability, which was attributed to the specific substrate structure.To shed more light on the synergistic effect between the Ni2P and MoS2 at their interface, DFT was adopted to calculate the Gibbs free energy (ΔGH∗), and the crystal structures of different catalysts with different H∗ adsorption sites are shown in Fig. 8 and Fig. 2S. The perfect activity for the HER is a zero value for ΔGH∗, which is caused by the ΔG offset from a proton reduction and ΔG from the removal of adsorbed hydrogen [56,57]. As shown in Fig. 8d, the H∗ adsorption sites of Mo and S have high absolute (ΔGH∗) values. Clearly, the introduction of Ni2P into the MoS2 structure improved HER activity. The absolute ΔGH∗ value of Ni2P or MoS2 was greatly reduced. The minimum value for S at the adsorption sites of Ni2P/MoS2 was 0.10 eV, which was much closer to zero, indicating that Ni2P/MoS2 has good HER performance. These results proved that the synergistic effect between Ni2P and MoS2 at their interface could improve the catalytic performance for the HER, which was also in accordance with the experimental results, demonstrating wide applications in composite materials [58–65].In summary, a Ni2P/MoS2-CC synergistic catalyst was prepared through a novel method. The synergistic interface between Ni2P and MoS2 greatly reduced the free energy barrier and improved the catalytic performance for the HER. The addition of N-doped carbon substrates and Ni2P provided excellent conductivity for electron transport, and the porous structure prevented the aggregation of MoS2 while also provided more passageways for the desorption of the adsorbed H∗ intermediate. Compared with Ni2P and MoS2 catalysts, the prepared Ni2P/MoS2-CC cocatalyst exhibited obvious improvements with a wide pH range for the HER. In addition, Ni2P/MoS2-CC cocatalyst showed low OP values of 280, 350 and 40 mV in acidic, phosphate-buffered saline and alkaline solutions, respectively, and the corresponding Ts values were 75, 121 and 95 mV dec−1, respectively. The DFT results revealed that the interface between Ni2P and MoS2 decreased the absolute (ΔGH∗) value to accelerate proton/electron transfer. This work has opened a new avenue to prepare a cocatalyst with a specific interfacial structure that facilitates the HER process.We declare that we have no financial and personal relationships with other people or organizations that can inappropriately influence our work, there is no professional or other personal interest of any nature or kind in any product, service and/or company that could be construed.The work described has not been submitted elsewhere for publication, in whole or in part, and all the authors listed have approved the manuscript that is enclosed.We greatly appreciate the financial support of the National Natural Science Foundation of China (No. 21872119, 22072127), the Natural Science Foundation of Hebei Province (No. B2021203016), the Science and Technology Project of Hebei Education Department (No. ZD2022147), and the Special Project for Local Science and Technology Development Guided by the Central Government of China (No. 216Z1301G).The following is the Supplementary data to this article:Materials, characterization, DFT calculation, preparation of the working electrode. Fig. 1S(a–d) CV curves of different catalysts; (e) N2 adsorption–desorption curve; (f) Pore size distribution curve. Fig. 2S. (a) Crystal structures of MoS2 using S as the adsorption sites; (b) crystal structures of MoS2 using Mo as the adsorption sites; (c) crystal structures of Ni2P using Ni as the adsorption sites; (d) crystal structures of Ni2P using P as the adsorption sites; (e) crystal structures of Ni2P/MoS2 using P as the adsorption sites. Multimedia component 1 Multimedia component 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.gee.2020.12.008.

Electrochemical catalysts for the hydrogen evolution reaction (HER) have attracted increasing attentions. Noble metal-free cocatalysts play a vital role in HER applications. Herein, a novel strategy to prepare a Ni2P/MoS2 cocatalyst through a simple hydrothermal-phosphorization method was reported, and the prepared cocatalyst was then loaded on an N-doped carbon substrate with excellent conductive performance. The large surface area of the carbon substrate provided many active sites, and the interface between Ni2P and MoS2 improved the catalytic performance for the HER. Compared with pure Ni2P catalyst and MoS2 catalyst, the prepared Ni2P/MoS2 cocatalyst exhibited enhanced catalytic performance. In addition, the results indicate that the prepared cocatalyst has a wide pH range and low onset potential values of 280, 350 and 40 mV in acidic, phosphate-buffered saline and alkaline solutions, respectively, and the corresponding Tafel slopes are 75, 121 and 95 mV dec−1, respectively. Density functional theory (DFT) was adopted to calculate the hydrogen adsorption free energy (ΔGH∗). The results showed that the interface between Ni2P and MoS2 reduced ΔGH∗, which was beneficial to the adsorption of hydrogen. Present preparation of cocatalysts with unique interfaces provides a new strategy for improving the catalytic performance of HER.

Data will be made available on request.Hydrogen is the green energy with the greatest potential in the 21st century. The hydrogen produced industrially in the water-to-gas conversion processes or steam reforming processes contains other impurities, including CO2, N2 and CO, etc. [1]. Therefore, it is required to purify or separate hydrogen. For this purpose, membrane separation technology with low energy consumption and a high degree of automation can be utilized [2]. The most often used inorganic membranes for gas separation are palladium metal membranes, microporous ceramic membranes, and carbon molecular sieve membranes [3–5]. Previous researches have demonstrated that microporous silica membranes offer remarkable gas separation performances. The most common preparation processes are chemical vapor deposition (CVD) and sol-gel techniques [6,7]. The sol-gel method is a promising candidate approach with favorable properties, such as large surface areas and superior gas separation performances.However, the primary issues with pure silica membrane include undesirable gas permeance and selectivity, as well as insufficient hydrothermal stability when exposed to water vapor environments [8]. An increase in selectivity is commonly at the expense of a decrease in membrane permeance. In general, it is quite difficult for one single material to overcome the contradiction between permeance and selectivity, so membranes need to be modified. In silica structures, the hydrolysis reactions of siloxane (SiOSi) bonds and water molecules result in the fracturing of SiOSi bonds and the formation of new moveable silanol (Si-OH) groups. Microporous structures become dense or even collapse as a result of the rearrangement of Si-OH groups and the continuation of condensation reactions. This circumstance frequently causes confusion in industrial gas treatment. Therefore, silica membranes used in industry must have greater hydrothermal stabilities in order to maintain relatively stable gas permeances and desirable separation performances throughout their operational lifetime. There are two methods for enhancing the hydrothermal stabilities of pure silica membranes. One strategy is to introduce hydrophobic groups to reduce the Si-OH concentrations on the membrane surfaces, hence decreasing the physical adsorption of water molecules, including alkylamine, methyl and phenyl, etc. [9–14]. The other is to dope transition metals/metal oxides into the silica networks, such as alumina, nickel, cobalt, tantalum, magnesium, niobium, titanium and zirconium, etc. [15–24]. This approach effectively prevents the densification of membranes following high-temperature treatment, thereby enhancing the hydrothermal stability and reproducibility of membranes. This is because oxygen atoms and transition metal atoms generate more stable covalent connections than SiO bonds, so structures of metal-doped silica membranes are more stable than those of the SiO2 membrane [8]. Numerous studies have demonstrated that this method has a vast array of positive effects.Nickel and cobalt elements have been proven to be promising metal nanoparticles in the field of gas separation. Nickel is a relatively inexpensive transition metal with high hydrogen affinity. Kanezashi et al. [16] investigated nickel-doped silica membranes with various nickel contents (Si/Ni = 4/1–1/1). The steady permeances of He and H2 for the nickel-doped silica membrane (Si/Ni = 2/1) at 500 °C and 90 kPa were 1.6 × 10−5 and 4.6 × 10−6 mol m−2 Pa−1 s−1, respectively, and the He/N2 and H2/N2 permselectivities were 1450 and 400, respectively. In addition, the high hydrothermal stability of cobalt-doped silica membranes in mixed air streams has the curiosity of academic researchers. According to Uhlmann et al. [17], the He permeance was 9.5 × 10−8 mol m−2 Pa−1 s−1 with an optimum activation energy of 15 kJ mol−1 of cobalt-doped silica membrane. The He/N2 permselectivity rose from 350 to 1100 under dry gas conditions after exposure to water vapor. Liu et al. [18] synthesized cobalt-doped silica membranes at 500 °C with a He/CO2 permselectivity of 479 and a He permeance of 3.3 × 10−7 mol m−2 Pa−1 s−1. After hydrothermal treatment, the He and H2 permeances decreased by 28 % and 22 %, respectively, and the He/CO2 permselectivity dropped to 190.The problems with pure SiO2 membrane applications are their undesirable H2/CO2 separation performance and poor hydrothermal stability in hydrothermal environment. In this investigation, hydrophobic group modification and metal doping are intended to solve the problems. MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) materials were synthesized using the sol-gel method, and membranes were fabricated using the coating process under N2 atmosphere. The physicochemical properties and microscopic morphologies of the materials and membranes were systematically characterized by Fourier transform infrared spectroscopy (FTIR), X-ray diffraction (XRD), transmission electron microscope (TEM), N2 adsorption-desorption measurements and scanning electron microscope (SEM). Subsequently, the H2 permeances, H2/CO2 permselectivities and hydrothermal stabilities of the membranes were discussed in detail. To our knowledge, this is the first study to investigate the effects of methyl groups and nickel-cobalt doping on gas separation performances and hydrothermal stabilities of silica membranes, with the hope of inspiring future researches on binary metal-doped silica membranes.The samples were prepared by the sol-gel method using tetraethylorthosilicate (TEOS, purchased from Xi’an chemical reagent Co., Ltd., Xi’an, China) and methyltriethoxysilane (MTES, purchased from Hangzhou Guibao Chemical Co., Ltd., Hangzhou, China) as silica sources, nickel nitrate hexahydrogen (Ni(NO3)2·6H2O, purchased from Tianjin Fuchen Chemical Reagent Co., Ltd., Tianjin, China) as nickel sources, cobaltous nitrate hexahydrate (Co(NO3)2·6H2O, purchased from Tianjin Fuchen Chemical Reagent Co., Ltd., Tianjin, China) as cobalt sources, absolute ethanol (EtOH, purchased from Tianjin Branch Micro-Europe Chemical Reagent Co., Ltd., Tianjin, China) as organic solvents, nitric acid (HNO3, purchased from Sichuan Xilong Reagent Co., Ltd., Chengdu, China) as catalysts and deionized water prepared in the lab. In our earlier research, Co/MSiO2 membranes with n Co/n TEOS = 0.08 exhibited optimal gas permselectivity and hydrothermal stability [25]. In this study, Ni-Co/MSiO2 materials and membranes were prepared with (n Ni + n Co)/n TEOS = 0.08. The molar ratio of TEOS/MTES/Ni(NO3)2·6H2O/Co(NO3)2·6H2O/EtOH/HNO3/H2O was 1/0.8/x/0.08-x/8.5/0.085/6.8 (n Ni and n Co were denoted by x and 0.08-x, respectively). The contents of n Ni (x) and n Co (0.08-x) in the samples for this investigation are summarized in Table 1.The TEOS, MTES and Co(NO3)2·6H2O solutions were completely dissolved in an ethanol solution, then placed in an ice-water bath, and strongly stirred for 0.5 h. Subsequently, a mixture of Ni(NO3)2·6H2O solution, deionized water and HNO3 solution was added to the aforementioned solution. The reaction mixtures were then heated and refluxed in a water bath at 60 °C for 3 h, then cooled to 25 °C (heating and cooling rate 0.5 °C min−1) to obtain the MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) sols [20].Sols of MSiO2 and NixCo0.08−x/MSiO2 were vacuum-dried at 30 °C to produce gels. The gels were crushed into powders and calcined at 400 °C under N2 atmosphere for 2 h in a tube furnace (heating rate of 0.5 °C min−1) to get the final unsupported MSiO2 and NixCo0.08−x/MSiO2 (x = 0.024, 0.04, 0.056, and 0.08) materials [26].The microporous α-Al2O3 composite disks supports (dimensions of Φ20 mm × 1.5 mm, purchased from Yixing Damai Ceramic Technology Research Institute, China) were lightly sanded with 600 and 1200 grit sandpaper, washed with water and ethanol to remove inorganic and organic contaminants, and dried overnight at 70 °C. The α-Al2O3 supports were submerged in sols for 1 min to ensure that the sols were adequately absorbed into the pores and were withdrawn at a pace of 10 cm min−1. The samples were dried at 30 °C for 3 h and calcined in a tubular furnace at 400 °C under N2 atmosphere for 2 h (heating rate 0.5 °C min−1). To minimize any defects caused by dust in the air, the deep coating process was repeated 4 times to obtain the final MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) membranes [27]. Fig. 1 depicts the schematic diagram of preparation process for MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) sols/materials/membranes.The as-prepared MSiO2 and NixCo0.08−x/MSiO2 membranes were exposed to saturated steam in an incubator at 200 °C and 75 % RH for 288 h to obtain the steam-treated MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) membranes. Next, MSiO2 and NixCo0.08−x/MSiO2 membranes were calcined at 400 °C in a tubular furnace under N2 atmosphere for 2 h (heating rate 0.5 °C min−1) to obtain regenerated MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) membranes [8]. The gas permeances of steam-treated and regenerated MSiO2 and NixCo0.08−x/MSiO2 membranes were investigated, respectively.Fourier transform infrared spectroscopy (FTIR) was performed on a Nicolet 5700 instrument to measure the functional groups of materials in the wavelength measurement range of 400–4000 cm−1 using the KBr compression method. X-ray diffraction (XRD) was performed on a Rigaku D/max 2200 X-ray diffractometer to determine the chemical compositions of materials via CuKα radiation (40 kV, 40 mA), and the scanning range was 4–90° with the speed of 8°/min. The crystallizations of materials were analyzed by transmission electron microscopy (TEM) using a JEOL-JEM 2100F instrument. The BET surface areas and pore volumes of materials were determined by the N2 sorption-desorption isotherms using an ASAP 2020 instrument. The morphologies of membrane surfaces and cross-sections were analyzed by scanning electron microscope (SEM) using a Hitachi JEOL-JSM-6300 instrument with a 5 kV acceleration voltage.The schematic diagram of the experimental setups of the gas dead-end permeance system is performed in Fig. 2. The membranes were sealed in a stainless steel module using O-rings. Testing could begin after ventilation achieved a stable level (at least 3 h). Single gas permeance of MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) membranes were examined at various pressures and temperatures. The internal and external pressure difference (0.1–0.4 MPa) and temperature (25–200 °C) were adjusted to meet the test parameters. H2 (0.289 nm) and CO2 (0.33 nm) with high-purity (99.99 %) were used as permeance gases. Each value for each membrane was based on three membranes. Each membrane was measured three times under the same condition. According to the nine values, the average value and standard deviation of each membrane were calculated. The values displayed in the data analysis plot were the average values, and the standard deviations were represented as the error bars. The gas permeance tests followed the standard dead-end volume procedure, in which the data were obtained from the end bubble flow meter and recorded when the equilibrium state was reached (2 h after gas stabilization). The ratio of H2 permeance to CO2 permeance determined the permselectivity of H2/CO2, known as the ideal selectivity. The gas permeance is represented by Eq. (1) and the permselectivity is expressed by Eq. (2): (1) F = Q × P A × ∆ P × T × R (2) α = F a F b where F is the gas permeance (mol m−2 Pa−1 s−1), Q is the gas flow through the flow meter (m3 s−1), P is the standard atmospheric pressure in Xi’an, China (Pa, 1.05 × 105), A is the effective area of the membrane (m2), ∆ P is the pressure difference across the membrane (Pa), T is the absolute temperature (K) and R is the gas constant (J mol−1 K−1), and α is the permselectivity.For steam-treated and regenerated MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) membranes, the H2 and CO2 permeances were determined by repeating the aforementioned procedures, respectively, and the H2/CO2 permselectivity values were calculated using the same approach, respectively.The infrared spectra curves of unsupported MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) materials are shown in Fig. 3. The bending vibration peak of water molecules is displayed at the absorption peak of 1645 cm−1 [28]. The absorption peak at 3450 cm−1 is attributed to the stretching vibration of -OH groups in structurally coordinated water and physically adsorbed water. The stretching vibration peak of -CH3 groups and the characteristic absorption peak of Si-CH3 groups for TEOS and MTES appear at 2978 cm−1 and 1276 cm−1, respectively, demonstrating that the hydrophobic -CH3 groups successfully branch onto the Si atoms. The absorption peaks at 790 and 443 cm−1 are considered as symmetrical stretching vibration of the OSiO bonds and the bending vibration of the SiO Si bonds, respectively. In the MSiO2 material, the absorption peak at 1050 cm−1 accompanied by a shoulder is ascribed to the asymmetric stretching vibration of the SiOSi bonds, confirming that the sol-gel process has been carried out satisfactorily [29]. The absorption peaks at approximately 1031, 1012 and 1023 cm−1 are assigned to the vibration of the SiOSi bonds for Ni/MSiO2, Co/MSiO2 and Ni0.024Co0.056/MSiO2 materials, respectively. This is because the introduction of nickel or/and cobalt elements can destroy the symmetry of SiO2 and lead to the shift of the SiOSi bond position. In addition, there is an absorption peak at approximately 960 cm−1 in nickel or/and cobalt-doped silica materials, but not in the MSiO2 materials. Metals can be covalently attached to siloxanes to form SiOM bonds, thereby forming a denser network structure [30,31]. Therefore, it can be speculated that the absorption peak is the SiOM (M = Ni, Co) bonds. The nickel and cobalt elements have been successfully integrated into the MSiO2 network to form Si-O-Ni or/and SiOCo bonds, which is conducive to overcoming the problem of the hydrolysis reaction of SiOSi bonds with water molecules under water vapor conditions. This minimizes the possibility of structural collapse after regeneration and sets the groundwork for enhancing the hydrothermal stability and repeatability of silica membranes.The XRD patterns of unsupported MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) materials are illustrated in Fig. 4. The distinct diffraction peak at around 2θ = 23° can be assigned to the typical amorphous SiO2 phase of all samples [32,33], suggesting that methyl modification, nickel or/and cobalt doping does not change the phase structure of silica particles. The doping of nickel or/and cobalt elements reduces the SiO2 peak intensity. This is because some silica atoms are replaced by the introduced nickel or/and cobalt elements to form Si-O-Ni or/and SiOCo bonds, resulting in the decrease of SiOSi bonds. In addition to the amorphous SiO2 phase, another absorption peak in the Ni/MSiO2 material is detected at 2θ = 43.31°, corresponding to the (200) crystalline plane of NiO (JCPDS no. 47-1049). The other diffraction peaks in the Co/MSiO2 and Ni0.24Co0.056/MSiO2 materials appear at 2θ = 42.49° and 61.64°, corresponding to the (200) and (220) crystalline planes of CoO (JCPDS no. 70-2855). The peaks of NiO and CoO are not pronounced due to the low concentrations of nickel and cobalt elements. The full-width at half maxima of the characteristic reflection with the highest intensity (200) is used to calculate the mean crystallite size with the aid of making use of the Scherrer equation [34], as indicated in Eq. (3): (3) D = k λ β cos θ where D is the size of NiO/CoO crystallites, k is the constant value of Scherrer (0.89), λ is the wavelength of X-ray source (0.154 nm), β is the full width at half maximum intensity, and θ is the Bragg angle.The mean size of NiO crystal in the Ni/MSiO2 material is calculated as 4.5 nm, and the mean size of CoO crystals in the Co/MSiO2 and Ni0.24Co0.056/MSiO2 materials is calculated as 4.8 nm. The synthesis of NiO and CoO is advantageous for facilitating the surface diffusion of H2 molecules, and has positive effects on improving the H2 permeances and separation performances. Fig. 5 depicts photos of unsupported MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) materials. In nickel or/and cobalt-doped silica materials, the color gradually becomes lighter as the cobalt content decreases and the nickel content increases. They are blue-violet, blue-gray, cyan-gray, gray-green and pale-yellow-green, respectively. It can be inferred that Co(NO3)2·6H2O and Ni(NO3)2·6H2O solutions are successfully thermally decomposed into CoO and NiO during the calcination processes (black-gray for CoO and yellow-green for NiO), which is consistent with the XRD analysis (Fig. 4).The TEM images of unsupported MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) materials are presented in Fig. 6. The darker-contrast particles are attributed to the metal oxides due to the differences in electronic density. The presence of different-sized particles in the image indicates that the oxides of nickel and cobalt have been effectively doped into the amorphous silica matrix and are irregularly distributed as single-dispersed particles. The lighter-contrast structures are attributed to the silica carriers, which exhibit an amorphous three-dimensional mesh structure. The metal-doped silica materials have more dispersed three-dimensional structures, which is more advantageous to reducing the densification of the MSiO2 material.The N2 adsorption-desorption isotherms and pore size distributions of unsupported MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) materials are demonstrated in Fig. 7. The N2 adsorption capacity rises dramatically at lower relative pressures (P/P0 < 0.1), which is caused by the strong adsorption potential energies in the pores. Therefore, it is speculated that there are a large number of micropores in the materials. Subsequently, as the relative pressure continues to increase, the amounts of N2 adsorption capacity is still slowly increasing. Except for single-layer adsorption, the pore structures contain multi-layer adsorption and even capillary condensation. This adsorption type corresponds to the Type I adsorption isotherm of the IUPAC classification. These samples possess the same characteristics as microporous materials. Notably, the adsorption isotherm of the MSiO2 material exhibits hysteresis and the loop is not closed. This may be because the MSiO2 material collapsing in liquid nitrogen at low temperatures and N2 molecules becoming trapped in micropores, resulting in non-overlapping adsorption-desorption curves. This phenomenon does not occur in other materials, because the doping of nickel or/and cobalt elements activates the pore size, resulting in larger microporous volumes, thereby forming more stable silica frameworks [35]. The pore size distribution curves in Fig. 7b are obtained by Horvath-Kawazoe (HK) and Barrett-Joyner-Halenda (BJH) methods. The pore size distributions tend to widen due to the doping of nickel and cobalt elements. The pore size distribution of the MSiO2 material ranges from 1.5 Å to 10 Å, and that of nickel or/and cobalt-doped silica materials ranges from 1.5 Å to 15 Å.The pore structure parameters of unsupported MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) materials are listed in Table 2. The doping of nickel or/and cobalt enlarges the BET surface area, total pore volume and average pore size of the MSiO2 material, which is because the bond lengths of NiO or/and CoO bonds is longer than those of the SiO bonds (the bond lengths of NiO, CoO and SiO are approximately 2.1, 2.3 and 1.64 Å, respectively). For membranes with gas-selective permeance, a larger total pore volume and a smaller average pore size can guarantee that the membrane works as an effective sieve, allowing for greater gas permeances while preventing the passage of larger gas molecules. The Ni0.24Co0.056/MSiO2 membrane is more beneficial as an effective sieve for gas separation. The optimized molecular structure models based on molecular dynamics simulations of MSiO2, Ni/MSiO2, Co/MSiO2 and Ni0.24Co0.056/MSiO2 materials are shown in Fig. 8. The molecular structure distribution of the Ni0.024Co0.056/MSiO2 material is more uniform and tighter, which lays a foundation for the membrane to achieve superior gas separation and hydrothermal stability. Fig. 9 illustrates SEM images of surfaces and cross-sections for MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) membranes. The surface coverages are relatively complete without visible pinholes and cracks, and the surface particles are relatively homogeneous. Compared with the MSiO2 membrane, a small number of large particles are observed on the surfaces of the nickel or/and cobalt-doped silica membranes due to the formation of NiO or/and CoO. The membrane cross-section shows an asymmetric configuration and can be divided into two layers. The upper part is the selective layer, which is the medium of gas permselectivity. The thickness of selective layers for MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) membranes are approximately 2.3–2.8 µm. The α-Al2O3 support layer provides mechanical strength as the membrane system’s foundation. The membrane cross-sections reveal irregular structures due to the shapes of the α-Al2O3 support bodies. Overall, the porous α-Al2O3 supports are successfully loaded with a selective layer for gas separation. Fig. 10 displays influence of pressure difference on H2 permeances and H2/CO2 permselectivities for MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) membranes at 25 °C. The H2 permeances of MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) membranes at 0.3 MPa and 25 °C are 2.4 × 10−7–6.4 × 10−7 mol m−2 Pa−1 s−1 and the H2/CO2 permselectivities are 19.2–56.8. With the increase of differential pressure, the H2 permeance and H2/CO2 permselectivity of the MSiO2 membrane is basically insignificant, but those of metal-doped silica membranes marginally increase. The H2 permeances of MSiO2, Ni/MSiO2, Co/MSiO2 and Ni0.24Co0.056/MSiO2 membranes at 0.3 MPa and 25 ℃ rise by 0.3 %, 1.0 %, 1.4 % and 2.0 %, and the H2/CO2 permselectivities increase by 2.0 %, 3.4 %, 4.3 % and 2.2 %, respectively, compared with those at 0.1 MPa and 25 °C. The rise in pressure difference facilitates the adsorption and transport of H2 molecules, so the H2 permeances and H2/CO2 permselectivities of silica membranes are improved at a relatively high differential pressure (0.3 MPa).The influence of temperature on H2 permeances and H2/CO2 permselectivities for MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) membranes at a pressure difference of 0.3 MPa are investigated in Fig. 11. The H2 permeances and H2/CO2 permselectivities of all tested membranes increase practically linearly as the temperature within the measuring range rises. The H2 permeance and H2/CO2 permselectivity of the Ni0.24Co0.056/MSiO2 membrane at 0.3 MPa and 200 °C are 1.2 and 2.0 times higher than those at 0.3 MPa and 25 °C, respectively. The transport of H2 molecules is dominated by the activation-diffusion process, so the relatively high temperature is more conducive to the movement of H2 molecules, thereby increasing the H2 permeance. The CO2 permeances are not optimal at relatively higher temperatures due to the violent movement of CO2 molecules and the increase in the average free path. These factors are conducive to promoting the increase of H2/CO2 permselectivity. The results show the relatively high temperature (200 °C) is favorable to promoting the H2 permeances and H2/CO2 permselectivities.The apparent permeance activation energy (E a) quantifies the energy required for gas molecules to permeate through membrane pores. According to the Arrhenius Law [36], the permeance F is the temperature-dependent parameter represented by Eq. (4): (4) F = F 0 exp ( – E a RT ) where F is the gas permeance (mol m−2 Pa−1 s−1), F 0 is the maximum permeance at infinitely high temperatures (mol m−2 Pa−1 s−1), E a is the apparent permeance activation energy (kJ mol−1), R is the gas constant (J mol−1 K−1) and T is the absolute temperature (K). Eq. (4) can be described in Eq. (5): (5) ln F = ln F 0 – E a RT Fig. 12 performs the Arrhenius relationships of H2 and CO2 permeances in MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) membranes. Table 3 lists the apparent permeance activation energy (E a) of H2 and CO2 for MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) membranes. There are percolative paths in the membranes that allow H2 and CO2 molecules to diffuse. The E a values of H2 molecules are positive, while E a values of CO2 molecules are negative, consistent with prior results [37]. The positive or negative E a values are correlated with the gas transport mechanisms. The negative values of E a are frequently interpreted as the strong adsorption of molecules on the pore surface. The magnitude of E a value depends on the pore diameter, porosity and interaction between pore walls and gas molecules. The higher E a value indicates the larger energy barrier that the gas must overcome during transport in the membrane pores, hence raising the difficulty coefficient of gas diffusion. The E a value of H2 in the Ni0.024Co0.56/MSiO2 membrane is lower than that in other membranes, so H2 molecules can permeate into the membrane with less repulsive force. This is because the doping of nickel and cobalt elements enlarges the total pore volume of the MSiO2 membrane, which successfully reduces the densification of the SiO2 network. In addition, the NiO and CoO have the good affinities for H2 molecules, which facilitates the surface diffusion of H2 molecules. However, the E a value of CO2 in the Ni0.024Co0.56/MSiO2 membrane is higher than that in other membranes. The NiO and CoO are alkaline metal oxides with significant adsorption effects on acidic CO2 molecules [38]. Therefore, CO2 molecules are adsorbed on the membrane pore wall, resulting in the contraction of pore walls and an increase in transport resistance of CO2 molecules [20]. The Ni0.024Co0.56/MSiO2 membrane exhibits the lowest diffusion resistance of H2 molecules and the highest diffusion resistance of CO2 molecules among all tested membranes, which effectively promotes H2 permeance and H2/CO2 permselectivity. Fig. 13 demonstrates the H2 permeances and H2/CO2 permselectivities for MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) membranes at a pressure difference of 0.3 MPa and 200 °C. The H2 permeances of MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) membranes are 3.6 × 10−7–7.6 × 10−7 mol m−2 Pa−1 s−1 and the H2/CO2 permselectivities are 41.2–113.5. The binary nickel-cobalt doping is more favorable to achieving higher H2 permeances and H2/CO2 permselectivities than doping with the single metal. The total amount of nickel and cobalt elements in binary nickel-cobalt-doped silica membranes remain unchanged, but the increase in cobalt content is more conducive to promoting H2 permeances and H2/CO2 permselectivities. So it becomes clear that this specific Ni/Co substitution ratio of Ni0.024Co0.056/MSiO2 membrane is responsible for the exhibited maximum H2 permeance and H2/CO2 permselectivity. Therefore, this specific Ni/Co substitution ratio seems to create an average pore size that is the most effective for H2/CO2 separation.The Ni0.024Co0.056/MSiO2 membrane exhibits superior H2 permeance and H2/CO2 permselectivity than the MSiO2 membrane. Compared with the MSiO2 membrane, the H2 permeance and H2/CO2 permselectivity of the Ni0.024Co0.056/MSiO2 membrane increase by 2.1 and 2.8 times, respectively. The possible mechanisms in MSiO2 and Ni0.24Co0.056/MSiO2 membranes for H2/CO2 separation are presented in Fig. 14. The separation of gas molecules is based on differences in pore sizes and differences in affinity for gas molecules on the surface of membranes. The introduction of nickel and cobalt elements enlarges the total pore volume and microporosity of the MSiO2 material (Table 2). The diffusion mechanism of H2 in the metal-doped membrane differs from that of the MSiO2 membrane. The NiO and CoO have the significant affinities for H2 molecules, which can enhance the surface diffusion of H2 molecules, hence facilitating adsorption and transport of H2 molecules, which is conducive to the growth of H2 permeance and H2/CO2 permselectivity [39]. Fig. 15 plots the H2 permeances and H2/CO2 permselectivities of steam-treated and regenerated MSiO2 and NixCo0.08−x/MSiO2 membranes at a pressure difference of 0.3 MPa and 200 °C. The H2 permeance and H2/CO2 permselectivity of the steam-treated Ni0.024Co0.056/MSiO2 membrane increase, while those of the steam-treated MSiO2 membrane decrease. Compared with the fresh samples, the H2 permeance and H2/CO2 permselectivity of the steam-treated MSiO2 membrane drop by 9.3 % and 8.7 %, respectively, and those of the steam-treated Ni0.024Co0.056/MSiO2 membrane rise by 9.8 % and 5.4 %, respectively. The hydrophilic silanol group (Si-OH) on the MSiO2 membrane surface is the active physical adsorption center, so the Si-OH group easily absorbs water vapor when the MSiO2 membrane is in a humid heat state for a long time. This causes a hydrolysis reaction between the SiOSi bonds in the silica structure and the water molecules, which causes the SiOSi bonds to break and form new Si-OH groups. The Si-OH groups occupy part of pore spaces of the MSiO2 membrane, which increases the diffusion resistance of H2 molecules, thereby reducing the H2 permeance and H2/CO2 permselectivity. The doping of nickel and cobalt elements broadens the total pore volume and average pore size of the MSiO2 membrane, consequently decreasing the H2 resistance through the steam-treated Ni0.024Co0.056/MSiO2 membrane and enhancing the H2 permeance and H2/CO2 permselectivity. In addition, the nickel and cobalt elements formed SiONi and SiOCo bonds can prevent the network structure of the steam-treated Ni0.024Co0.056/MSiO2 membrane from breaking, therefore enhancing the hydrothermal stability.The H2 permeance and H2/CO2 permselectivity of the regenerated MSiO2 membrane exhibit the recovery trend, which is because the physically adsorbed water in membrane pores evaporate at high temperature, so the diffusion resistances of H2 through the MSiO2 membrane decrease, resulting in the inevitable increase of H2 permeance and H2/CO2 permselectivity. The H2 permeance and H2/CO2 permselectivity of the regenerated MSiO2 membrane are still lower than those of fresh samples, but those of the regenerated Ni0.024Co0.056/MSiO2 membrane continue to increase. Compared with the fresh samples, the H2 permeance and H2/CO2 permselectivity of the regenerated MSiO2 membrane decrease by 7.5 % and 7.4 %, respectively, and those of the regenerated Ni0.024Co0.056/MSiO2 membrane increase by 12.8 % and 8.3 %, respectively. The Si-OH groups on the MSiO2 membrane surface underwent further rearrangement and condensation reactions during the regeneration process, resulting in the compact and even collapse of the micropore structure, showing lower H2 permeance and H2/CO2 permselectivity than the fresh samples. This effect may be mitigated by the presence of Ni-O-Si and Co-O-Si bonds, so the Ni0.024Co0.056/MSiO2 membrane demonstrates more repeatability. The findings demonstrate that the introduction of nickel and cobalt significantly improves the hydrothermal stability and repeatability of the MSiO2 membrane. Table 4 demonstrates the pore sizes, E a values, gas permeances and permselectivities, hydrothermal stability of various silica membranes. There are two methods to characterize the pore sizes. One is to measure the pore sizes by gas permeation measurements using the supported membranes, and the other is to calculate pore sizes by the N2 adsorption-desorption isotherms using the unsupported membrane materials. Since the drying stresses of unsupported silica membrane materials and supported silica membranes during heat treatment are not exactly the same, the two materials cannot be expected to have exactly the same structure. The particle size and pore structure data of unsupported membrane materials cannot be quantitatively converted to the situation of supported membranes, but can qualitatively indicate the changing trend of material structure in the treatment process. The silica membranes can be prepared by multi-step coating process which can aid in reduction of the number of defect sites and decrease the pore size that gas molecules go through. Generally speaking, a larger pore size may lead to a lower apparent permeance activation energy, a higher gas permeance and a lower gas permselectivity. The gas permeance and permselectivity are the two most important parameters of silica membranes, although there is always a trade-off between both. The increase of gas permeance is always at the expense of gas permselectivity and vice versa, so it is difficult to simultaneously enhance gas permeance and permselectivity. The introduction of nickel and cobalt elements can boost the surface diffusion of H2 molecules, hence promoting the adsorption and transport of H2 molecules, which is advantageous for achieving higher H2 permeances and H2/CO2 permselectivities. Furthermore, the NiOSi and CoOSi bonds formed by nickel and cobalt doping can improve the hydrothermal stability and repeatability of silica membranes. Binary nickel-cobalt metal doping provides a novel avenue to develop high-performance membranes with enhanced gas permeance and hydrothermal stability.In summary, MSiO2 and NixCo0.08−x/MSiO2 (x = 0, 0.024, 0.04, 0.056 and 0.08) materials and membranes were successfully prepared using TEOS, MTES, Ni(NO3)2·6H2O and Co(NO3)2·6 H2O solutions. The physicochemical properties and microscopic morphologies were systematically characterized by FTIR, XRD, TEM, N2 adsorption-desorption and SEM. The H2/CO2 permselectivities of membranes were evaluated by differential pressure, temperature, Ni/Co content and hydrothermal stability and other factors. The phase structure analysis proved that nickel and cobalt elements were successfully incorporated into SiO2 network in the form of SiONi/SiOCo bonds and NiO/CoO. The apparent activation energies of H2 permeances in MSiO2 and Ni0.024Co0.056/MSiO2 membranes were 2.67 ± 0.04 and 1.13 ± 0.06 kJ mol−1, respectively. When operating at a pressure difference of 0.3 MPa and 200 °C, the H2 permeances of MSiO2 and NixCo0.08−x/MSiO2 membranes were 3.6 × 10−7 × 10−7 mol m−2 Pa−1 s−1and 7.6 × 10−7 mol m−2 Pa−1 s−1, respectively, and the H2/CO2 permselectivities were 41.2 and 113.5, respectively. Compared with the MSiO2 membrane, the H2 permeance and H2/CO2 permselectivity of the Ni0.024Co0.056/MSiO2 membrane rose by 2.1 and 2.8 times, respectively. After steam treatment and regeneration, compared with the fresh samples, the H2 permeance and H2/CO2 permselectivity of the MSiO2 membrane fell by 7.5 % and 7.4 %, respectively, and those of the Ni0.024Co0.056/MSiO2 membrane increased by 12.8 % and 8.3 %, respectively. The doping of nickel and cobalt elements enhanced H2 permeances and H2/CO2 permselectivities, and counteracted numerous effects of water on silica matrixes, thereby improving the hydrothermal stability of the MSiO2 membrane. We will conduct additional comprehensive tests of binary nickel-cobalt-doped silica membranes for separating H2 from various gases (N2, CH4 and O2) in order to determine their practical efficiency in industrial applications. Mengyu Yan: Conceptualization, Writing – original draft. Jing Yang: Conceptualization, Writing – review & editing, Funding acquisition. Ruihua Mu: Methodology, Project administration. Yingming Guo: Software, Project administration. Xinshui Cui: Supervision, Data curation. Jinghua Song: Formal analysis.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This work was supported by the Key Research and Development Projects of Shaanxi Province, China [2022SF-287 and 2021GY-147]; and the Scientific Research Plan Projects of Shaanxi Education Department, China [19JC017 and 21JK0650].

Silica membranes possess gas separation characteristics, but their undesirable performances and poor hydrothermal stabilities hinder the application in industrial gas treatments. To solve the problems, we have devised a new membrane preparation method involving methyl group modification and nickel-cobalt doping. In this paper, methyl-modified silica (MSiO2) and nickel-cobalt-doped methyl-modified silica (NixCo0.08−x/MSiO2, x = 0, 0.024, 0.04, 0.056 and 0.08) materials and membranes were synthesized using the sol-gel technique. The physical-chemical properties of materials were characterized by FTIR, XRD, TEM, N2 adsorption-desorption and SEM. The H2 permeances and permselectivities of membranes were evaluated with pressure difference, temperature, Ni/Co content and hydrothermal stability as the inferred factors. In Ni-Co/MSiO2 materials, nickel and cobalt elements were found in the form of Si-O-Ni/Si-O-Co bonds and NiO/CoO. The H2 permeance and H2/CO2 permselectivity of the Ni0.024Co0.056/MSiO2 membrane were 7.6 × 10−7 mol m−2 Pa−1 s−1 and 113.5, respectively, which were 2.1 and 2.8 times higher than the MSiO2 membrane at 0.3 MPa and 200 °C. After steam treatment and regeneration, the Ni0.024Co0.056/MSiO2 membrane increased from the original values in the H2 permeance and H2/CO2 permselectivity, while the MSiO2 membrane decreased. The final results revealed that Ni-Co/MSiO2 membranes possessed excellent H2/CO2 permselectivity and hydrothermal stability.

Although the lockdowns and other measures implemented to cope with the Covid-19 pandemic substantially reduced the global energy demand, it is expected that the trend will be on the rise again in the near future. By and large, fossil fuel is still an important source of energy for industries, households, especially for the various means of transportation. Due to the environmental impact such as greenhouse emissions and the climate change, many forms of alternative energy have been proposed to substitute the use of fossil energy. Considered as carbon neutral, energy from plants in the form of biofuel has gained interests worldwide since there is a wide variety of raw materials that can be used such as linseed oil, jatropha oil, coconut oil, soybean oil, sunflower oil, palm oil, etc. [1–7]. However, food security as well as economic viability are still the main issues to be considered when developing the processing plants.Crude palm kernel oil (CPKO) is vegetable oil obtained from the inner part of palm fruit without bleaching or refining process. Using CPKO as raw material for the production of biofuel certainly helps improve the economic performance of the process. CPKO consists of both saturated and unsaturated fatty acid chains esterified to the glycerol backbone, which can potentially be converted into other products such as alkanes, alkenes, aromatics, isomer compounds, or cyclic compounds depending on the type of catalyst and the operating conditions. Hence, CPKO can be used as raw material to produce biojet, in which can be used for the aviation industry. The conventional production of biojet relies heavily on the fossil resources. Four alternative methods have been proposed for producing jet fuel from non-fossil raw materials. The first method is alcohol-to-jet fuel using methanol, ethanol, butanol, and long-chain fatty alcohol as raw material. Another method is called gas-to-jet fuel, converting biogas, natural gas, syngas into jet fuel via fermentation and Fisher Tropsch synthesis. Various types of sugar such as sugarcane, corn, fruit residue can be used to produced jet fuel by the process called sugar-to-jet. The fourth method is oil-to-jet, which is the chemical conversion of vegetable oil into bio-jet fuel via hydro-processing, hydrotreating, deoxygenation, isomerization/hydrocracking [8–10]. This work is related to the application of oil-to-jet method, which involves the use of hydrogen gas to transform unsaturated hydrocarbon, aromatics, and heteroatoms into biojet fuel.Hydrocracking is a chemical reaction that usually requires high temperatures exceeding 400 °C in order to produce smaller molecules. Deoxygenation reaction (DO) of triglyceride with hydrogen can occur with the presence of metal catalyst at moderate temperatures (250–450 °C) and the pressure of 10–300 bar [8,11–15]. There are three different chemical reactions in this category including hydrodeoxygenation (HDO), decarbonylation (DCN), and decarboxylation (DCX). The first reaction eliminates oxygen in the form of water molecule. On the other hand, DCN and DCX release oxygen in the form of CO and CO2, respectively, causing a reduced number of carbon in the hydrocarbon product [16]. Platinum-group metals (Ru, Rh, Os, Ir, Pd, and Pt) have been used as catalyst for the conversion of vegetable oil to biofuel. Snare et al., [17] studied the effect of different metal catalysts (Pd, Pt, Ru, Mo, Ni, Rh, Ir, and Os) supported on carbon and metal oxide on the deoxygenation of stearic acid, which is a saturated fatty acid with 18 carbon atoms. The main product was n-heptadecane (n-C17). Both 5%Pd/C and 5%Pt/C provided high conversion and high selectivity toward C17 exceeding 90%. Pt/C was also reported as an outstanding catalyst for hydrogenation and deoxygenation reactions especially DCN and DCX [18,19]. This helps reduce the amount of water as by-product (from HDO) and enhances the overall biofuel yield. Fu et al. [20] performed hydrothermal conversion of five different fatty acids including stearic, palmitic, lauraic, oleic, and linoleic acid at high temperatures over Pt/C catalyst. Saturated fatty acids (stearic acid, palmitic acid, and lauric acid) were converted to n-alkanes via decarboxylation reaction with high selectivity exceeding 90%. For unsaturated fatty acid (oleic acid and linoleic acid), the major pathway was the hydrogenation reaction producing saturated fatty acid followed by decarboxylation to form n-alkanes. This research showed that Pt/C can be used as an efficient catalyst for both deoxygenation and hydrogenation reactions. Scaldaferri et al. [21] applied a group of catalysts including Pd/C, NbOPO4, ZSM5, and beta-zeolite for deoxygenation of soybean oil under 10 bar of nitrogen pressure in a batch reactor. Among these catalysts, Pd/C provided the highest yield of bio-jet fuel (62%) followed by niobium phosphate (50%). The fraction of oxygenated compound was only 1% for the case of Pd/C catalyst. The stable and oxygen-free hydrocarbon product tends to have low viscosity, improving flow characteristics at low temperatures [22]. These findings in the literature confirm that Pt/C catalyst is one of the promising catalysts for biojet application.Freezing point is the key property for jet fuel that should be kept below −47 °C in order to maintain flow properties at high altitudes (low temperature). The content of iso-alkanes, cycloalkanes and small molecules of hydrocarbons can greatly affect the freezing point of jet fuel [23]. The heat of combustion must be of at least 42.8 MJ/kg while the content of aromatics must not exceed 25 vol% according to standard specification for aviation turbine fuel (ASTM D1655-04a, ASTM D7566, JP-8 MIL-DTL-83133E) [24,25].The choice of reactor type is one of the important elements for the production of biojet. Many newly developed catalysts have been tested in a batch reactor where the reacting mixture is thoroughly mixed. In this system, the overall rate of reaction is impeded due to the dilution with reaction products and chemical equilibrium. Another issue is the separation and recycling of catalyst after the operation completed. To reduce the effects of dilution and chemical equilibrium while enhancing the productivity, this work applied a continuous fixed bed reactor, where the solid catalyst is placed inside. Moreover, the small footprint of a continuous process also improves the safety of the operation especially when dealing with high pressures.Therefore, this study applied 5%Pt/C as catalyst for deoxygenation of CPKO. The catalyst was packed in a continuous fixed bed reactor. The effect of operating parameters including reaction temperature, pressure, amount of catalyst, hydrogen flow rate, and CPKO flow rate on the reaction performance were investigated. The fuel properties of the product at the optimal conditions were compared with various standards and the reaction performance was compared with other systems as reported in the literature.CPKO was obtained from Univanich Palm Oil Public Company Limited, Thailand. It was kept in an amber glass bottle at −20 °C. Standard n-alkane (nC8-nC20, 40 mg/L each in hexane) solution was supplied by Sigma-Aldrich. Cyclohexane (GC grade, 99.8%) and methyl alcohol (HPLC grade, 99.9%) for esterification and transesterification reactions were purchased from Fisher Chemical. Sulfuric acid (ACS reagent grade, 98%) was purchased from Merck. The 5% Pt/C catalyst and H-ZSM-5 zeolite (Si/Al = 15) were supplied by Riogen Inc. Isopropanol (AR grade, 99.7%) was purchased from QReC. Sodium sulphate anhydrous (99.0%) and potassium hydroxide pellet (85%) were obtained from Carlo-Erba. Ultra-high purity helium (99.999%), compressed air (air zero), and high purity hydrogen (99.99%) were supplied by Linde (Thailand).The catalyst was characterized by Scanning Electron Microscope and Energy Dispersive X-ray Spectrometer (S-4800, Hitachi) to study the surface morphology and dispersion of metal on the support. Nitrogen physisorption (3Flex surface characterization analyzer, Micromeritics) was used to determine the specific surface area according to Brunauer-Emmett-Teller theory (BET). The total pore volume and the average pore size were calculated by Barrett-Joyner-Halenda (BJH) method. Ammonia (NH3) temperature programed desorption (BELCAT II, Thermo Finnigan) was used to determine the acidity of the catalyst.Pulse H2 chemisorption (AutoChem II 2920 chemisorption analyzer, Micromeritics) was applied to determine the metal dispersion of Pt on the activated carbon. The molecular coordination of hydrogen to metal was 1:2 [26]. The amount of 0.1 g of catalyst was placed in a U-glass tube. At 330 °C under inert atmosphere, the stream of argon flowing at a rate of 30 mL/min was used to flush the impurity off the catalyst for 90 min. Then the temperature was changed to 45 °C. Pulse chemisorption was performed until saturation was achieved.To obtain the profile of free fatty acids of CPKO (acid value was 33.95 mg KOH/g oil), two reaction steps were performed. The first step was the esterification of CPKO. After removing the moisture in CPKO, excess methanol and 4 wt% sulfuric acid were added. The reaction was carried out at 65 °C in a round-bottom flask with reflux. The product was washed with deionized water. After esterification reaction, the acid value was 0.83 mg KOH/g oil. Then the oil was transesterified with methanol at 65 °C for 1.5 h using 1.5 wt% KOH as catalyst. The obtained fatty acid methyl ester (FAME) was washed with deionized water, followed by removing the moisture. The sample was then analyzed by GC–MS.Thermal gravimetric analysis (TGA/DSC 3+, Mettler Toledo) was used to study the decomposition of CPKO in the temperature range of 35–650 °C with the heating rate of 10 °C/min and the nitrogen flow rate of 60 mL/min.The catalyst particles were packed in a tubular reactor made of stainless steel (od. ¼ inch × 0.049 in WT). Two layers of glass wool (0.1 g for each layer) were used to sandwich the catalyst bed in the middle. The reactor was placed inside a tubular furnace, where the reaction temperature was adjusted by means of a temperature controller. For a typical experiment, the catalyst bed was treated with hydrogen flowing at a rate of 70 mL/min at 330 °C for 90 min. A certain amount of CPKO was ultrasonicated to eliminate any dissolved gas prior to be fed as reactant. Then the flow rate of hydrogen was adjusted to the desired value via a mass flow controller while the flow rate of CPKO was adjusted via HPLC pump. Both streams were mixed in a J-mixer (od. 1/16 in, 0.020 in thru hole) with liquid stream at right angle. Then the reacting mixture entered the reactor previously packed with catalyst particles. The exit-end of the reactor was equipped with a back-pressure regulator and a separator that was used to separate gas and liquid products as shown in Fig. 1 . Liquid samples were collected every hour for the total of 10 h at the sampling port. Anhydrous NaSO4 was used to remove water from the samples prior to the analysis. The operating conditions applied are summarized in Table 1 .For the analysis by GC-FID and GC–MS, the oil samples were diluted with cyclohexane (CAS number 110-82-7). The volume of 1 μL of diluted oil sample was injected with a split ratio of 40:1 to the GC-FID (HP6890, Agilent) equipped with DB-5HT GC column (30 m, 0.25 mm in diameter and 0.25 µm film thickness). The injector temperature and detector temperature were set at 350 °C. The oven was heated from the initial temperature of 50 °C and held for 2 min followed by a temperature ramp to 130 °C at a heating rate of 10 °C/min and a final ramp to 365 °C at a rate of 15 °C/min. The system was held at this temperature for 15 min. Helium was used as a carrier gas with a constant flow rate of 2.0 mL/min. Standard normal-alkanes (linear alkane) were used to characterize the hydrocarbon product based on the retention times of GC-FID chromatogram. Eq. (1) was used to calculate the fraction of certain compound(s) in the sample. This can be used to determine the yield of that product fraction via Eq. (2). In this work, the desired product was categorized in two groups, which were nC8-nC16 and non nC8-non nC16. Note that the latter was distinguished by the peaks associated with retention times in the range of nC8-nC16 but not the same as that of standard n-alkanes [27]. Both groups were considered as biojet fuel range and GC–MS was used to analyze the potential biojet products. The productivity was calculated via Eq (3) in order to compare the performance with other systems as reported in the literature. Eqs. (4)–(6) were used to determine the yield of liquid, water, and gas product. The water yield was calculated based on the difference of the weight of oil before and after the use of NaSO4 as shown in Eq. (5). After that, the mass balance was used to determine the mass of gas product. (1) % A r e a A = A r e a o f A T o t a l a r e a o f o i l p r o d u c t × 100 (2) % Y i e l d A = % A r e a o f A × w e i g h t o f o i l p r o d u c t w e i g h t o f C P K O (3) P r o d u c t i v i t y = % A r e a × w e i g h t o f o i l p r o d u c t a m o u n t o f c a t a l y s t × t i m e where A is either nC8-nC16 or non nC8-non nC16 compounds. (4) L i q u i d y i e l d ( % ) = M a s s o f l i q u i d p r o d u c t M a s s o f s t a r t i n g C P K O × 100 (5) W a t e r y i e l d ( % ) = M a s s o f l i q u i d p r o d u c t M a s s o f s t a r t i n g C P K O × 100 (6) G a s y i e l d ( % ) = M a s s o f g a s p r o d u c t M a s s o f s t a r t i n g C P K O × 100 Fatty acid profile of CPKO were identified via GC–MS (5975C, Agilent) equipped with DB-FastFAME column, Agilent G3903-63011 (30 m × 0.25 mm × 0.25 µm). The volume of 1 μL of diluted oil sample was injected. Initial temperature of 50 °C was ramped to 194 °C with heating rate 30 °C/min and held for 3.5 min followed by another temperature ramp to 250 °C at a heating rate of 5 °C/min. The system was held at this temperature for 1 min.GC–MS (5975C, Agilent) equipped with DB-5MS column, Agilent 122–5532 (30 m, 0.25 mm in diameter and 0.25 µm film thickness) was used for the quantification of iso-alkanes, cycloalkanes, alkenes, aromatic, oxygenated compounds and other components in the oil product for P9 and P9HZSM0.06. The injector was set at 300 °C. The oven was heated from the initial temperature of 40 °C and held for 5 min followed by a temperature ramp to 300 °C at a heating rate of 2 °C/min. The system was held at this temperature for 10 min. The ion source temperature was at 230 °C. Helium was used as a carrier gas with a constant flow rate of 1.5 mL/min. The compounds in CPKO and oil product were identified based on the NIST mass spectral database.Mettler Toledo DSC 3+ was employed to identify the freezing point and melting point of CPKO and product samples. A small amount of sample (10.50 ± 0.5 mg) was subjected to two thermal treatments. The first temperature profile was a cooling process from the initial temperature of 40 °C to −75 °C at a rate of −5°C/min under the nitrogen flow at a rate of 50 mL/min. Then another temperature profile was imposed by ramping up from the final temperature of the previous step to 40 °C at a rate of 5 °C/min.The boiling range of oil product (ASTM D2887) was determined by another GC-FID (Varian CP3800) equipped with Zebron ZB-1XT Simdist capillary column (15 m × 0.53 mm × 0.25 µm). A temperature program started from the initial temperature of 40 °C where it was held for 1 min before ramping up to 370 °C at a heating rate 10 °C/min. The oven temperature was held at 370 °C for 5 min to complete the profile. The injector temperature and detector temperature were set at 370 °C. The volume of sample injected into the column was 1 µL using a spitless mode.To assess the change of oxygen content in the oil. Both CPKO and the product were analyzed by the Elemental analyzer (CHNS/O Analyzer, 628 series, Leco Coporation, USA) at 1300 °C. The acid value of CPKO was determine via titration (EN 14104) with 0.1 M alcoholic potassium hydroxide. Isopropanol was used as solvent and phenolphthalein was used as indicator. The acid value can be calculated by Eq. (7). (7) A c i d v a l u e = C K O H × V K O H × 56.1 W o i l where C K O H  = concentration of potassium hydroxide solution (mol/L) V K O H  = volume of potassium hydroxide solution used for titration (mL) W O i l  = weight of oil sample (g) C K O H  = concentration of potassium hydroxide solution (mol/L) V K O H  = volume of potassium hydroxide solution used for titration (mL) W O i l  = weight of oil sample (g)The heating value of oil product was measured using Parr 6400 calorimeter. Note that the minimum heating value of biojet fuel is 42.8 MJ/kg (ASTM D1655-04a, ASTM D7566).From nitrogen physisorption experiment, the adsorption isotherm of our fresh catalyst as presented in Fig. 2 a showed characteristics of microporous material (pore width < 2 nm). The pore size distribution is shown in Fig. 2b with the average pore diameter of 2.21 nm. Note that this average pore diameter was close to the boundary between micropores and mesopores. The adsorption isotherm revealed the S-shape with a narrow hysteresis loop. This was probably resulted from the partially fused micropores. The analysis revealed the pore volume of 0.57 cm3/g and high BET specific surface area of 1,038.66 m2/g, which was in the range of surface area of activated carbon [28], indicating that the surface area of catalyst was not significantly affected by the platinum loading.The morphology of catalyst surface is shown in Fig. 3 . The SEM images indicate the porous structure with deposits of small particles, commonly observed for loaded activated carbon. The elemental mapping via SEM-EDX analysis suggested that the metal was uniformly distributed on the activated carbon support. The metal dispersion determined via H2-chemisorption was 6.49%, close to the amount of Pt loading in the manufacturing process (5 wt% Pt). Obtained from the NH3-TPD experiment, the acidity of catalyst is represented in Fig. 4 , with two major peaks at 225 °C and 900.8 °C corresponding to the weak acid sites (0.05 mmol/g) and strong acid sites (1.773 mmol/g), respectively. It was possible that the shoulder peak around 600 °C was caused by the partial decomposition of carbon [29–31]. The peak between 500 and 900 °C from our result represented the strong acid sites of Pt/C catalyst, according to NH3-TPD technique. Prior to the temperature ramp, ammonia molecules adsorped on acid sites of catalyst. Upon increasing of temperature, ammonia molecules desorped from catalytic surfaces. The amount of ammonia leaving the catalyst was interpreted as the amount of acid sites despite the partial decomposition of carbon occurring at high temperatures. This was in line with the work of Lawal [32], reporting the NH3-TPD profiles of 5% Pt/C and 5% Pt/graphite. However, the cause of strong acid sites was not specified. Ob-eye et al. [33] reported the strong acidity of activated carbon. There is also a possibility that the strong acid sites resulted from the interaction of activated carbon and Pt. A further study is required to identify whether the strong acid sites were related to Pt or activated carbon, or the synergistic interaction. Note that high acidity of catalyst can be associated with the catalytic performance of the deoxygenation. The detailed characterization of HZSM-5 (used for improving the fuel properties in this research) was reported by Bangjang et al. [34].The decomposition of CPKO was obtained from thermal gravimetric analysis (TGA) and derivative thermogravimetric analysis (DTG), as shown in Fig. 5 a. It was observed that the thermal degradation of CPKO consisted of 2 steps for decomposition of polyunsaturated and monounsaturated fatty acid, respectively [35,36]. Approximately 15.76% loss occurred between 204.49 °C and 313.37 °C, due to the decomposition of polyunsaturated fatty acids. The second step involved the decomposition of monounsaturated fatty acids as observed between 368.47 °C and 441.54 °C, accounting for the weight loss of 83.16%. According to the TGA/DTG analysis, the reaction temperature for our hydroprocessing of CPKO should not exceed 450 °C. Further increasing the temperature could cause the excess thermal cracking of liquid hydrocarbon to undesired product. Fig. 5b shows the temperature profiles for melting and freezing of CPKO obtained from differential scanning calorimetric analysis (DSC). On the cooling segment, the solidification of CPKO was observed from 15.10 to 11.54 °C, while the melting was observed from 12.63 to 28.28 °C with trace amount melted at −19.75 °C. These results will be used to compare with that of the hydroprocessing product.The acidity of CPKO, determined via titration method (EN14104), was 33.95 mg KOH/g oil. The fatty acid profile of CPKO was identified by GCMS. The total saturated fatty acid content, total unsaturated fatty acid, fatty acid profile, and the average molecular weight of CPKO are summarized in Table 2 . The content of fatty acids in the jet fuel range (C8-C16) was 71.74%. Hence, the deoxygenation of CPKO over Pt/C can potentially lead to the high yield of jet fuel product. Besides, the unsaturated chain can undergo hydrogenation, isomerization, and cracking at the double bond chains. The saturated chains may also be dehydrogenated on metal catalyst followed by isomerization and cracking to produce jet fuel [37,38].One of the most important parameters for catalytic reaction strongly affecting the reaction performance is the reaction temperature. In this work, the temperature range of 350–420 °C was used to study the effect of reaction temperature on the yield and properties of product. Considering experimental conditions P1-P4, where the pressure, amount of catalyst, feed rate of oil, and feed rate of hydrogen were kept constant at 500 psi, 0.05 g, 17.5 mL/min, and 0.04 mL/min, respectively. It was found that the color of liquid product changed with increasing reaction temperature. Fig. 6 displays four sets of 10 samples collected hourly from the phase separator at each reaction temperature. Note that the collection time was based on the time when CPKO was introduced to the reactor. The color did not significantly change when the reaction temperature was below 400 °C; however, the samples appeared distinctively darker at 420 °C possibly due to the contribution of thermal cracking (see Fig. 8 ). All samples were stored at −20 °C. The product did not change the physical appearance as well as chemical composition after 4 months of storage at −20 °C.At high temperatures, the cracking reactions of triglycerides in CPKO led to small molecules of product, while the deoxygenation and hydrogenation reactions yielded saturated compounds in the form of normal chain length hydrocarbons (n-compounds) [17]. As shown in Fig. 7, the fraction of non n-alkane in the biojet range (peaks between nC8-nC16) including iso-alkane, cycloalkanes, oxygenated compounds increased with increasing temperature. Table 3 shows the yield of liquid and gas products obtained from different operating conditions. The liquid yield was separated into two fractions, one was aqueous (mostly water) and another was organic (liquid hydrocarbon). For P1-P4, the yield of gas product increased with increasing reaction temperature. The presence of CPKO was observed in the liquid product obtained from P1-P4. The gas product constituted of CO (from DCN), CO2 (from DCX), light hydrocarbons such as CH4, and remaining H2. Note that the content of gas and aqueous products can reflect on the degree of conversion of CPKO via deoxygenation reactions. According to the aqueous content, indicating the presence of water produced as by-product from DCN and HDO, the reactions performed at 420 °C yielded a slightly higher content of water. However, this slight proportion suggested that DCX was the major reaction pathway. This was in line with the previous study by Liu et al. [39], who applied deoxygenation of palmitic acid over 5%Pt/C. In this case, the major product was pentadecane (C15), produced via decarboxylation (DCX).This was related to the increased conversion of CPKO, suggesting that the hydrogenation and hydrodeoxygenation reactions were significantly promoted. The fraction of nC8-nC16 was rather constant throughout the time-on-stream of 10 h. For the reaction temperature of 420 °C (P4), the average yield of biojet range fuel was 59%. Although the reaction performance was somewhat subdued, the biojet yield of approximately 40% was obtained at 400 °C. Also, the product had a slightly yellowish appearance (not dark). Both reaction temperatures (400 °C and 420 °C) were used for subsequent experiments in order to optimize other operating parameters.It is conceivable that the amount of catalyst should be balanced with the flow rate of CPKO to allow for efficient catalytic reactions, while suppressing the undesired reactions. Therefore, the effects of both catalyst amount and flow rate of CPKO on the reaction performance were investigated. In this set of experiments, the pressure and hydrogen flow rate were maintained at 500 psi and 17.5 mL/min, respectively. Fig. 8 shows samples collected for the reaction temperature of 400 °C and 420 °C. Based on the physical appearance of the samples, the color of product was affected by the operating conditions applied for our hydro-processing. For the reaction temperature at 400 °C and catalyst amount of 0.05 g (P3 and P5), the color of product appeared lighter when the flow rate of CPKO was decreased from 0.04 mL/min (P3) to 0.02 mL/min (P5). Then we increased the catalyst amount from 0.05 g (P5) to 0.07 g (P6) while other parameters were kept constant. At this condition, the color of product was transparent. This was owing to the efficient catalytic reaction of CPKO and hydrogen. In other words, the amount of catalyst of 0.07 g was sufficient to handle the CPKO flow rate of 0.02 mL/min. On the contrary, the product was slightly yellowish similar to that of P5 when the flow rate of CPKO was increased from 0.02 mL/min (P6) to 0.04 mL/min (P7). It was probable that a larger portion of intermediates was present. This trend was more apparent for the comparison between P3 and P4. Upon decreasing the flow rate of CPKO (P8) and increasing the amount of catalyst (P9), the color of product became lighter and almost transparent. The trend reversed when the flow rate of CPKO was increased from 0.02 mL/min (P9) to 0.04 mL/min (P10).Oil samples obtained from different operating conditions (P5-P10) were analyzed by GC-FID and the chromatograms are shown in Fig. 9 . The main components in CPKO are represented by the peaks at retention times greater than 25 min. It can be observed that the product of P5 and P6 contained a significant portion of high-molecular weight hydrocarbons despite the complete disappearance of raw material. The presence of raw material was evident when the flow rate of CPKO was increased from 0.02 mL/min (P6) to 0.04 mL/min (P7), indicating that CPKO was not completely converted. A similar trend was also observed when changing the operating conditions from P8 to P9 and then to P10. However, the chromatograms of P8-P10 show that the distribution of product shifted to the left compared to that of P5-P7, suggesting a greater portion of low-molecular weight hydrocarbons. Consequently, we analyzed the yield of biojet range product of samples obtained from P8-P10 in comparison with that of P4, as presented in Fig. 10 . The difference in the total yield of biojet was not significant. However, P9 provided the highest yield of nC8-nC16 and the lowest yield of non-nC8-non nC16. Note that normal-alkane (linear alkane) can burn very cleanly, enhancing the combustion characteristics. P10 was undesirable since the product still contained some unconverted CPKO. Fig. 10d shows the boiling point profiles of the product obtained from different reaction conditions. The shifting of boiling point profile is noticeable. For reaction at 400 °C, the fraction of low-boiling point was smaller than that of 420 °C as shown in Table 4 . Moreover, the boiling point profile of P9 showed a significantly smaller content of high-boiling point hydrocarbons, providing 44.69% of jet fuel (boiling point in the range of 150–280 °C) as shown in Table 4. This efficient conversion of CPKO was a combination of high reaction temperature, sufficient amount of catalyst, and appropriate feed rate of CPKO. The DSC analysis of the product obtained from P8-P10 revealed that P9 yielded the lowest freezing point (-0.34 to −46.95 °C). Hence, this operating condition of P9 (420 °C, 500 psi, 0.07 g of Pt/C, CPKO flow rate of 0.02 mL/min, hydrogen flow rate of 17.5 mL/min) was used to study the effect of hydrogen flow rate.The flow rate of hydrogen is considered as an important parameter affecting the performance of deoxygenation and hydrogenation of vegetable oil. High ratios of H2-to-oil during the hydroprocessing could lead to the product that contains the majority of saturated hydrocarbons with high ratio of hydrogen-to-carbon. This type of product can be associated with high combustion efficiency [40]. A set of experiments were performed by varying the hydrogen flow rate as 17.50, 35.0, and 70.0 mL/min designated by P9, P11, and P12. This corresponded to the H2-to-CPKO molar ratio of 28.02, 55.66, 110.92, respectively. Other operating parameters were kept constant at 420 °C, 500 psi, 0.07 g of Pt/C, and CPKO flow rate of 0.02 mL/min. Fig. 11 a–c shows the effect of hydrogen flow rate on the %yield of biojet fuel. Results indicated that similar yields of biojet were obtained for all experiments (500 psi). Although the H2-to-CPKO ratio increased with increasing hydrogen flow rate, the residence time was negatively affected. In other words, the higher hydrogen flow rate, the shorter the residence time. The analysis via simulated distillation also revealed that the boiling range of the product was not significantly different, as indicated in Fig. 11d. The samples were analyzed by DSC to obtain the freezing profile, which was −0.34 to −46.95 °C, −0.67 to −48.30 °C, and 5 to −44.58 °C for P9, P11, and P12, respectively. Therefore, the operating condition P12 was not considered further due to the broad freezing profile, especially on the higher temperature range. Since, the fuel properties of product obtained from P9 and P11 were similar, P9 (lower hydrogen flow rate) was chosen for further investigation on the effect of reaction pressure. Decreasing the pressure from 500 psi to 250 psi led to the decreased content of n-alkane, owing to the promoted hydrogenation and deoxygenation. At this condition, the content of hydrocarbons in the jet fuel boiling range decreased to 18.60%. Hence, the suitable operating conditions were at 420 °C, 500 psi, 0.07 g of Pt/C, CPKO flow rate of 0.02 mL/min, hydrogen flow rate of 17.5 mL/min.To address the importance of catalyst for this process, a blank test was performed at the optimal conditions (P9). This result was compared with that obtained using 5%Pt/C, as shown in Fig. 12 . Apparently, the color of fuel product was markedly different. Without catalyst, thermal reactions resulted in a dark brown liquid. The aqueous phase was not observed. It was possible that a small degree of deoxygenation occurred for the blank test. Therefore, for the liquid product collected, 7.14% of n-alkane in the range of biojet fuel (nC8-nC16) was possibly produced via random cracking of CPKO followed by hydrogenation. According to the fatty acid profile of CPKO (Table 2), the content of fatty acids in the jet fuel range (C8-C16) was 71.74%. Hence, hydrocarbons in the biojet fuel range were likely produced in the system. The content of n-alkanes of approximately 7.14% was much lower than 27.68% of the product obtained from P9. The content of non-n-alkanes from the blank experiment was 52.42%, suggesting that the deoxygenation was not significantly involved. It was probable that the n-alkanes were produced via cracking reactions of saturated chain fatty acids. The freezing point of 11.47 °C (obtained from DSC analysis) was much higher than those obtained with the use of catalyst. Therefore, in this case, it was necessary to apply the catalyst in order to achieve the desired properties of biojet fuel through the deoxygenation, isomerization, and hydrocracking. Table 5 shows the comparison of reaction performance of hydroprocessing of vegetable oil using different systems such as batch, semi-batch, and continuous reactors. Despite the more extreme conditions in terms of reaction temperature and pressure, our system offered an extraordinarily high productivity (>9 gproduct/gcatalyst∙h) compared to other systems (<1.6 gproduct/gcatalyst∙h), due to the small amount of catalyst and the relatively high weight hourly space velocity (WHSV). The yield of biojet product was comparable to the literature data.One of the important fuel properties for jet fuel is the freezing point. The product obtained from P9 was analyzed by DSC. The result showed that the freezing point was in the range of −0.34 to −46.95 °C, which was higher than −47 °C according to the specification of aviation fuel (ASTM D1655-04a, ASTM D7566, and JP-8 MIL-DTL-83133E) [24,25]. Hence, the fuel product should be upgraded to improve the freezing point. For this, we modified the catalyst bed slightly to incorporate a section of HZSM-5 located behind the bed of Pt/C, with a layer of glass wool sandwiched in between. The operating conditions were kept the same as P9 except the catalyst bed. The amount of HZSM-5 was varied as 0, 0.02 g, 0.04 g, and 0.06 g. The product samples collected at 10 h of reaction time were analyzed by the elemental analyzer, DSC, and simulated distillation. The samples were designated as P9HZSM0.02, P9HZSM0.04, and P9HZSM0.06 for the amount of HZSM-5 of 0.02 g, 0.04 g, and 0.06 g, respectively.According to the elemental analysis, the oxygen content of CPKO, P9, P9HZSM0.02, P9HZSM0.04, and P9HZSM0.06 was 18.83, 14.96, 15.02, 12.46, and 7.64%w/w, respectively. These results confirmed that the oxygen content was reduced after our hydroprocessing. This was in line with the aqueous fraction of the product that increased with increasing amount of HZSM-5, produced via hydrodeoxygenation (HDO) [46]. The product tended to be non-polar compounds, suitable for blending with jet fuel. HZSM-5 also promoted the aromatization, cracking/isomerization [47–49]. This led to the increase of gas product when compared to the product obtained from P9, especially for the first 7 h of time-on-stream (see Fig. 13 a). The chromatograms of these products are presented in Fig. 9h–j. Evidently, the use of HZSM-5 caused a significant shift of product peaks to the left side, indicating lower boiling point of product. This was also supported by the simulated distillation analysis as shown in Fig. 13b. The maximum content of jet fuel of 83.34% was obtained from P9HZSM0.06. Table 6 shows the composition of product obtained from P9 and P9HZSM0.06. The content of aromatics (BTXs) significantly increased with the use of Pt/C and HZSM-5. According to the work of Qian et al. [50], who investigated the properties of diesel blends, increasing the content of aromatics lowered the boiling point and viscosity while increasing the density of the blend. Note that the increased fuel density can help minimize the fuel storage [23]. The catalytic performance of HZSM-5 is generally associated with the Si-to-Al ratio [51]. The lower the ratio, the higher acidity of catalyst, leading to the higher degree of aromatization [52]. In our system, the Si-to-Al was 15. There are two major reaction pathways involving hydrotreating over HZSM-5 catalyst [53]. The first one starts with cracking reactions, producing carbonium ions. Then the isomerization and hydrogenation occurred, resulting in iso-alkanes. The second pathway involves unsaturated hydrocarbons produced via cracking reactions, followed by aromatization via Diels-Alder reaction. Hence, the major product of this pathway is aromatic compounds. In our case, results suggested that the reaction mechanism was dominated by the second pathway. The use of zeolite with a larger Si-to-Al ratio may lead to the larger content of iso-alkanes.As mentioned previously, the content of cycloalkane in the biojet fuel can help decrease the freezing point and n-alkanes exhibit high freezing point [23]. According to Table 6, despite the fact that the content of cycloalkane (6.8%) in the product obtained from P9 was larger than that of P9HZSM0.06, this effect was overwhelmed by the content of aromatic compounds. It was conceivable that the reactions proceeded via hydrogenation and deoxygenation for the case of P9, resulting a larger content of n-alkane, alkyl-cycloalkane, and cycloalkane as compared to that of P9HZSM0.06.The samples were analyzed for the freezing point via DSC. Results are shown in Fig. 14 . Increasing the amount of HZSM-5 led to the decrease of freezing point since high degree of cracking reaction caused high content of small molecules [23,47,49]. The content of BTXs significantly improved the cold flow property of our product [54]. The use of 0.06 g of HZSM-5 together with Pt/C could lower the freezing point by 30 °C compared to that of P9. It was also observed that using 0.02 g of HZSM-5 in the case of P9HZSM0.02 was not sufficient to cause a major difference in terms of gas yield, freezing point, and boiling range. P9 provided the content of 44.69% of jet fuel as compared to 50.53% obtained from P9HZSM0.02 (see Table 4). Therefore, the amount or ratio of HZSM-5 and Pt/C is one of the important parameters for the optimization of this process. The heating value of 43.43 MJ/kg of the upgraded biojet product was not significantly altered as compared to that of the non-upgraded product (43.15%). The production of biojet fuel (C8-C16) via deoxygenation reaction using Pt/C as a catalyst resulted in the high yield of biojet fuel. However, one of the important properties of jet fuel is the freezing point. Apparently, our product (P9) did not meet the criteria. Using HZSM-5 in combination with Pt/C offered the improved freezing point property of biojet product. Due to the possible over-cracking in our system, the acid strength of HZSM-5 should be varied to explore the effect on the chemical composition of the biojet product.The production of biojet fuel from CPKO via hydroprocessing was demonstrated in a continuous process (fixed bed reactor) using 5%Pt/C as catalyst for hydrogenation, cracking, and deoxygenation. The effects of operating parameters including reaction temperature, pressure, CPKO flow rate, hydrogen flow rate, and the amount of catalyst were investigated. The content of n-alkane C8-C16 was significantly improved at 500 psi as compared to that obtained at 250 psi. The hydrogen flow rate in the range investigated did not significantly affect the product quality. The optimal operating conditions were at 420 °C, 500 psi, hydrogen flow 17.50 mL/min, and CPKO flow 0.02 mL/min (H2-to-oil molar ratio of 28.02). The yield of biojet was 58.29%, with the major content of nC8-nC16 of 27.68%. The content of hydrocarbons in the biojet boiling range (ASTM D2887) was 44.69%. The oxygen content was 14.96% as compared to that of the feedstock of 18.83%. The biojet product was upgraded by the use of HZSM-5 in combination with Pt/C as catalyst to promote cracking, aromatization and isomerization. The freezing point was shifted down by 30 °C and the oxygen content was reduced to 7.64% when using 0.06 g of HZSM-5 and 0.07 g of Pt/C. The product with low content of oxygenated compound was classified as non-polar hydrocarbons, suitable for blending with commercial jet fuel. The content of hydrocarbons in the biojet boiling range was 83.34% for the case of combined catalyst. In our system, the high acid strength of HZSM-5 led to the high content of aromatics in the product, contributing to the shifting of boiling range of biofuel. The resulting small hydrocarbons also improved the cold flow property of product. Montakan Makcharoen: Visualization, Methodology, Investigation, Writing – original draft. Amaraporn Kaewchada: Supervision, Writing – review & editing, Conceptualization, Methodology, Writing – original draft. Nattee Akkarawatkhoosith: Visualization, Writing – review & editing, Conceptualization, Investigation. Attasak Jaree: Supervision, Writing – review & editing, Conceptualization, Methodology, Investigation, Resources, Project administration, Funding acquisition.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.The authors acknowledge the financial support from the National Research Council of Thailand (NRCT).

This work focused on the conversion of crude palm kernel oil (CPKO) to biojet fuel via deoxygenation reaction in a fixed bed reactor. The catalyst was Pt supported on activated carbon (Pt/C). The investigation involved the effects of various operating parameters such as reaction temperature (350–420 °C), pressure (250 and 500 psi), flow rate of CPKO (0.02 and 0.04 mL/min), flow rate of hydrogen (17.5, 35.0, and 70.0 mL/min), and the amount of catalyst (0.05 and 0.07 g) on the reaction performance and the fuel properties of product. The biojet yield of 58.29%, with the major content of linear alkane in the range of jet fuel (nC8-nC16) of 27.68% and the productivity of 9.32 g product/g catalyst·h were achieved at the optimal operating conditions (420 °C, 500 psi, CPKO flow rate at 0.02 mL/min, hydrogen flow rate at 17.50 mL/min (H2-to-CPKO molar ratio of 28.02), and 0.07 g of Pt/C catalyst). The oxygen content was reduced from 18.83% (in CPKO) to 14.96% (in the product). To improve the freezing point of biojet fuel via cracking and aromatization, the catalyst bed was modified by adding HZSM-5 as a catalyst bed adjacent to the bed of Pt/C. The freezing point of product was significantly lowered by 30 °C.

Data will be made available on request.The number of daily activities that possess a serious threat to humanity is increasing, which worsens the situation globally. According to the World Economic Forum and the World Health Organization, antibiotic resistance is arguably the biggest threat of the 21st century and might be a “potential tragedy” for human welfare and the global economy [1]. Alarmingly, antibiotic resistance is on the rise, necessitating the development of new classes of antibacterial candidates. Certain antibiotic drugs like Daptomycin, GAR936 Linezolid, and Oitavancin, had created the marketplace for antimicrobial chemotherapy [2,3]. Utilizing metallo-drugs is one of the therapeutic approaches under consideration. Metal complexes are explored as potential candidate as these complexes inhibit enzymatic action, interact with intracellular biomolecules, increase lipophilicity, modify the function of plasma membrane, break cell cycle, and many more [4]. In a similar way, chelation drastically alters the organic functionality of metal–ligand complexes. A variety of metal quinoline based anti-infection drug like Norfloxacin and Ciprofloxacin, were investigated to exert better action than antibiotics alone [5–7]. The cytotoxic study of cisplatin has given a tremendous push to explore novel metal complexes. The success of cisplatin had accelerated metal-derived medicines and sparked a widespread research inspiring scientists to develop alternative methods with better pharmacological properties [8]. SB metallo-drugs were reported against a variety of bacterial and parasite species in countless researches. In 1864, Hugo Schiff firstly explained about the condensation reaction pathway between an amine and an aldehyde resulting SB as a product (Fig. 1 (a)). SB ligands have ability to coordinate with diverse metal ions via imine nitrogen or other donor atoms i.e., oxygen and sulphur (Fig. 1(b)). Numerous researches were focused on the synthesis, adaptability, and reactivity of the central metal atom as well as the existence of the azomethine group, which played a vital role in understanding the biological transformation process and racemization reaction [9,10]. Particularly, the medicinal science was dominated by heterocyclic Schiff base metal chelates due to the diverse features [11,12]. A variety of SB complexes with heterocyclic moieties such as semicarbazone, thiosemicarbazone, 1,2,4-triazoles, pyrazoles, 4-aminoantipyrene, benzoxazoles, coumarins, and triazines have attracted a lot of attention [13–17]. Owing to the prospective applications in analytical chemistry, dying industry, food industry, pharmaceutical industries, and agrochemical endeavours, SB ligands and their metal chelates have been accounted repeatedly [18]. This review aimed to describe the synthesis of SB and their metal complexes along with the metal–ligand stoichiometry. The review also covers the interaction of the Co(II), Ni(II) and Cu(II) complexes affecting bio-activities.Table 1. The synthesis of SB under solvent free or without solvent can proceed in the presence/absence of catalyst [19]. These are as followed:Microwave-assisted synthesis is a green chemical method. It is a beneficial technique in the organic synthesis due to its simplicity, responsive and reducing hazard capacity. It can often minimize reaction time under solvent-free or less-solvent conditions, resulting in higher yields and easier work-up in comparison to traditional methods [20] . The synthesis of SB under non-solvent environment [21] has been described below ( Scheme 1 ).As the reaction mixture is mashed in a mortar-pestle, the formation of SBs can be achieved effectively at room temperature using a catalyst such as SnCl2 and CH3COOH.In this process, mixture of primary amine and aldehyde/ketone is well mixed with the help of mortal-pastel. The reaction takes two to three minutes to complete.Methanol and ethanol are suitable solvents which are generally used in the synthesis of SB ligands followed by the refluxing in acidic, basic or neural medium. The purification of the product is done by either recrystallization or chromatography (TLC/column) techniques. Schemes 2 and 3 provide the reaction mechanisms in acidic and basic medium, respectively [22].In 2019, Vinusha and co-workers described the synthesis of SB ligand 3 by taking 5-amino-4H-1,2,4-triazole-3-thiol 1 and 3-hydroxy-4-methoxy benzaldehyde 2. The metal (Co(II), Ni(II) and Cu(II)) chloride salts were taken in 2L:1M stoichiometric ratio for the synthesis of metal complexes 4–6 ( Scheme 4 ). The structural analysis of 3 and its metal chelates was carried out by using proton/carbon NMR, mass, FT-IR and TGA spectroscopic techniques. The analysis results confirmed tridentate nature of 3 and octahedral geometry of 4–6 complexes. All compounds were examined for in vitro antibacterial activity by employing agar well diffusion method with nine food pathogens which included three gram + ve bacteria (B1, B12, B14) and six gram –ve bacteria (B2, B4, B9, B15, B16, B17). DMSO and Amoxicillin were utilized to control negative and positive action, respectively. All the gram +ve and gram –ve bacteria showed inhibition zone with compounds 4–6. Whereas 3 showed better inhibition against B1, B2, B4, B12, B14 and B15 [23].In 2020, SB 9 was synthesized by taking pyrazine-2-carbohydrazone 7 and 2-hydroxy-5-methylacetophenone 8. Co(II) 10, Ni(II) 11 and Cu(II) 12 complexes with the general empirical formula [M(L)(Cl)(H2O)2] were also reported ( Scheme 5 ). Microanalytical, magnetic susceptibility and various spectroscopic techniques such as proton/carbon NMR, IR, P-XRD, SEM, ESR and TGA were employed for the characterization of the compounds. The spectroscopic analysis validated the tridentate donor behavior (ONO) of 9. The molar conductance and physico-chemical suggested the non-electrolytic behavior and monomeric octahedral geometry of 10–12, respectively. All the compounds were screened for their antimicrobial activity using disc-agar diffusion method. Fungal strains F1 and F2 were used for antifungal screening and clotrimazole was used as standard drug. The results confirmed good antifungal activity of 9–12. In vitro antibacterial screening was carried out by taking gram +ve (B1, B3), gram –ve (B2, B18) strains. Ciprofloxacin, an antibacterial standard drug was used for the comparison purpose. The antibacterial actions of 10–12 complexes were more prominent against gram +ve as compare to gram –ve strains. The order of the antibacterial activity was 11 > Ciprofloxacin > 12 > 10. The chelation was found to be responsible for the enhanced antimicrobial activity of the 10–12 than 9 [24].In the same year, Dhanaraj and Raj [25] described the synthesis of SB ligand 17 in two steps. First step included the refluxing of ethanolic solution of 4-aminoantipyrine 13 and acetamide 14 to yield 4-aminopyridine derivative 15. In the second step, ethanolic solutions of p-phenylenediamine 16 and 15 were refluxed in order to get the SB ligand 17. The acetate salts of Co(II), Ni(II) and Cu(II) were used for the synthesis of complexes 18, 19 and 20, respectively ( Scheme 6 ). Compounds 17–20 were analyzed by elemental and several spectroscopic techniques such as mass, IR, UV–Visible and XRD. The nano-crystalline structure of 17–20 was confirmed by results. The complexes 18–20 were analyzed for DNA binding and cleavage activities. Antimicrobial studies were carried out against some bacterial (B1, B2, B3 and B4) and fungal (F1 and F2) species using agar well diffusion method. Cytotoxicity and anticancer action of all the compounds were also tested against L929 fibroblast cell line and SK-MEL-28 cell line, respectively. The compound 20 possessed the maximum potential for DNA binding, DNA cleavage and antimicrobial assay in comparison to 17, 18 and 19.In 2020, the synthesis of quinoxaline-based ligand 27 was carried out in multiple steps. Initially, aqueous solution of o-pheneylenediamine 21 and oxalic acid 22 was heated at 100°C to obtained 1,4-dihydro-quinoxalin-2,3-dione 23. The next step involved the refluxing of 23 and ethylenediamine 24 for 2 hrs, resulting 3-[2-(aminoethyl)amino]quinoxalin-2(1H)-one 25 which after reaction with salicylaldehyde 26 precipitated 27. The Co(II) 28, Ni(II) 29 and Cu(II) 30 metal chelates were also synthesized (Scheme 7 ) and characterized using FTIR, ESR, NMR, EDAX, magnetic moment, conductance, analytical and electronic spectral data. The ligand 27 acted as OO bidentate donor and coordinated through carbonyl oxygen and phenolic oxygen of quinoxaline ring. Complex 30 was analyzed theoretically. The compounds were tested for in vitro antimicrobial efficiency by well diffusion technique against gram +ve (B1 and B3), gram –ve (B4 and B5) bacteria and fungul (F1 and F2) strains. The compounds were also analyzed for their anticancer, in vitro antioxidant and DNA binding studies. The complexes 28 and 29 were found potent against human breast cells lines MCF7. The antioxidant activities of 28–30 were found moderated [26].A novel SB ligand 33 was synthesized [27] via condensation reaction of methyl-6-acetamide-2-amino-4,5,6,7-tetrahydrobenzo[b]thiophene-3-carboxylate 31 and 2-hydroxy-3-methoxybenzaldehyde 32. Metal chelates were also prepared by refluxing 33 with CoCl2·6H2O and NiCl2·6H2O ( Scheme 8 ). The spectroscopic analysis confirmed octahedral geometry for Co(II) 34 and Ni(II) 35 complexes. The compounds 33–35 were screened for their antioxidant activity by four different methods i.e.,(a) FTC, (b) FRAP method, (c) DPPH free radical scavenging activity using Blosis method and (d) reduction force determination using CUPRAC method. The results confirmed better antioxidant activity of 34 and 35 in comparison to 33.Alothmanet al.,[28] described the synthesis of 38 through condensation reaction of 1-aminoquinolin-2(1H)-one 36 and 2-hydroxybenzaldehyde 37 along with its nano-sized Co(II) 39, Ni(II) 40 and Cu(II) 41 complexes ( Scheme 9 ). The structure elucidation was carried out via various elemental and spectroscopic techniques, confirming the neutral (NOO) tridentate nature of 38 and octahedral geometry of 39–41. The anticancer potential of 38–41 was carried out against calf thymus CT-DNA through UV–Visible absorption process. To describe the DNA cleavage activity, solid-state DC electrical conductivity was measured at 576-696℃, confirming the semiconducting nature of 39–41. The potent cytotoxicity against Artemiasalina was possessed by 39 and 40 complexes with the value of LD50 = 2.68 × 10-6 and 2.74 × 10-6, respectively.In 2020, a new antipyrine based tridentate ligand 48 was synthesized into three steps [29]. Firstly, 2-hydroxy-3-methoxybenzaldehyde 42 and 4-aminoantipyrin 43 were refluxed to obtain compound 44. Another compound 47 was synthesized through the reaction of 3-nitrobenzaldehyde 45 and hydrazine hydrate 46. Both the compounds 44 and 47 were refluxed to yield 48. Different complexes of Co(II) 49, Ni(II) 50 and Cu(II) 51 were also reported via condensation pathway ( Scheme 10 ). Various techniques such as magnetic susceptibility, molar conductance, elemental analysis, UV–Visible, FT-IR, mass spectroscopy, proton/carbon NMR and TGA were employed for the characterization of 48–51. The analysis confirmed the octahedral geometry of 49 and 50 and distorted octahedral geometry of 51. In vitro antimicrobial actions against gram +ve (B1, B7 and B8), gram –ve (B2 and B4) bacterial and fungi (F1, F2, F3, F4 and F5) strains using Broth micro dilution method were evaluated confirming the considerable activity of 48–51. The complexes were also tested for the screening of anticancer efficiency against liver bilobular cancerous cells (LBir2754), SOD efficacy and DNA cleavage.In the same year, a reaction was carried out by refluxing glycylglycine 52, 4-nitrobenzaldehyde 53 and metal salts. The Co(II) 54, Ni(II) 55 and Cu(II) 56 transition complexes ( Scheme 11 ) were characterized using molar conductance, electronic spectra, magnetic moment and various spectroscopic methods such as TGA, ESR, P-XRD, IR and NMR, confirming octahedral geometry for 54–56. In vitro antimicrobial screening was performed against B1, B2, B3, B9, F1, F2 and F3 employing disc diffusion process. DNA cleavage against E.coli DNA and anticancer efficiency against colon cancer cells and human cervical cell lines were evaluated. The complex 55 was found effectively potent in comparison to 54 and 56 [30].A novel SB ligand 59 was synthesized by the condensation of o-phenylenediamine 58 and 5-acetyl-4-hydroxy-2H-1,3-thiazine-2,6(3H)-dione 57. The Co(II) 60, Ni(II) 61 and Cu(II) 62 complexes were also prepared by using condensation pathway ( Scheme 12 ). The elemental analysis, magnetic moment, molar conductance and spectral analysis (IR, proton/carbon NMR, ESR, TGA, mass) were done. The monobasic tridentate nature of 59 was suggested due to the presence of three donor sites (phenolic oxygen, azomethine nitrogen and nitrogen of amino –NH2 group). 61 and 62 complexes possessed octahedral geometry whereas complex 60 acquired square planner geometry. Compounds 59–62 were used for in vitro antimicrobial screening against B1, B10 strains and fungus F2 using disc agar diffusion method. The antitumor action of 59 and 60 was also tested against human hepatocelluar carcinoma cell [31]. Complex 62 showed highest antitumor activity than 59 with the IC50 value 120 µg/mL.A naphthalene–functionalized SB 65 was reported and synthesized using 2-hydroxy-1-naphthaldehyde 63 and o-phenylenediamine 64 via condensation along with Co(II) 66, Ni(II) 67 and Cu(II) 68 chelates (Scheme 13 ). The stoichiometry and structure elucidation were done on the basis of elemental analysis and spectroscopic methods (UV–Visible, FTIR, NMR, and mass). The analysis confirmed the tridentate NNO donor nature of 64 and six-coordinated geometry of 66–68. Microanalytical study verified 1:1 stoichiometry ratio of metal and ligand. All the compounds 65–68 were tested for antimicrobial, anticancer and cytotoxic activities. The antimicrobial screening was carried out against B1, B6, B2, B11, F1 and F6 strains employing disc diffusion method and anticancer activity was performed against colon carcinoma HCT-116 cell lines. The compounds 66–68 were found to be more active than 65 [32].Alothman et. al. [33] in the same year, refluxed the ethanolic solutions of 1,8-diamino-3,6-dioxaoctane 69 and 3,5-dichloro-salicylaldehyde 70 to synthesize a novel SB ligand 71. Co(II) 72, Ni(II) 73 and Cu(II) 74 complexes of 71 were prepared via facile synthesis strategy (Scheme 14 ). The analytical and spectroscopic analysis confirmed the hexadentate donor nature of 71. Absorption (UV–Visible) studies confirmed the octahedral geometry of 72-74 complexes. Agar well diffusion was used to check in vitro antimicrobial activities against three bacteria B1, B2, B9 and three fungi F1, F2 and F3strains. In vitro cytotoxic analysis was also carried out against MCF7 cancer cells. The experimental analysis confirmed complex 74 to be more active as compared to 71–73. Hence, complex 74 possessed enough potential to use as anticancer agent.In 2020, ethene-1,2-diamine based ligand 78 was prepared bycondensation of furan-2-carbaldehyde 75, 1,2-ethenediamine 76 and 2-hydroxybenzaldehyde 77. Co(II) 79 and Ni(II) 80 complexes were also prepared. The structural analysis was performed using UV–Visible and magnetic moment analysis, suggesting octahedral geometry of 79 and 80 (Scheme 15 ). The compounds 78–80 were tested against B1, B2, F2 and F4 strains for antimicrobial potential. The poor inhibition zone (disc-agar diffusion method) confirmed the inactiveness of 78–80 against F4 strain, whereas 79 was moderately efficient and 80 was inefficient against F2. The metal coordination inability and low lipophilicity of 79 and 80 were responsible for the low antimicrobial efficiency of the complexes. Cytotoxicity analysis of 78–80 was carried out against breast carcinoma (MCF7) and contra liver carcinoma (HEPG2).Complex 80 demonstrated the admirable anticancer activity as compared to 78 and 79 with higher LD50 values. The DNA binding ability of complexes was also performed with calf thymus DNA (CT-DNA) and analyzed using viscosity and absorption methods. The complex 80 intensively bounded with CT-DNA as per the results [34].In the same year, a new class of SB ligand 83 was synthesized by refluxing 2-aminothiophenol 81 and 2-(p-tolyoxy)-quinoline-3-carbaldehyde 82. The Co(II) 84, Ni(II) 85 and Cu(II) 86 metal complexes were also synthesized (Scheme 16 ) and characterized via various spectroscopic (IR, UV–Visible, mass, TGA, XRD, SEM and EDX) techniques [35]. Elemental and magnetic moment data confirmed 1:1 ligand and metal stochiometric ratio. In vitro antimicrobial activity and DNA cleavage study were carried out to check the potential of 83–86. Bacterial (B1, B2, B4 and B11) and fungal (F1 and F7) strains were used for microbial activity using cup plate agar diffusion method. The order of antibacterial activity was and 85 > 86 > 84 against B2 and B4, 86 > 85 > 84 against B11 and 86 > 84 > 85 against B1. The antifungal potential order was found to be 86 > 85 > 84 against F1 and F7. The DNA cleavage study was performedusing agarose gel electrophoresis method and pUC18 DNA was used in cleavage process. Complexes 86 and 84 were abled for DNA cleavage and no cleavage was noticed with 83 and 85.Daravath et. al. described the synthesis of a thiazol-based ligand 89 by refluxing methanolic solution of 6-aminobenzothiozole 87 and 5-hydroxysalicylaldehyde 88. A series of Co(II) 90, Ni(II) 91 and Cu(II) 92 complexes were also synthesized via condensation reaction (Scheme 17 ). Elemental analysis and various spectroscopic techniques reported the square planer geometry of 90–92 in 1:2 stoichiometry (metal:ligand) ratio. Metal complexes were examined for CT-DNA binding and cleavage of Pbr322-DNA by using UV–Visible absorption, fluorescence titrations and agarose gel study, respectively. The antimicrobial activity of 89–92 were carried out against two gram +ve (B1 and B14) and gram –ve (B2 and B4) bacterial strains. The antifungal screening was carried out against F8 and F9 by using paper disc method. The results showed that among all the complexes, 92 showed the higher degree of DNA cleavage, binding and antimicrobial activities [36].In 2020, a series of Co(II) 96, Ni(II) 97 and Cu(II) 98 complexes of new ligand 95 was synthesized by performing template reaction [37], in which methanolic solution of metal salts were added dropwise into the methanolic solution of o-vanillin 93 and glycine 94 and refluxed for two hrs (Scheme 18 ). The structure elucidation was carried out by using FT-IR, NMR, UV–Visible techniques and X-ray crystallography. Tridentate dianionic donor nature of 95 and octahedral geometry of 96–98 was confirmed by the data. All the compounds 95–98 were assessed for their molecular docking and DNA cleavage activities. DNA cleavage activity was carried out against E.coli genome through agarose gel electrophoresis process. Tyrosine Kinase (1 T46) and EGFR (1 M17) were the two targets that were used in molecular docking study. It was concluded that 98 showed the highest activity for DNA cleavage as compared to 95–97.In the same year, the Cu(II) complexes 104 and 105 of two SB ligands were synthesized by Medani and co-team. Ligand 101 and ligand 103 were synthesized by refluxing 2-phenylacetohydrazide 99 with 2-hydroxyacetophenone 100 and 1-hydroxy-2-napthaldehyde 102, respectively (Scheme 19 ). The structure elucidation was carried out using various elemental and spectroscopic methods. The X-ray study suggested the formation of binuclear complexes with 101 and 103. Compounds 101, 103–105 were screened for in vitro antimicrobial activity against bacteria (B1 and B2) and fungi (F2 and F3) stains. Ampicillin and Amphotericin B were used as standard antibacterial and antifungal drug, respectively. The antioxidant survey was also done through DPPH free radical scavenging method. On the basis of the antimicrobial screening (agar well diffusion method), antioxidant analysis and molecular docking studies, confirmed the DNA binding ability of 101, 103–105. Complexes 104 and 105 also exhibited better potential for DNA interaction as compared to 101 and 103 [38].Three novel hydrazone SBs named as (HL1=(E)-N'-(pyridin-2-ylmethylene)benzohydrazide 106, H2L2=(E)-2-(2-hydroxybenzylidene)hydrazine-1-carboxamide 107 and HL3=(E)-2-(pyridin-2-ylmethylene)hydrazine-1-carboxamide 108 (Scheme 20 ) and their Co(II) 109, Ni(II) 110 and Cu(II) 111 complexes were synthesized by Fekri et al. [39]. The compounds were analyzed using numerous techniques and tested for in vitro antibacterial as well as anticancer activities. Antibacterial activities were performed against B1, B2, B3 and B4 employing minimum bactericidal concentration and minimum inhibitory concentration methods. The anticancer activity was done against human colon cancer (SW742) and human gastric cancer (AGS) cell lines using MTT assay. The results illustrated the higher efficiency of 109 – 111 for anticancer and antibacterial activities.In the same year, Salicylaldehyde based three novel SB ligands (123–125) with aromatic systems and aliphatic spacers were synthesized successfully via diamine precursors taking dinitro compounds 112 and diamine compounds (113–115) to yield intermediates (116–118) and diamine precursors (119–121) [40]. These percussors and salicylaldehyde (1 2 2) were refluxed in ethanolic medium to give the respective SBs and further used (Scheme 21 ) in the synthesis of Cu(II) complexes (126–128). Firstly, The FT-IR data showed the appropriate chelation sites of 123–125 for the metal ions. The p-p* transition in phenyl ring and n-p* transition of azomethine group in the Cu(II) complexes were confirmed by the electronic absorbance spectroscopy. The biological investigation (antibacterial, antitumor, DPPH free radical scavenging, brine shrimp and DNA cleavage activities) of 123–125, 119–121 and 126–128 were also studied. The antibacterial activity of 116–126 and 128 was done against two gram +ve (B1 and B12) and four gram -ve (B2, B15, B19 and B20) strains using well diffusion method. None of the tested compounds demonstrated appreciable antibacterial activity against these six bacterial strains. The ineffectiveness of the compounds was attributed to the microbial cell’s impermeability, which prevented the tested compounds from interacting with all the six bacterial strains. The antifungal activity of 116–126 and 128 was carried out against F1, F3, F10, F4 and F11. Furthermore, the growth of fungal strains was inhibited to various degrees by the compounds. The Potato disc antitumor analysis was used to test the efficiency of 116–126 and 128. Compounds 116-126 and 128 illustrated the significant tumor suppression in a concentration-dependent fashion. The cytotoxicity of the compounds was checked against Artemiasalina and confirmed cytotoxic nature of 119–121, 126 and 128 and non-cytotoxic nature of 116–118 and 123–125. DPPH free radical examination showed that complexes 123–125 were extremely antioxidant. The nature of the compounds 116-122 were concentration-dependent DNA protective whereas 126 and 128 caused a considerable harm to the plasmid-DNA at all concentrations.A SB ligand 4-bromo-2-(((3-(methylamino)propyl)imino)methyl)phenol 131 was synthesized by refluxing N-methyl-1,3-diaminopropane 129 and 5-bromosalicylaldehyde 130 in ethanolic environment along with its Cu(II) complex 132 (Scheme 22 ). The structures of 131 and 132 were analyzed by various techniques such as FT-IR, EPR, electronic, solvatochromic studies, single crystal crystallography and Hirshfeld surface analysis. 131 and 132 were tested for their in vitro cytotoxic and antibacterial activities. Cytotoxic activity was performed against Dalton’s lymphoma as cites cell, reporting 36 mg/m LIC50 value for 132. The antibacterial activity was carried out against B1, B2, B4 and B14 bacterial strains using well diffusion method. The complex 132 showed highest activity against B1 and B14 than B2 and B4 [41].In 2020, Co(II) 136 and Cu(II) 137 complexes of novel SB [(L) 2-((1H-Benzo[d]imidazole-4ylimmino)methylphenol] ligand 135 were synthesized through condensation of 2-aminobenzimidazole 133 and hydroxybenzaldehyde 134 in ethanolic medium (Scheme 23 ). The structure elucidation of 135–137 was carried out by magnetic moment, molar conductance and various spectral techniques (FT-IR, UV–Visible, AAS, proton NMR). The spectral analysis and DFT study confirmed the bidentate donor nature of 135 and octahedral geometry of 136 and 137. The compound 137 possessed maximum antimicrobial efficiency against B1, B2, B9 and F2 using agar well diffusion method. The CT-DNA binding using UV–Visible absorption studies demonstrated better DNA binding capacity of 137 than 135 and 136 [42].Kareem et. al., carried out condensation reaction between 1,7-bis-(4-hydroxy-3-methoxyphenyl)-1,6-heptadiene-3,5-dione (curcumin) 138 and amino ethylene piperazine 139 to prepare a novel SB ligand 140. The Co(II) 141, Ni(II) 142 and Cu(II) 143 complexes were also synthesized (Scheme 24 ) and structural analysis was carried out using several spectroscopic methods. The molar conductance signified the non-electrolytic behavior of 141–143. Micro analytical analysis confirmed the 1:1 stoichiometric ratio of 140 and its metal complexes. EPR and UV–Visible spectroscopic analysis ascribed the octahedral geometry of 141 and 142 while 143 has the square planer geometry. The ability to scavenge free radicals was tested using revised Brand-Williams methods which proved 143 to be more potential antioxidant agent than 141 and 142. Compounds 140–143 were also examined for in vitro cytotoxic activity against HeLa, KCL22 and MDA-MB231 cancer cell lines as well as normal PBMCs cells using MTT method. Complex 143 was found to be more effective on KCL22 and MDA-MB231 cell lines than 140–142 [43].A reaction of 2-aminobenzylalcohol 144 and ortho-vanillin 145 was carried out to produce a novel SB ligand ABOVL 146. Co(II) 147 and Cu(II) 148 complexes of ABOVL were also prepared (Scheme 25 ). The analytical and spectral techniques have been employed to analysis the structure of 146–148. The spectral data confirmed the bidentate (O, N) donor nature of 146, octahedral and square planer geometry of 147 and 148, respectively. The CT-DNA binding analysis was achieved using absorption, viscosity and fluorescence data. The absorption values of 147 and 148 were 6.24 ± 0.04 × 104 M−1 and 5.76 ± 0.03 × 104 M−1, respectively. The viscosity experiments also disclosed the CT-DNA binding with 147 and 148 via intercalation. Furthermore, the data of Ksv (Stern-Volmer quenching constant) for 148 and 147, obtained from fluorescence experiment were 3.99 × 103 M−1 and 3.21 × 103 M−1, respectively. All the above analysis confirmed potential binding of CT-DNA with 148. Antimicrobial activity against B2, B21, B22, B23 bacterial and fungal F8 and F9 strains using agar diffusion method confirmed better microbial inhibition of 147 and 148 than 143. In vitro cytotoxic analysis was carried out against murine melanoma cancer cells (B16F10), human pancreatic carcinoma (MiaPacac2) and human cervical adenocarcinoma (HeLa) tumor cells line. cis-platin was used as control system using MTT assay. 148 demonstrated the maximum inhibitory efficiency with the higher IC50 value 49.13 mg/mL than 147 [44].A novel SB ligand (4-nitrophenylimino)methyl)benzylideneamino)phenol 152 originated from the condensation reaction of 2-aminophenol 149, 1,3-isopthaldehyde 150 and 4-nitroaniline 151. The Co(II) 153, Ni(II) 154 and Cu(II) 155 complexes of 152 were also synthesized (Scheme 26 ). The structure analysis was carried out using molar conductance, elemental and spectroscopic techniques. The tridentate nature and octahedral geometry were confirmed for 152 and 153–155, respectively. Utilizing the cyclic voltammetric and absorption studies, the interaction of 153–155 with CT-DNA (calf thymus) was investigated. The findings of the experiments showed that complexes can bind with CT-DNA in the intercalation manner. DNA cleavage analysis was also carried out against Pbr322 plasmid DNA by utilizing the agarose gel electrophoresis analysis in H2O2. The DNA cleavage data confirmed the inactive nature of 152 and 154 whereas 153 and 155 showed higher DNA cleavage tendency into linear and open circular manner. In vitro antimicrobial screening utilizing paper disc assay against bacterial (B1, B2, B3 and B18) and fungal (F1, F2, F12 and F13) strains, illustrated better antimicrobial activity of 153–155 than 152 [45].Two SBs(Z)-1-(1H-benzo[d]imidazol-2-yl)-N-benzylidenemethanamine 159 and 1-(1H-benzo[d]imidazol-2-yl)-N-(4-nitrobenzylidene) methanamine 160 were obtained from the reaction of 2-(aminomethyl)benzimidazoledihydrochloride 156 with benzaldehyde 157 and 4-nitrobenzaldehyde 158 to yield respective ligands (159 and 160). Co(II) 161, Ni(II) 162 and Cu(II) 163 complexes were also reported (Scheme 27 ) and analyzed via different physicochemical and spectroscopic techniques. The square planer geometry of 161–163 were confirmed on the basis of magnetic susceptibility, UV–Visible and molar conductivity studies. Compounds 161–163 were tested for CT-DNA binding using fluorescence spectroscopy, absorption, viscosity, circular dichroism and cyclic voltammetry. Moreover, the metallo nucleases properties of 161–163 were reported using agarose gel electrophoresis assay. In vitro antimicrobial activity using disc diffusion assay against bacterial (B1, B2, B3, B4 and B18) and fungal (F1, F2, F11, F12 and F13) strains confirmed better inhibition of 161–163 than 159 and 160. Complex 163 exhibited highest efficiency against all microbial strains as compare to 161 and 162 [46].Two novel SB ligands, HPSL 167 and HPSA 168 were prepared by refluxing sodium-5-sulfonate-2-hydroxybenzaldehyde 164, two forms of amino acids i.e., D,L-leucine 165 and phenylalanine 166 in aqueous medium. Water soluble Cu(II) complexes, Cu-HPS 169 and CU-PSA 170 were also prepared (Scheme 28 ). The structural analyses established tridentate dibasic chelating nature of 167 and 168. Magnetic susceptibility analysis confirmed the paramagnetic behavior and square planar geometry of 169 and 170. Compounds 167–170 were screened for their antimicrobial, CT-DNA binding and anticancer activities. The antimicrobial screening using agar well dilution method against bacterial (B1, B2 and B24) and fungal (F2, F3 and F14) pathogens strains against the standard drug Gentamycin and Fluconazole, respectively. The results demonstrated the better antimicrobial activity of 169 and 170 than 167 and 168. CT-DNA binding of 167–170 was done using viscosity and gel-electrophoresis. Compounds 167–170 possessed low lipophilicity owing to the presence of salting (Na-sulfonato) groups, resulting in significant electrostatic interaction with CT-DNA. The order of the interaction was HPSL < HPSA < Cu-PSL < Cu-PSA. The anticancer activity was performed against breast carcinoma (MCF7), colon carcinoma (HCT-116) and hepatocellular carcinoma (HepG2) cells. The findings confirmed the viability of 169 and 170 as anticancer therapeutic candidates [47].A new SB ligand (2-((E)-(4-trifluoromethoxy)phenylimino)methyl)-6-tert-butylphenol 173 was synthesized in methanolic medium condensation of 3-(tert-butyl)-2-hydroxybenzaldehyde 171 and 4-(trifluoromethoxy)benzenamine 172. [Co(L)2(H2O)2] 174, [Ni(L)2] 175 and [Cu(L)2] 176 complexes of 173 were also synthesized (Scheme 29 ) and characterized via different elemental and spectroscopic techniques. Compounds 175 and 176 possessed square planer geometry whereas 174 owned octahedral geometry with 2:1 (ligand:metal) stoichiometry ratio, which was also supported by single crystal XRD. Compounds 173–176 were subjected for their CT-DNA binding, Pbr322 DNA cleavage and antimicrobial (paper disc technique) activities. The effective CT-DNA binding with 174–176 was analyzed by viscosity measurements, fluorescence quenching and absorption spectroscopy. The order of binding compatibility was 176 > 175 > 174 > 173. The Pbr322 DNA cleavage activity was performed via gel electrophoresis method through oxidative and photolytic mechanism. Complex 176 reported efficient cleavage of Pbr322 supercoiled DNA into the liner or circular forms. The compounds 174–176 exhibited better antimicrobial inhibition against bacterial (B2 and B12) and fungal (F8 and F9) strains as compared to 173 [48].Co(II) 180, Ni(II) 181 and Cu(II) 182 complexes of neoteric salen-based quadridentate SB ligand 179 were obtained from the reaction of 4-fluoro-1,2-phenylenediamine 177 and 2-hydroxy naphthaldehyde 178 (Scheme 30 ). The characterization was carried out using thermal, elemental analysis and different spectroscopic methods. The electronic transition spectral analysis suggested the square planer geometry and 1:1 metal:ligand stoichiometry ratio of 180–182 whereas magnetic moment data suggested the paramagnetic and diamagnetic nature of 180 and 182 and 181, respectively. Compounds 179–182 were analyzed for their antimicrobial, DNA binding (fluorescence emission, UV–Visible absorption and viscosity techniques), DNA cleavage and antioxidant activities. Mancozeb and Streptomycin were utilized as standard drugs for antifungal and antibacterial activities, respectively by employing the disc diffusion method. Complexes 180–182 exhibited better antimicrobial potential against F8, F9, B1, B2, B9 and B12 strains than 179. Complex 182 showed the maximum potential for DNA binding than 180 and 181. The DNA cleavage action was performed against the Pbr322 DNA using photolytic and oxidative pathway by utilizing the agarose gel electrophoresis method. 180–182 were able to cleavage the DNA into its linear and nicked form and 179 was found to be unaffected into the DNA cleavage process. On the other hand, the antioxidant activity of 180–182 was performed using free radical DPPH scavenging method with the standard ascorbic acid. The IC50 value was maximum for 182 but less than the standard ascorbic acid [49].Rao and co-team [50] described the synthesis ( Scheme 31 ) and characterization of newly prepared Co(II) 186, Ni(II) 187 and Cu(II) 188 complexes of SB ligand 2-((E)-(6-ethoxybenzo[d]thiazol-2-ylimino)methyl)-4-nitrophenol 185, resulted from the reaction of 5-nitrosalicylaldehyde 183 and 2-amino-6-ethoxybenzothiazole 184. The ligand 185 exhibited the bi-dentate monobasic behavior and the complexes 186–188 possessed square planer geometry. All the compounds 185–188 were analyzed for their antimicrobial, DNA binding and DNA cleavage studies. CT-DNA binding was carried out using fluorescence emission, UV–Visible absorption and viscosity measurement techniques, confirming the interaction through intercalative manner whereas CT-DNA cleavage efficiency was checked using agarose gel electrophoresis process by utilizing H2O2 as oxidant reagent. The antibacterial (disc diffusion method) activity was also performed against the B4 ATCC-15380, B2 ATCC-25922, B1 ATCC-25923, B9 ATCC-12454 and B8 ATCC-35552 bacterial strains. The cytotoxic activity was analyzed using MTT-assay against MCF7 and HeLa cell lines, which confirmed the better efficiency of 186–188 towards both cell lines as compare to 185. The order of IC50 value was 188 > 186 > 187 > l85. Compound 188 showed the highest potential for all the biological (DNA binding and cleavage, antioxidant and antibacterial activities.A series of Co(II) 192 and Ni(II) 193 complexes was synthesized by the condensation reaction of respective metal salts and SB ligand 191 derived from the reaction of 1,2-diaminopropane 189 and 2-hydroxy-6-isopropyl-3-methyl-bezaldehyde 190 in 1:2 M ratio ( Scheme 32 ).The synthesized compounds were analyzed through elemental analysis, magnetic and electrochemical measurements, molar conductivity and various spectroscopic techniques such as SEM-EDX, cyclic voltammetry. Furthermore, single crystal-XRD technique confirmed a dimeric form with the empirical formula [NiL]2 and the distorted square planar geometry around the Nickel center for 193. Compounds 191–193 were analyzed for their antimicrobial, DNA cleavage and antioxidant activities. All compounds showed negligible antifungal activity against F1, F2, F3 and F15 than the standard drugs i.e., Fluconazole and Miconazole. Whereas, complex 192 exhibited maximum antibacterial activity against B1, B2, B3 and B9 strains using Ampicillin and Ciprofloxacin as standard drugs. Complexes 192 and 193 exhibited higher MIC values than Ampicillin but lower than Ciprofloxacin. 192 and 193 showed good antioxidant activity than191on comparing with ascorbic acid. The DNA cleavage activity was performed against Pbr322 DNA in the presence of H2O2 using agarose gel electrophoresis method. Complex 193 possessed maximum potential to convert the supercoiled form (I) into naked DNA form (II) [51].A novel SB ligand 196 was derived from the reaction of 4-methoxy salicylaldehyde 194 and 6-aminobenzothiazole 195. Co(II) 197, Ni(II) 198 and Cu(II) 199 complexes of 196 were also prepared using condensation pathway ( Scheme 33 ). All the compounds were characterized through various analytical and spectral techniques such as IR, UV–Visible, TGA and ESR. According to the data, all the complexes 197–199 possessed square planar geometry. The anticancer efficiency was checked against different cell lines such as cervical cancer cell (HeLa), breast cancer cell (MCF7) and adenocarcinomic human alveolar basal epithelial cells (A549) with cis-platin as standard drug. Complex 199 showed the maximum IC50 value. 97–199 were more active against HeLa cells than A549 and MCF7 cancer cells. The toxicity order was cis -platin > 199 > 198 > 197 > 196, due to the reduction of charge on metal ions which allowed the complexes to pass through the lipid layer of the cell membrane easily. The DNA binding study through intercalative mode via electronic absorption, viscosity and fluorescence quenching confirmed efficient binding of 199 with CT-DNA. On the other side, DNA cleavage action of 197–199 was performed against super coiled Pbr322 DNA using agarose gel electrophoresis method. 199 reported endorsed translations of super coiled Pbr322 plasmid DNA into linear form more efficiently. Complex 199 also possessed highest antibacterial efficiency against gram +ve B3 and gram –ve B2, B4 and B26 strains amongst all the synthesized compounds, using disc diffusion method. This was due to chelation leading the complexes to behave as effective bactericidal agents [52].In recent decades, almost every field of science and technology has experienced enormous advancements. Despite the advancements, there is still a long way to go until therapeutic interventions against microorganisms and cancer treatment have been explored. Due to the symptoms and medication obstruction, the direct access of antibacterial and anticancer drugs is restricted. Although there has been a notable advancement in our understanding of the subatomic causes of microbial disease and tumor growth, optimal therapeutic approaches are still lacking. In light of these facts, it is imperative to promote the development of novel antibacterial and anticancer agents. It is possible to fully address the usefulness of the various types of atoms (both ligands and metal complexes) for designing of new and effective antibacterial and anticancer agents. Investigating SB metal chelates with various subatomic features and topologies as antibacterial and anticancer agents is therefore essential. Additionally, emphasizing and launching techniques may be beneficial for human development to overcome the drawbacks of available operators. Therefore, we concentrated on the pharmacologically potent SB and their chelates. Notably, metallo-compounds shown more potential pharmacological efficacy than the conventional SB, necessitating more research. Tweedy's chelation theory and Overtone's idea were used by the authors to explain the enhanced biological efficiency. We have made an effort to emphasize the synergistic work and opportunities of SB derived metal complexes in accordance with the scope of the study. The designing and synthesis of novel SBs are incredibly significant for inventing unique medication libraries and eliminating the challenge of various drug resistances. Many SB metal complexes have shown excellent effectiveness against microorganisms and in the treatment of cancer, however none of these complexes have gone in-depth research or been released onto the market. This might be because people neither think SB complexes can be utilized as antibacterial/anticancer medications nor they do not have a professional interest in them. Researchers have been inspired to create synthetic strategies for the synthesis of innovative metal-frameworks based on SB science by the advantageous properties. An effective technique to obtain novel analogues with improved properties is to coordinate the carbonyl group by the SB reaction to amine compounds that represent various pharmacological groups. Many materials, including metal–organic frameworks, gels, porous organic cage, and nanocomposites, can be created through molecular self-assembly using a small subset of these substances. These assemblies currently have a broad variety of applications, from physics and materials science to pharmacology and health, and can be further tailored to simulate biological settings. This study covered various Co(II), Ni(II) and Cu(II) complexes with excellent DNA binding/cleavage, antimicrobial, anticancer and antioxidant inhibitory properties that may be useful for developing innovative therapeutic strategies for treating a variety of diseases. On the basis of this analysis, it is possible to draw the conclusion that creating SB metal complexes with crucial features is necessary in order to deliver medication to the product. Alka: Writing – original draft. Seema Gautam: Methodology. Rajesh Kumar: Formal analysis, Investigation. Prashant Singh: Visualization, Writing – review & editing. Namita Gandhi: Data curation. Pallavi Jain: Methodology, Validation, Supervision.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.The authors would like to thank SRM Institute of Science and Technology, NCR Campus, Modinagar, Ghaziabad, India for the guidance and support.

Schiff bases are versatile chemical compounds that are frequently utilized and manufactured by reacting various amines with carbonyl compounds (aldehydes/ketones), resulting in the formation of the azomethine/imine (–CN–) group by condensation reaction. Medical science is driven to generate novel drugs with revolutionary bioactivities and functionalities to cure diseases that are rapidly evolving. Schiff base (SB) is a dynamic pharmacophore that, through chelation, can create complexes with metals of various oxidation states. SB metal complexes have already been recognized as an effective branch of investigation in coordination science. In the recent years, scientists have paid close attention to SBs and the metal complexes owing to versatile potential in the pharmaceutics sector, such as antifungal, antibacterial, antiviral and antimalarial, anti-HIV, anti-cancer, anti-tuberculosis, and many others. These compounds have also been identified as potent oxidants, with applications in sensing and nanotechnology. The ligand environment, metal ion complexation, and lipophilic nature have an impact on the biological activity of transition metal complexes. SB metal complexes are the attractive targets for the development of broad-spectrum medicines due to their combination of pharmacological properties. The review focuses on the synthesis, spectroscopic characterization and in-vitro biomedical applications (antimicrobial, anticancer, antitumor, DNA binding and cleavage, antioxidant) of SBs as well as their Co(II), Ni(II) and Cu(II) transition metal complexes.

Due to dwindling reserves of easily accessible fossil resources and the increasing demand for fuels and chemicals, there is a growing need to develop efficient catalytic routes from renewable lignocellulosic biomass to fuels and chemicals. Cellulose, which is a renewable feedstock derivable, can be converted to platform chemicals such as 5-hydroxymethylfurfural (HMF) [1], levulinic acid (LvA) [2], and γ-valerolactone (GVL) [3]. In particular, GVL has attracted much attention in recent years, because it can be used in the chemical industry either directly or as an intermediate to food additives, nylon and green solvents [4]. GVL is also increasingly considered as a platform for the production of liquid biofuels [5].GVL can be obtained by gas- or liquid-phase hydrogenation of LvA using a suitable metal-based catalyst and a hydrogen source. In the past few years, many different catalyst systems including noble [6–8] and non-noble metals [9–13] have been evaluated for the hydrogenation of LvA to GVL. In addition to the use of renewable feedstock, it is also of increasing importance to develop catalysts that are based on cheap and abundantly available metals. Many catalysts used in the chemical industry are currently based on expensive precious group metals. Accordingly, there is a strong incentive to replace them with non-noble metals.The catalytic hydrogenation of LvA in the presence of molecular hydrogen using batch [14,15] and flow reactors [15–19] has been extensively studied in the last decade. The first example of catalytic hydrogenation of LvA was however already reported more than 50 years ago by Broadbent et al. [20] These researchers employed an unsupported Re black catalyst and were able to reach a 71 % yield of GVL after 18 h reaction at 106 °C at a H2 pressure of 150 bar. The remaining products were mainly polymeric esters. Afterwards, a wide range of supported noble metal catalysts featuring mainly Ru [7,18,21], Ir [22], Rh [23], and Pt [17] as the key hydrogenation components were evaluated for their activity in LvA hydrogenation. Ruthenium-based catalysts have emerged as promising candidates, because they typically combine high activity and selectivity. Non-noble copper-based catalysts such as Cu/ZrO2 [24], Cu/SiO2 [25], Cu-Cr [26], and Cu-Fe [27] have also been reported to be effective for producing GVL from LvA, although they typically require a higher reaction temperature and/or long reaction time for achieving high LvA conversion. Homogeneous catalysts have also been studied providing good results under relatively mild conditions. For instance, Yi and collaborators reported the hydrogenation of levulinic acid to GVL using a homogeneous Fe complex in aqueous solution, obtaining GVL yields as high as 97 % in 2 h at 100 °C and 50 bar [28].There is growing evidence of the positive effect of bimetallic catalyst formulations for the hydrogenation of oxygenated substrates [29,30]. For instance, the group of Weckhuysen reported on the beneficial effect of Ru-Au nano-alloying for the catalytic conversion of LvA to GVL [31]. Bimetallic catalysts containing noble and non-noble metals (i.e., Ni-Ru, Ni-Pt, Ni-Au, and Ni-Pd) supported on supports such as zeolite, ZrO2, γ-Al2O3, and SiO2 have also been employed for the upgrading of biobased intermediates derived from lignin [32]. Supported Ni-Re [33] and Pt-Re [34] catalysts have been shown to be highly active for the selective hydrogenation of carboxylic acids. Higher conversion and selectivity of carboxylic acid hydrogenation were achieved with a Ni-Re catalyst compared to its single-metal and Pt-based counterparts [33]. Recently, Ni/Al2O3, Ni-Cu/Al2O3, Ni-Nb/TiO2 and Ni/HZSM-5 were employed for the hydrogenation of LvA [15,19,35]. A high reaction temperature (220−275 °C) was however required to achieve reasonable performance. Shimizu and co-workers first reported a noble-metal-free Ni-MoOx/C catalyst with a TON (turnover number) of 4950 [36], which is comparable to a state-of-the-art Ru catalyst for the hydrogenation of LvA to GVL at 250 °C [37]. Grunwaldt et al. reported a solvent-free method to obtain a 92 % GVL yield for LvA hydrogenation using Ni/Al2O3 [14]. However, reuse of the Ni catalyst resulted in a significantly lower activity. Shimizu’s group reported that Re/TiO2 is a promising catalyst for the selective hydrogenation of aromatic and aliphatic carboxylic acids. In their study, 3-phenylpropanol was produced in 97 % yield from 3-phenylpropionic acid under mild conditions (50 bar H2 at 140 °C) [38].Here we report a novel TiO2-supported Fe-Re bimetallic catalyst system, which is highly active in the hydrogenation of LvA to GVL under mild conditions. The catalysts were extensively characterized using H2-TPR, XPS, XANES, EXAFS, Mössbauer spectroscopy, CO-IR spectroscopy and TEM. Strong interaction between Fe and Re was observed in terms of the formation of a Fe-Re-oxide phase, which upon reduction is partially converted into a metallic Fe-Re alloy covered by FeOx species. The interactions of Re and Fe with the titania support also play an important role in the formation of catalytically active nanoparticles. The strong synergy in levulinic acid hydrogenation is attributed to the interface between metallic Fe-Re particles and FeOx.All the supported catalysts were prepared by an incipient wetness impregnation method. Titania (P25 TiO2, Evonik-Degussa) was dried at 110 °C overnight, prior to impregnation of the metal precursor. For the preparation of Fe-Re bimetallic catalysts, appropriate amounts of Fe(NO3)3·9 H2O (≥98.0 %, Sigma Aldrich) and perrhenic acid (HReO4) (99.99 %, 75−80 wt% in H2O, Sigma Aldrich) precursors were dissolved in deionized water. Then, the required amount of titania was added very slowly under continuous stirring at room temperature. The sample was dried at 110 °C overnight, and ground thoroughly and reduced in a furnace at 500 °C for 2 h (ramp rate 2 °C/min) in a flow of 10 % H2/He (total 100 mL/min). Two monometallic reference catalysts 2.0 wt% Fe/TiO2 and 13 wt% Re/TiO2 were prepared using the same method.Temperature-programmed reduction (TPR) experiments were performed in a Micromeritics AutoChem II 2920 instrument equipped with a fixed-bed reactor, a computer-controlled oven, and a thermal conductivity detector. Typically, samples (50 mg) were loaded in a tubular quartz reactor. Prior to reduction, samples were pretreated at 150 °C for 2 h. The sample was reduced in 4 vol% H2 in N2 at a flow rate of 8 mL/min, whilst heating from room temperature up to 900 °C at a heating rate of 10 °C/min. The H2 consumption was monitored by a gas chromatography equipped with a thermal conductivity detector (TCD) and calibrated using a CuO/SiO2 reference catalyst.XPS measurements were performed using a Kratos AXIS Ultra spectrometer, equipped with a monochromatic X-ray source, and a delay-line detector (DLD). Spectra were obtained using an aluminum anode (Al Kα = 1486.6 eV) operating at 150 W. Survey scans were measured at a constant pass energy of 160 eV and region scans at 40 eV. The background pressure was kept at 2 × 10−9 mbar. Quasi-in situ XPS measurements for all of the catalysts were performed after reducing them in a tubular quartz reactor with 10 °C/min heating rate from room temperature to 500 °C in a flow of 10 vol% H2 in He (total flow 100 mL/min). After cooling to room temperature, the lids at the inlet and outlet of the reactor were closed to prevent air exposure. The samples were prepared for XPS measurements in an Ar-flushed glove box and transferred in an air-tight transfer holder to the XPS apparatus. Data analysis was performed using CasaXPS software. The binding energy was corrected for surface charging by taking the C 1s peak of adventitious carbon as a reference at 284.6 eV.X-ray absorption fine structure (QEXAFS) measurements were done at the Fe K-edge (∼7112 eV) and Re L-3 edge (10535 eV) in transmission mode on beamline BM26 at ESRF (DUBBLE, Grenoble). The photon flux of the incoming and outgoing X-ray beam was detected with two ionization chambers I0 and It, respectively. The obtained absorption data were background-subtracted, normalized and fitted as difference spectra using Athena software. EXAFS analysis was performed using VIPER on k3-weighted data. The amplitude reduction factor S02 was determined by fitting the first Re-Re coordination to 12, of Re foil. In a typical experiment, ca. 15 mg catalyst sample (in tablet form) was placed in a stainless-steel XAS reactor equipped with two fire-rods and glassy carbon windows as described in ref [39]. Catalysts were reduced in this cell by heating at a rate of 3 °C/min from 40 °C to 500 °C followed by an isothermal dwell of 0.5 h in a flow of 20 vol% H2 in He at a total flow rate of 50 mL/min. During reduction, the state of the samples was followed by XANES, while EXAFS spectra were recorded at 50 °C after the reduction.Transmission 57Fe Mössbauer spectra were collected at -153.2 °C with a sinusoidal velocity spectrometer using a 57Co(Rh) source. Velocity calibration was carried out using an α-Fe foil at room temperature. The source and the absorbing samples were kept at the same temperature during the measurements. The Mössbauer spectra were fitted using the Mosswinn 4.0 program.Low-temperature infrared spectra of CO adsorbed on the catalysts was recorded using a Bruker Vertex V70v FT-IR spectrometer. The IR spectra were acquired at a resolution of 2 cm−1 and 32 scans were averaged for each spectrum. Typically, an amount of ca. 20 mg catalyst was pressed into a thin self-supporting wafer with a diameter of 13 mm, which was then placed inside a controlled-environment IR transmission cell capable of heating and cooling, gas dosing, and evacuation. Prior to CO adsorption, the catalyst wafer was reduced at 500 °C for 1 h in flowing 10 vol% H2 in He, followed by cooling to 100 °C. The cell was then evacuated to ∼10-6 mbar and further cooled to liquid nitrogen temperature. The sample was then subjected to pulses of CO via a sample loop (10 μL) connected to a six-port sampling valve. CO was pulsed until saturation was reached as observed by saturation of the CO IR adsorption bands.Transmission electron micrographs were acquired on a FEI cubed Cs corrected Titan at 300 kV. Typically, a small amount of the sample was ground and suspended in pure ethanol, sonicated and dispersed over a Cu grid with a holey carbon film. Samples were firstly reduced in 10 vol% H2 in He (total flow 100 mL/min) at 500 °C for 2 h, followed by passivation in 1 vol% O2 in He for 10 h. Bright field images (BF) were taken using a rather large objective aperture to enhance the contrast, specifically for lattice imaging. HAADF-STEM imaging was done to analyze the particle size. Elemental analysis was done with an Oxford Instruments EDX detector X-MaxN 100TLE.Aqueous-phase catalytic hydrogenation of LvA to GVL was performed in a 10 mL autoclave (HOKE Swagelok) at various temperatures (130–200 °C) and a (cold) H2 pressure of 40 bar. In a typical reaction, 2 mmol LvA and 23 mg reduced catalyst were loaded into the autoclave in a nitrogen-flushed glove-box. The autoclave was sealed using a rubber plug before removing it from the glove-box. An amount of 4 mL degassed water was injected into the autoclave via the rubber plug using a syringe. The autoclave was then sealed and purged four times with H2 before the pressure was increased to 40 bar. The reaction was started by heating the autoclave to the desired reaction temperature under continuous stirring (1000 rpm). At the end of the reaction, the autoclave was cooled rapidly to room temperature in an ice bath, after which the remaining H2 was released. The catalyst was separated from the solution by filtration (0.45 μm filters). The reaction products were subjected to NMR analysis.Quantitative analysis of the liquid products (LvA and GVL) was carried by 1H-NMR using 1,4-dioxane as an internal standard. An amount of 100 μL 1,4-dioxane was added to the reaction mixture after the catalytic reaction. An aliquot of 300 μL of the reaction mixture was transferred to a 5 mm NMR tube together with 300 μL deuterated dimethylsulfoxide-d6 (DMSO-d6) solvent. For quantitative 1H NMR analysis, 32 scans were averaged using a relaxation delay of 5 s. All spectra were integrated using MestReNova software.The conversion of LvA ( X ) was calculated as follows: X   % = C L v A , 0 − C L v A C L v A , 0 × 100 % The yield of the liquid component i ( Y i ) was calculated as follows: Y i   % = C p r o d u c t   i C L v A , 0 × 100 % A set of Fe-Re catalysts supported on TiO2 with different atomic Fe/Re ratios were prepared by wetness impregnation. The Fe loading for all of the Fe and Fe-Re catalysts was kept at 2.0 wt%. The atomic Fe-to-Re ratio was varied between 5:1 and 1:2. The bimetallic catalysts are denoted as Fe-Re(x:y)/TiO2 in which x:y stands for the atomic Fe/Re ratio. Monometallic Fe-2.0 wt%/TiO2 and Re-13 wt%/TiO2 (denoted as Fe(2.0)/TiO2 and Re(13)/TiO2, respectively) were prepared in the same way and served as reference catalysts.We firstly screened these catalysts for their performance in the hydrogenation of LvA to GVL. For this purpose, the catalysts were reduced at 500 °C for 2 h and then tested in a batch reactor in water at 140 °C and 40 bar H2 for 4 h. 0 Fig. 1 compares the performance of the reduced catalysts. Fe(2.0)/TiO2 showed a very low GVL yield of less than 1%. The yield for Re(13)/TiO2 was 3%. A much higher catalytic performance was achieved using Fe-Re bimetallic catalysts. With Fe-Re(1:1)/TiO2 catalyst, a yield of 12 % GVL was obtained at the conversion of 14 %. The catalytic performance increased with increasing Re-to-Fe ratio. The best catalytic performance was obtained for Fe-Re(1:2)/TiO2, which gave 17 % yield of GVL at the conversion of 18 %. These results evidence a significant synergy between Fe and Re. The addition of Re to Fe strongly improved LvA conversion.TPR traces of the reduction of all catalysts are presented in Fig. 2 . The Fe(2.0)/TiO2 sample shows a very small reduction feature around 285 °C, which is due to the (partial) reduction of Fe3+ to Fe2+. A broad feature around 650 °C can be attributed to the reduction of Fe2+ to metallic Fe [40]. The active phase in Re(13)/TiO2 sample is reduced at 350 °C. This suggests that it is easier to reduce Re than Fe on titania. For the Fe-Re samples, the main reduction peak becomes sharper and its position shifts to slightly higher temperatures compared with the Re-only and Fe-only samples. It is also noted that the first (partial) reduction peak at 285 °C, which is due to reduction of Fe3O4 or Fe2O3 to FeO, is not present for the bimetallic Fe-Re catalysts and the second reduction peak at 650 °C shifts to higher temperature (cf. the dashed line in Fig. 2). A previous study showed that higher reduction temperature for bimetallic Fe-Re/SiO2 catalysts compared to the monometallic ones can be attributed to a strong interaction between Fe and Re in the mixed oxide [40]. For the Fe-Re/TiO2 samples, besides the small reduction feature of Fe2+ → Fe°, a single main reduction feature suggests that Fe and Re are present in a mixed-oxide phase.Reduced Fe, Re, and Fe-Re samples were further characterized by XPS, which is a surface-sensitive technique. Fig. 3 shows the fitted Re 4f XPS spectra. Quantitative XPS data are collected in Table 1 . A wide range of oxidation states of Re between +2 and +7 is observed for the reduced catalysts. At low Re content in Fe-Re(5:1)/TiO2, the intensity of the Re signal was too low for reliable fitting. For Fe-Re(2:1)/TiO2, the Re2+ : Re4+ : Re5+ : Re6+ : Re7+ ratio was 16 : 16 : 10 : 21 : 37. No metallic Re was observed for this sample. Samples with a higher Re content, Fe-Re(1:1)/TiO2 and Fe-Re(1:2)/TiO2, contained both metallic and oxidic Re. For reduced Fe-Re(1:1)/TiO2, the metallic Re fraction was 53 % and the remainder was present as Re-oxide species with a large contribution of Re2+. The Re° content increased to 70 % for Fe-Re(1:2)/TiO2. It is important to mention that the metallic Re° content in the Fe-Re(1:1)/TiO2 and Fe-Re(1:2)/TiO2 samples are higher in the Re(13)/TiO2 sample (29 %). These results confirm that the presence of Fe resulted in a higher reducibility of the Re component in the reduced materials.We also analyzed the oxidation state of Fe in these samples by XPS. The XPS spectra in the Fe 2p region and the results of their deconvolution are shown in Fig. 4 and Table 1, respectively. The reduced Fe(2.0)/TiO2 sample contains a large amount of Fe2+ (78 %) with the remainder being Fe3+ (22 %). No metallic Fe was observed in this catalyst. This confirms that the first reduction peak in TPR is due to the partial reduction of Fe3+ to Fe2+. It is seen that the amount of Fe2+ is lower for the reduced Fe-Re(5:1)/TiO2, Fe-Re(2:1)/TiO2, and Fe-Re(1:1)/TiO2 samples in comparison to Fe(2.0)/TiO2. The finding that the Fe3+/Fe2+ ratios of these three catalysts are similar indicates that Re does not promote the reduction of Fe at a low Re content. Instead, its reducibility is decreased, which is likely due to the strong interaction of Fe in a mixed Fe-Re-oxide phase. On the other hand, at higher Re loading (Fe-Re(1:2)/TiO2), the fraction of Fe2+ in the reduced materials is already ∼85 %, suggesting that the presence of Re promotes the reduction of Fe. The well-known mechanism for this kind of reduction promotion is the spillover of H atoms from the first reduced metal phase to the other components [41,42]. The difference is obviously due to the formation of metallic Re at higher Re content. Since the Fe content is constant in our samples, this may imply that a certain fraction of free Re-oxide is needed to obtain metallic Re.We also investigated the reduction of Fe and Re in more detail by in situ XANES, which is bulk sensitive in contrast to XPS. The XANES spectra of the selected Fe-, Re-, and Fe-Re samples were collected at reduction temperatures in the 50−500 °C range under a 50 mL/min flow of a 10/40 v/v mixture of H2 and He. Reference materials including the corresponding metals and metal oxides of varying oxidation states were also measured. The energy of the half-edge-step was used to compare the oxidation state of Fe and Re during the reduction process. Fig. 5 a and c show the XANES spectra of Fe and Re reference samples, respectively. Fig. 5b and d show the energy corresponding to the half-edge-step for Fe and Re, respectively, during reduction, which is employed here as a qualitative indicator of the oxidation state. Although the pre-edge feature of Fe XANES spectra can be analyzed to analyze the oxidation and coordination state of Fe, [43] the data quality of the measured spectra was too low to extract useful information. Analysis of the energy of the half-edge-step shows a gradual reduction of Fe3+ to Fe2+ and Fe(0). The addition of Re increases the rate of Fe reduction. This is in line with the earlier finding that the addition of Re promotes the reduction of Fe3+ to lower oxidation state. On the other hand, while the Re(13)/TiO2 shows the highest reducibility, it is seen that the addition of Fe delays the reduction of Re to higher temperature. However, by comparing the energy at the half-edge-step of the Re-containing samples with the Re foil (Re°) reference, it can be concluded that all of the Re species can be reduced to their metallic state above 350 °C. This seems to be inconsistent with the XPS results, where a larger amount of oxidized Re was observed. This difference is likely due to the in situ character of the XANES measurements, while XPS was carried out in a quasi in situ mode, including cooling to room temperature and sample transfer at ambient conditions via a glove-box. Although air exposure was avoided, traces of oxygen and water will likely oxidize reduced surface Re species. Obviously, the difference is substantial also because XPS is a surface-sensitive technique, while XANES probes the bulk as well. Notably, the Fe-Re bimetallic catalysts gave lower energies at the half-edge-step than the Re(13)/TiO2 catalyst and the Re foil (Fig. 5d). This points to the formation of a Fe-Re alloy with a different electronic structure than reduced Re nanoparticles. The presence of this Fe-Re alloy can explain the higher reducibility of the Re component in reduced samples, revealed by XPS analysis.In order to obtain more detailed structural information of the Fe-Re samples, extended X-Ray absorption fine structure (EXAFS) data were collected both at the Fe K-edge and the Re L3-edge after reduction at 500 °C followed by cooling to 50 °C. Due to the low Fe content in combination with the strong X-ray absorption by the titania support, the Fe Ke-edge EXAFS data quality was too low for reliable fitting. Fig. 6 depicts the experimental and fitted k 3-weighted R-space spectra for the Re foil and the Re(13)/TiO2 and Fe-Re(1:2)/TiO2 samples. The EXAFS fitted parameters are summarized in Table 2 . The amplitude reduction factor was chosen such that the Re foil with its typical cubic closest packed structure has a Re-Re coordination number of 12. For the Re(13)/TiO2 sample, the first Re-Re shell coordination number due to a metallic Re-Re bond is significantly lower (CN = 7.2), implying the formation of nanoparticles. This sample also contains a first shell contribution from Re-O scattering at 2.072 Å with a coordination number of 1.1, indicating that the Re phase was not fully reduced. This may be attributed to the strong interaction with the titania support. The difference in Debye-Waller factors between this sample and the reference also hints at such interactions [44]. The fitting results for Fe-Re(1:2)/TiO2 are very different. First of all, the Re-Re bond distance is reduced compared with the other two Re-only samples and a second shell is seen, which can be fitted with a Re-Fe scatterer. The theoretically determined Re-Fe and Re-Re bond length values in a stoichiometric Re-Fe alloy are 2.587 Å and 2.689 Å, respectively. Both are shorter than the Re-Re bond length (2.75 Å) in Re(13)/TiO2 and the Re foil. The shorter Fe-Re bond length is due to the smaller size of the Fe atom with respect to the Re atom. The Re-Re bond length is 2.653 Å with a coordination number of 7.6, implying a similar coordination number as for the Re-only sample. The occurrence of a Fe-Re coordination of 1.2 at 2.453 Å indicates that a Re-Fe alloy is formed with a high Re/Fe ratio. The inclusion of a Re-O contribution did not significantly improve the fitting results for the Fe-Re(1:2)/TiO2 catalysts.Given the poor quality of the Fe K-edge XAS data, we used Mössbauer spectroscopy to characterize the Fe phase in our samples. The Mössbauer absorption spectra were recorded at -153.2 °C after reduction at a temperature of 500 °C. The obtained spectra are presented in Fig. 7 , while the Mössbauer fit parameters are summarized in Table 3 . The Fe(2.0)/TiO2 sample can only be reduced to metallic Fe to a small extent (6%). The remaining Fe is present as Fe3+ (bulky hematite, Fe2O3, 15 %), super-paramagnetic Fe3+ (small hematite particles, Fe2O3, 52 %), and Fe2+ (wüstite-like structures, FeO, 27 %). The hyperfine sextet with a hyperfine field of 51.9 T is characteristic for large bulky hematite particles, indicating that a part of the Fe phase sintered during the reduction. The sample containing Re had a higher Fe reduction degree, evidenced by the increase of the relative intensity of the Fe° singlet. The hyperfine sextet is no longer visible in the Re-containing samples, indicating that the sintering of Fe-oxide did not occur. This is very likely due to the interaction between Fe and Re. For the FeRe(1:2)/TiO2, the intermediate Fe2+ phase is not visible, which we take as an indication that only part of the Fe atoms can be promoted by Re and reduced at the applied temperature. The (remaining) non-reducible Fe3+ species are likely experiencing a strong interaction with the TiO2 support. The Fe reduction degree in this sample is close to 40 %. Given that the reduction degree probed by XPS was much lower, we speculate that the reduced Fe species might be part of a Fe-Re alloy. We cannot exclude however that part of the difference in Fe reduction degree can be due to the sensitivity of reduced Fe species to oxygen during the transfer from the XPS pre-chamber to the high-vacuum chamber.The reduced and passivated samples were analyzed on a Cs-corrected TEM. In all samples, nanoparticles were found on micron-sized agglomerates of titania particles. Although the bright-field images showed the presence of 1−2 nm nanoparticles, the active phase could not be clearly observed in this way. Accordingly, HAADF-STEM images were recorded as well from which the particle size distribution was determined. EDX maps and particle size distributions are shown in Figs. 8–11 . The average particle size for Fe-Re(2:1)/TiO2, Fe-Re(1:1)/TiO2, and Fe-Re(1:2)/TiO2 are roughly similar at 1.0 ± 0.4 nm. Re(13)/TiO2 contains slightly larger nanoparticles with an average diameter of 1.3 ± 0.8 nm. All samples contain few nanoparticles larger than 2 nm. Notably, the Re-only sample contains a fraction of significantly larger nanoparticles (cf. Fig. 11). By comparing Fe-Re(1:2)/TiO2 and Re(13)/TiO2 with the same Re content, it can be stated that the presence of Fe in the bimetallic catalyst results in a better Re dispersion.EDX mapping in STEM mode showed the presence of Fe and Re over the surface of the titania support. The reported quantitative Fe signals (determined by wide electron probe EDX) were corrected for signals due to secondary electrons using the Co signal. In general, the Fe/Re ratio in areas where no clear nanoparticles are visible is higher than in areas where Re nanoparticles can be observed (Table 4 ). For instance, the Fe/Re ratio on the titania surface for Fe-Re(2:1)/TiO2 is higher (∼2.0) than the Fe/Re ratio on the nanoparticle displayed in Fig. 8. The EDX maps of the nanoparticle shown in Fig. 8d–f also suggest a core-shell structure in which the shell contains more Fe than the core. The presence of Fe and Re across the titania surface is in agreement with recent aberration-corrected TEM images of a Ni-Re catalyst, which showed that Re is dispersed on the surface in the form of atoms, clusters, and nanoparticles [33]. In the present study, STEM can only image the nanoparticles. Accordingly, we can conclude that the titania-supported Fe-Re samples contain Re nanoparticles in close contact with Fe and very highly dispersed Fe and Re species homogeneously distributed over the titania surface.IR spectroscopy of adsorbed CO was used to investigate the active phase of the reduced catalysts. As XPS showed that the samples contain oxidic Re and Fe, we recorded the CO IR spectra at liquid N2 temperature. We also included the bare TiO2 support for comparison. Fig. 12 shows the IR spectra of CO adsorbed on the samples as a function of the CO partial pressure in the cell (0.02−1 mbar CO range). The IR spectra in the CO stretching region contain bands at 2178 cm−1, 2157 cm−1, and 2042 cm−1. According to the literature [45,46] CO stretching bands between 2200−2100 cm−1 can be assigned to CO adsorption on Lewis-acidic metal cations and OH groups, while metals in a lower oxidation state usually give rise to lower CO stretch frequencies. In particular, bands between 1900 and 2100 cm-1 can be assigned to linearly adsorbed CO on the surface of metals. For the bare TiO2 support, a sharp band at 2181 cm−1 increased with CO pressure concomitant with a red shift to 2178 cm−1. This feature is due to CO adsorbed on Lewis acidic Ti4+ sites. Another weaker band at 2156 cm−1 can be assigned to weaker Ti−OH···CO complexes [47]. The presence of surface OH groups is also evident from the OH stretching region. Similar IR spectra were obtained for Fe(2.0)/TiO2 (Fig. 12b). The presence of Fe on TiO2 (Fe(2.0)/TiO2) resulted in a lower intensity of the OH stretching bands in the 3500−3800 cm−1 region, suggesting that during the preparation Fe3+ has reacted with titania OH groups forming Fe-O-Ti species. It is seen that the more acidic OH groups are preferentially consumed. There are no indications of the presence of metallic Fe, consistent with the other characterization data (Fig. 4, Table 2). The spectra did not contain other bands than those observed for bare TiO2, which suggests that Fe is present as a dispersed Fe-oxide phase, which does not adsorb CO. Based on the reported OH density of this type of titania (4.5 OH/nm2), we can estimate that a Fe content of 2 wt% corresponds to about 80 % of the monolayer capacity.For the Re-containing samples, the intensity of the broad band at 2157 cm−1 is much higher than for TiO2 and Fe(2.0)/TiO2. Moreover, these bands already appear at a much lower CO coverage and, importantly, before the band at 2178 cm-1 appears. This completely different behavior suggests that a different and stronger CO adsorption complex gives rise to the 2157 cm-1 band in these samples. This is further underpinned by the strong erosion of the band due to OH groups in the 3500−3800 cm-1 range when Re is present. It is interesting to note that the OH stretch intensity becomes weaker with increasing Re content. On the other hand, the reduced Re(13)/TiO2 only contains a very weak band at 2158 cm-1 band, which must be due to the consumption of most of the weakly acidic Ti−OH groups. This is consistent with the 13 wt% Re loading corresponding to 1.8 monolayer coverage of the titania surface hydroxyl groups. All of the IR spectra of the Re-containing samples contain a broad band in the 2030-2048 cm-1 regime, which can be associated with Re°. The intensity of this band for the Fe-Re(1:2)/TiO2 sample is much weaker than for the Re(13)/TiO2 one. We also notice a red shift of the Re° feature with increasing Re content, which could be due to the close proximity of CO adsorbed on Re to FeOx and/or the formation of a bimetallic Fe-Re alloy.The CO IR spectra show that the Fe-Re(1:1)/TiO2 and Fe-Re(1:2)/TiO2 samples contain a lower amount of reduced Re° surface sites than Re(13)/TiO2. On the other hand, while TEM shows that the nanoparticles in these catalysts are approximate ∼1 nm, XPS and TPR point to a substantially higher reduction degree of Re in the bimetallic catalysts. Together with the Fe/Re ratios on the nanoparticles derived by STEM-EDX maps, we can conclude that the metallic Re particles are covered by small Fe-oxide clusters that partially block the reduced Re sites. A similar conclusion has been drawn in studies of related Fe-Re/SiO2 [40] and Pd-FeOx/SiO2 [48] catalysts.The CO IR spectra of the bimetallic Fe-Re catalysts contain a feature at 2157 cm−1, which shows a maximum at intermediate Fe/Re ratio. It is not likely that this feature is related to the highly dispersed Fe-oxide and Re-oxide species on the titania support, because the signal is absent for the monometallic catalysts. Therefore, we speculate that the 2157 cm−1 is due to Lewis acid cations, likely Fe cations, at the interface between metallic Re nanoparticles and a partially reduced Fe-oxide (Scheme 1 ).In attempting to explain the Fe-Re synergy, we compare the highly active Fe-Re(1:2)/TiO2 catalyst with the nearly inactive Re(13)/TiO2 one. The Re reduction degree is much higher for the bimetallic catalyst, demonstrating that reduced Re is the active phase for LvA hydrogenation in line with previous literature [20]. Our characterization data show that the oxides of Fe and Re strongly bind to the titania surface via reaction with the OH groups. When the total Fe + Re content is higher than the monolayer capacity of TiO2 as for Fe-Re(1:2)/TiO2, hardly any OH groups are observed in the IR spectrum. Therefore, the higher Re reduction degree in the bimetallic catalyst can be related to a fraction of Re-oxide species that are not bound to the titania surface. When the Re loading is too low, a significant fraction of Re remains strongly bound to the titania surface and cannot be reduced, leading to a low hydrogenation activity. Thus, we can conclude that a role of Fe is to compete for surface OH groups of the titania support and decrease the interaction of Re with the support, thereby resulting in a higher reducibility of Re. This is supported by the Mössbauer data, which shows that only 41 % of Fe can be reduced to the metallic state with the remainder being Fe3+ (Fig. 7, Table 2). Despite the higher Re reduction degree in bimetallic catalysts, the Re surface area probed by CO IR spectroscopy is small, which is due to the coverage of part of the Re nanoparticles with Fe-oxides (Scheme 1). It is less likely that a very small amount of reduced Fe forming an alloy can cause this, since the combined XPS and Mössbauer data suggest that most of the reduced Fe species are in the core of a bimetallic Re-Fe alloy phase. We speculate that a second role of partially reduced Fe-oxides (FeO) is to dissociate water and provide slightly acidic OH groups, which can catalyze the dehydration step of 4-hydroxypentanoicacid intermediate to γ-valerolactone in the mechanism of LvA hydrogenation (Scheme 1) [49,50]. This kind of promoting effect has been discussed for earlier dehydration relevant to aqueous phase reforming by bimetallic Ir-Re, Pt-Re, and Rh-Re catalysts [51–53]. A similar mechanism involving acid-catalyzed dehydration followed by Pt-catalyzed hydrogenation for selective glycerol hydrogenolysis was also proposed by Davis and co-workers [54]. The promotion of metal catalysts with partially oxidized oxophilic MOx species, such as ReOx-promoted Rh has also been suggested by a DFT study [55]. In those cases, the oxophilic nature of Re facilitated the activation of water, while we speculate that for the Fe-Re bimetallic catalysts partially reduced Fe-oxide species play this role. The close proximity of metallic Re, Fe-Re alloys and their oxides including titania can also improve the hydrogenation activity via fast heterolytic activation of H2 at their interface. Such metal-support interfaces are known to facilitate heterolytic H2 activation [56].Encouraged by these findings, we further optimized the reaction conditions by varying the reaction temperature. Fe-Re(1:2)/TiO2 was selected for this optimization study as it was the most active catalyst in the screening stage. Fig. 13 shows that this catalyst is active for LvA hydrogenation at a temperature as low as 130 °C. Increasing the reaction temperature resulted in a remarkable increase of the catalytic performance. Nearly full conversion was achieved after reaction at 180 °C for 4 h, yielding 95 % GVL. The results at 200 °C were similar and demonstrate that the catalyst is very active and selective for LvA hydrogenation to GVL in water. The sigmoidal activation with respect to temperature might be due to the higher water coverage on the reduced Re surface at too low temperature, which may also lead to partial re-oxidation. Further operando spectroscopy would be required to investigate the catalytic surface during aqueous phase LvA hydrogenation.We performed additional LvA hydrogenation reactions by varying the reaction time from 30 min to 6 h at a temperature of 180 °C. The performance of Fe/TiO2, Re/TiO2, and Fe-Re(1:2)/TiO2 are compared in Fig. 14 . It is clear that the bimetallic catalyst is much more active in LvA hydrogenation. About 50 % yield of GVL at 50 % LvA conversion was achieved after 2 h reaction for Fe-Re(1:2)/TiO2, whereas Re(13)/TiO2 and Fe(2.0)/TiO2 were nearly inactive. Notably, the two monometallic catalysts became slightly active after prolonged reaction but afforded only GVL yields of 40 % and 5%, for Re(13)/TiO2 and Fe(2.0)/TiO2 after 6 h. For Fe-Re(1:2)/TiO2, the maximum yield of GVL ∼95 % was already reached after 4 h.Hydrogenation of levulinic acid towards γ-valerolactone is one of the most promising reactions in the fields of biomass valorization to fine chemicals and liquid transportation fuels. A series of Fe-Re supported on TiO2 (P25) catalysts were tested for hydrogenation of levulinic acid in water. Remarkable improvements in catalytic performance were observed for the Fe-Re bimetallic catalysts, in comparison with their monometallic counterparts, suggesting a synergistic effect. H2-TPR results show that the reduction peak of Fe-Re samples shifts to higher temperature regime due to the close interaction between Fe and Re species. XANES shows that the presence of Re promotes the reduction of Fe and there is an interaction between Fe and Re. EXAFS analysis further reveals the presence of Fe-Re alloy. XPS and low-temperature CO-FTIR results evidenced large fractions of FeOx and ReOx are present and part of the metallic Re is covered by FeOx. The Mössbauer study shows that only part of Fe can be reduced to metallic Fe, due to the strong interaction between Fe and TiO2 support. The coverage of TiO2 surface hydroxy by Fe species was believed to be the reason for the improved reducibility of Re. The Fe-Re alloy, improved Re reducibility and FeReOx species are likely present in the bimetallic samples and believed to be the main reason for the enhanced catalytic activity. The presence of FeOx and ReOx are highly oxophilic and might introduce Re−OH acidic groups via hydration during the reaction, facilitating the dehydration, a key intermediate step for levulinic acid hydrogenation. Under optimized conditions, nearly full conversion of levulinic acid could be achieved after reaction at 180 °C for 4 h, obtaining 95 % yield of GVL. Xiaoming Huang: Methodology, Investigation, Data curation, Formal analysis, Validation, Writing - original draft, Visualization. Kaituo Liu: Methodology, Investigation, Data curation, Formal analysis, Validation, Visualization. Wilbert L. Vrijburg: Formal analysis. Xianhong Ouyang: Investigation. A. Iulian Dugulan: Investigation, Formal analysis. Yingxin Liu: Investigation. M.W.G.M. Tiny Verhoeven: Investigation, Formal analysis. Nikolay A. Kosinov: Methodology, Investigation, Formal analysis. Evgeny A. Pidko: Conceptualization, Supervision. Emiel J.M. Hensen: Conceptualization, Writing - review & editing, Supervision, Project administration.There are no conflicts to declare.This work was performed in the framework of the European Union FP7 NMP project NOVACAM (“Novel Cheap and Abundant Materials for Catalytic Biomass Conversion”, FP7‐NMP‐2013‐EU‐Japan‐604319). E.A.P. thanks the Government of the Russian Federation (Grant 074‐U01) and the Ministry of Education and Science of the Russian Federation (Project 11.1706.2017/4.6) for supporting his research in the framework of his personal ITMO professorship. The authors would like to thank Rim van de Poll and Alexander Parastaev for helping with the XAS measurements, Bart Zijlstra for the FEFF calculations and Miao Yu for the useful discussions about EXAFS fitting.

Hydrogenation of levulinic acid to γ-valerolactone is a key reaction in the valorization of carbohydrates to renewable fuels and chemicals. State-of-the-art catalysts are based on supported noble metal nanoparticle catalysts. We report the utility of a bimetallic Fe-Re supported on TiO2 for this reaction. A strong synergy was observed between Fe and Re for the hydrogenation of levulinic acid in water under mild conditions. Fe-Re/TiO2 shows superior catalytic performance compared to monometallic Fe and Re catalysts at similar metal content. The hydrogenation activity of the bimetallic catalysts increased with Re content. H2-TPR, XPS, XANES, EXAFS, Mössbauer spectroscopy, TEM, and low-temperature CO IR spectroscopy show that the bimetallic catalysts contain metallic Re nanoparticles covered by FeOx species and small amounts of a Fe-Re alloy. Under reaction conditions, the partially reduced surface FeOx species adsorb water and form Brønsted acidic OH groups, which are involved in dehydration of reaction intermediates. Under optimized conditions, nearly full conversion of levulinic acid with a 95 % yield of γ-valerolactone could be achieved at a temperature as low as 180 °C in water at a H2 pressure of 40 bar.

Adsorption energyBond lengthMayer bond orderReaction EnergyActivation EnergyWith the adjusted energy structure and change in the supply-demand relationship, the production capacity of fossil fuels in the petrochemical industry is decreasing gradually, while the production capacity of chemical raw materials must be urgently strengthened. Hydrocracking technology is an important process for the petrochemical industry to convert distillates into chemical raw materials (Bezergianni et al., 2009; Choudhary and Saraf, 1975; Köseoḡ;lu and Phillips, 1987; Scherzer Jg, 1996). In general, hydrocracking catalysts contain acidic zeolites as the cracking center (Ali et al., 2002; Martens et al., 2001; Speight, 2020; Zhang et al., 2007), and the nitrogen contents, particularly the basic nitrogen compounds in the cracking feedstock, are strictly limited. To remove the nitrogen compounds in the feedstock, a hydrocracking pretreatment catalyst is required in the hydrocracking process (Badoga et al., 2020; Kohli et al., 2019; Oh et al., 2019; Prada Silvy et al., 2019).The prevailing commercial pretreatment catalysts are highly active Mo–Ni bimetal γ-alumina-supported hydrotreating catalysts. Strong acidic supports and electronegative elements can significantly improve the removal of nitrogen compounds (Hu et al., 2019; Tung et al., 2017; Valles et al., 2019; Yao et al., 2017; Tang et al., 2017). These prompters could cause electron deficiency or bring extra protons to the Ni–Mo–S active nanoclusters via inductive effects or charge transfer (Prins et al., 1997; Tominaga and Nagai, 2010). With the rapid development of computer technology and the progress of quantum chemical calculations, theoretical calculations of complex catalytic processes, such as charge distribution effects on hydrodenitrogenation, can be implemented.In this study, quinoline, which is a typical basic two-ring nitrogen compound for hydrodenitrogenation (HDN) research (Li et al., 2012; Lu et al., 2007), is used as the probe, and a series of model Ni–Mo–S with different charge distributions are used as the active sites. The key processes of quinoline HDN, including adsorption, hydrogenation saturation, and C–N bond cleavage on the Ni–Mo–S, are calculated by quantum chemistry calculations.The neutral Ni–Mo–S model in this study was a hexagonal single-layer nanocluster. The stable state of the Ni–Mo–S active sites under the hydrogenation reaction is shown in Fig. 1 (Ding et al., 2018a). Previous studies have shown that on the Ni(Co)–Mo–S or MoS2, the hydrogenolysis active centers are mainly located on the (10-10) plane, denoted as the Ni(Co)–Mo edge (Ding et al., 2017a, 2017b, 2018a,b; Sun et al., 2004; Sylvain et al., 2004). In this study, the issues of quinoline HDN are focused on the Ni–Mo edge of the Ni–Mo–S active sites. Considering the symmetry of the calculation model, in the electron deficiency case, three pairs of electrons were subtracted from the Ni–Mo–S, and the model is denoted as E-Ni-Mo-S. For additional protons, one proton was added to each Ni–Mo edge, and the model was denoted as P–Ni–Mo–S.Calculations were performed using the DMol3 code. The calculation function is the general gradient approximation-Perdew-Burke-Ernzerhof function, and the basis set is a double numerical plus polarization basis (Chigo Anota and Cocoletzi, 2014; Delley and B. 1982). To analyze the transition state, the open shell mode was used to treat the electron spin. The symmetry in the calculation was also canceled to meet the anisotropy in the HDN process. The orbital cut off is unified to 5.0 Å for every atom. To balance the calculation speed and accuracy, the effective core potential (ECP) method was used to simplify the core electron treatment, and thermal smearing was set to 5 × 10−4 Hartree. The self-consistent field density convergence (SCF) was set to 2 × 10−5, and the energy tolerance for the geometry optimization and transition state was 2 × 10−5 Hartree. The force tolerance was 4 × 10−3 Ha/Å geometry optimization and 3 × 10−3 Ha/Å for the transition search. The Grimme 06 correction method was used to calculate the atomic dispersion. The exchange-correlation dependent factor s 6 was set to 1.0, and the damping coefficient was set to 20.0. The dispersion parameters for the atoms involved in this calculation can be found in Table 1 (Grimme, 2010, 2011).During the process of HDN, the adsorption of reactants on the active sites relies on the interactions between the lone or conjugated electron pairs of the reactants and the unoccupied molecular orbitals of the active sites. According to acid-base theory, E-Ni-Mo-S can protonate basic nitrogen compounds. The changes in molecular orbitals before and after the protonation of quinoline (Q), tetrahydroquinoline (THQ), and decahydroquinoline (DHQ) (THQ and DHQ are important intermediates in the hydrodenitrogenation process of quinoline (Luan et al., 2009) and are shown in Table 2 . The highest occupied molecular orbital (HOMO) of nonprotonated basic nitrogen compounds is mainly contributed by the lone pair electrons on the nitrogen atoms. When the nitrides are protonated by E-Ni-Mo-S, the lone pair electrons of atoms combine with H+. The newly generated HOMO has barely related to the nitrogen atoms, and the orbital eigenvalue is significantly reduced. This change will weaken the binding ability between the active center and the nitrogen compounds.The effects of charge distributions on the lowest unoccupied molecular orbital (LUMO) are shown in Table 3 . On the neutral Ni–Mo edge, the LUMO is attributed to the d orbital of the tetracoordinated Ni atom and the pentacoordinated Mo atom with S atoms. The LUMO eigenvalue is −4.53 eV. On the E-Ni-Mo-S, the composition and morphology of the LUMO orbitals do not change much, and they still consist of unoccupied d orbitals from aligned metal atoms. However, the LUMO eigenvalue significantly decreases to −8.20 eV. On the P–Ni–Mo edge, the H+ bonds are stably coordinated with the pentacoordinated Mo atom, which is near the exposed Ni atom. This combination will satisfy the stable hexacoordination of the Mo atom. The Ni atom close to H+ will be more electron deficient, leading to a reduction in the LUMO eigenvalue. It could be concluded that both the lack of electrons and the extra protons will lower the LUMO eigenvalue and enhance the ability of receiving electrons from the reactants.During the HDN process, the reactant, some important intermediates and the ammonia have strong adsorption ability on the active centers. The calculation results of the adsorption of Q, THQ, DHQ and NH3 on Ni–Mo-edge affected by different charge distributions are shown in Table 4 . On the neutral Ni–Mo–S, the formation of the nitrogen compounds adsorption is point to point. Specifically, the nitrogen atom of Q, THQ and DHQ bonded with nickel atom, forming an N–Ni bond with 2.2–2.3 Å and 0.3–0.4 Mayer bond order. The nitrogen atom of NH3 prefers to bond with Mo atoms. The bond direction is in accord with the orientation of the LUMO morphology listed in Table 2. Because of the similarity of LUMO morphology, the adsorption morphology of the nitrogen compounds on the E-Ni-Mo-S and Ni–Mo–S active sites are similar as well, whereas the significant difference is the adsorption energy. The adsorption energies of nitrogen compounds on E-Ni-Mo-S are approximately 20–30 larger than those on the neutral Ni–Mo–S. When the nitrogen compounds adsorb on the P–Ni–Mo–S active sites, the H+ will transfer to the nitrogen, combing with the long pair electrons. The adsorption of nitrogen compounds will turn to flat model without forming the N–Ni bond. Despite the lacking of the single strong chemisorption bonds, the weak interaction between the conjugate π-electrons and the unoccupied orbitals the extra dispersion force from the increasing contact area both enlarge the adsorption energy. According to the calculation results, both the electron deficiency and the extra proton will enhance the adsorption of nitrogen compounds on the Ni–Mo-edge, whereas the ammonia desorption is inhibited which is negative to the recovery of the active center during the HDN process.On the Ni–Mo edge, hydrogen activation is carried out by H2 molecule dissociation on the metal or sulfur atom. Hydrogen dissociation with adsorption of a quinoline molecule was calculated, and the results are shown in Table 5 . On the Ni–Mo edge of neutral Ni–Mo–S, hydrogen dissociation is a strong endothermic step with a high energy barrier. At the corresponding position of E-Ni-Mo-S, this dissociation is an obvious exothermic process, and the activation energy significantly decreases to 108.51 kJ/mol. On the P–Ni–Mo–S, the thermal effects and activation energy charge were less significant than those on E-Ni-Mo-S. It could be predicted that electron deficiency will promote hydrogen dissociation.The newly generated active hydrogen must transfer to the nitrogen compounds quickly in the case of self-combination. Among the several hydrogen transfers of quinoline HDN, the conversion from THQ to penta-hydroquinoline (PHQ) is a key speed control step (Ding et al., 2017; Jian and Prins, 1998). This elementary reaction on the Ni–Mo edge with different charge distributions is shown in Table 6 . The active hydrogen breaks the conjugated aromatic rings. The reaction energy is up to 40–70 kJ/mol, and the activation energy exceeds 100 kJ/mol. In comparison, hydrogen transfer on neutral Ni–Mo–S is relatively easier and most difficult on E-Ni-Mo-S. The difficulty of hydrogen transfer is adverse to hydrogen dissociation, indicating that the stronger the interaction between the hydrogen and active sites, the easier the hydrogen dissociation and the harder the hydrogen transfer.For quinoline, the main pathway of C–N bond cleavage is the E2 elimination of DHQ. This process contains two elementary steps: the first step is hydrogen elimination of β-C, forming nona-hydroquinoline, and the second step is cleavage of the C–N bond, forming a CC bond and amino group (Li et al., 2012). Table 7 shows the elimination of the β-H of DHQ on Ni–Mo-edges with different acid types. According to the calculated results, the transfer of β-H to the active sites is an endothermic process with high activation energy. During this step, the S accepts the hydrogen atom, and the β-C atom bonds with the Mo atom. The influence of the charge distribution is limited, whereas the H+ provided by B–Ni–Mo–S returns to the active sites, and the reaction energy and activation energy both decrease. The C–N bond cleavage of NHQ is shown in Table 8 . The results show that the C–N break on the neutral Ni–Mo–S is a strong endothermic step with very high energy barrier. Meanwhile, the C–N bond cleaves the newly generated CC bonds attached with the Mo atom. The electron deficiency on Ni–Mo–S does not change the pathway of C–N bond cleavage, and the influence is quite limited. Attributable to the stronger adsorption ability of the LUMO, the energy barrier decreased by approximately 10 kJ/mol on the E-Mo-Ni-S. Notably, on the P–Ni–Mo–S, the proton transferred to the Ni–Mo-edge in the elimination step returns back to nitrogen compounds during C–N bond cleavage. The proton not only lowers the electron density but also increases the coordination of the N atom, leading to a more stable transition state of C–N bond cleavage. The activation energy decreased by approximately 40 kJ/mol, indicating that flexible H+ transfer between the nitrogen compounds and the active center significantly lowered the C–N bond cleavage in the HDN of quinoline.In this study, the HDN catalytic activities of Ni–Mo–S with different charge distributions are calculated. The conclusions are as follows: 1. Electron deficiency and extra protons could both lower the LUMO eigenvalue of Ni–Mo–S. The effects of electron deficiency on the morphology are limited, whereas extra protons could change the local morphology of LUMO. 2. Electron deficiency and extra protons could both enhance the adsorption ability of Ni–Mo–S active sties to nitrogen compounds. On neutral Ni–Mo–S and E-Ni-Mo-S, the nitrogen compounds adsorb via the chemisorption N–Ni bond, whereas on P–Ni–Mo–S, the nitrogen compounds take flat adsorption. However, ammonia desorption is inhibited by electron deficiency and extra protons during the HDN process. 3. Electron deficiency on N–Mo–S promotes the generation of active hydrogen but restricts hydrogen transfer to nitrogen compounds. 4. During C–N bond cleavage, the proton of P–Ni–Mo–S can flexibly transfer between the nitrogen compounds and the active sites. In this way, the cleavage of C–N is significantly promoted. Electron deficiency and extra protons could both lower the LUMO eigenvalue of Ni–Mo–S. The effects of electron deficiency on the morphology are limited, whereas extra protons could change the local morphology of LUMO.Electron deficiency and extra protons could both enhance the adsorption ability of Ni–Mo–S active sties to nitrogen compounds. On neutral Ni–Mo–S and E-Ni-Mo-S, the nitrogen compounds adsorb via the chemisorption N–Ni bond, whereas on P–Ni–Mo–S, the nitrogen compounds take flat adsorption. However, ammonia desorption is inhibited by electron deficiency and extra protons during the HDN process.Electron deficiency on N–Mo–S promotes the generation of active hydrogen but restricts hydrogen transfer to nitrogen compounds.During C–N bond cleavage, the proton of P–Ni–Mo–S can flexibly transfer between the nitrogen compounds and the active sites. In this way, the cleavage of C–N is significantly promoted.The authors acknowledge the financial support from the Sinopec Science and Technology Department (Grant No. 121014-1).

The charge distribution on Ni–Mo–S active sites can affect hydrodenitrogenation (HDN) activity. In this study, a series of model Ni–Mo–S were developed with various charge distributions. For comparison, the charge distribution effects on quinoline HDN were studied. The results show that a lack of electrons and extra protons can both lower the orbital eigenvalue of the Ni–Mo–S, leading to stronger adsorption of nitrogen-containing compounds and inhibition of ammonia desorption. Electron deficiency will improve the generation of active hydrogen on the active sites but inhibit hydrogen transfer to the nitrogen compounds; extra protons can provide H+ to the nitrogen compounds, which will flexibly transfer between the nitrogen compound and active sites, thus improving the cleavage of the C–N bond.

Lignin, the most abundant renewable aromatic material on Earth, can be potentially exploited for the sustainable supply of fuels and chemicals which are currently derived from rapidly depleting and greenhouse gas emitting fossil resources [1–5]. Lignin, constituting 15–30 % of the biomass weight and up to 40 % of the biomass energy, is an amorphous and highly cross-linked macromolecule composed of the three primary phenylpropane monomers of p-coumaryl, coniferyl and sinapyl alcohols [6–10]. It can be processed via various depolymerization techniques for the production of high-value platform chemicals such as phenolics, aromatics and alkanes [11]. Today, industries such as pulp and paper manufacturing and lignocellulosics-to-ethanol processes produce large amounts of lignin as a by-product which is mostly burnt for use as an internal energy input [12–16]. Therefore, development of efficient processes for feasible utilization of lignin is highly important both in terms of environmental and economic aspects. In recent years, different thermochemical approaches (e.g., liquefaction and pyrolysis) have been applied for processing of lignin materials under reductive, neutral and oxidative atmospheres to produce value-added products (e.g., aromatic and cycloalkane hydrocarbons and phenolic compounds) [17,18]. However, there is a major problem in the processing of lignin, which is the high formation of char solid residues remaining from the conversion of lignin, causing a low yield of desired target products [19–24]. This happens since lignin fragments from degradation of lignin polymer are highly reactive and undergo rapid repolymerization to form large amounts of char [25]. This necessitates the development of the catalytic systems which can effectively suppress char-forming condensation reactions.In the reductive approaches, aiming to develop lignin-to-hydrocarbon processes, the most commonly tested catalysts can be divided into three groups: (i) transition (e.g., Ni, Cu)/noble (e.g., Pd, Pt, Ru) metal-based catalysts used in metallic form; (ii) metal oxide catalysts (e.g., MoOx, ReOx); (iii) conventional sulfide catalysts (e.g., NiMo/Al2O3, CoMo/Al2O3) currently being applied in refineries for hydrotreating purposes [19,26–36]. Meanwhile, different metal phosphide (e.g., MoP, Ni2P), nitride (e.g., Mo2N) and carbide (e.g., Mo2C) catalysts have also been widely tested for the hydrodeoxygenation (HDO) of lignin model compounds [3]. The problem with the first two groups is that, although they mostly have high hydrogenation efficiency, their application is limited to sulfur-free lignin materials since they can be readily poisoned and deactivated in the presence of sulfur. The conventional sulfide catalysts have also been reported to give a high char yield from the conversion of lignin. Agarwal et al. [31] reported a char yield of 23.3 wt% produced in the liquefaction of kraft pine lignin at 450 °C and 100 bar H2 using CoMo/Al2O3 as a conventional hydrotreatment sulfide catalyst. In another work performed by the same group [32] for hydrotreatment of kraft lignin at 350 °C and 100 bar H2, the solid residue yields of 20.5 and 35.4 wt% were obtained over conventional sulfide catalysts of NiMo/Al2O3 and CoMo/Al2O3, respectively. Considering that the majority of commercially available lignins have sulfur content (sulfite and kraft lignins with 3.5–8.0 % and 1.0–3.0 % sulfur, respectively), development of novel sulfur-resistant catalysts with high HDO efficiency can be an important strategy for the future supply of fuels and chemicals from lignin feedstocks [2,37]. Sulfur-resistant catalysts could also be applied for co-processing of lignin materials with other sulfur-containing feedstocks. This is particularly important for the feasibility of the integration of lignin processing with the existing petroleum refinery units with sulfur-containing input streams to improve the cost-effectiveness of lignin valorization.To meet the above-mentioned challenges, and as a step towards an applicable and efficient lignin-to-hydrocarbon process, this work aimed to develop a catalyst with three major properties: (i) sulfur resistance; (ii) high char-suppressing potential; (iii) high HDO efficiency. In this study, rhenium sulfide was tested as a catalyst for HDO of m-cresol (as a model lignin-derived phenolic compound) and reductive liquefaction of kraft lignin, and its performance was compared with that of nickel-molybdenum sulfide which is a well-established conventional sulfide catalyst. To the best of our knowledge, this is the first use of rhenium sulfide for the conversion of a lignin feedstock. Rhenium sulfide has been reported in literature to be an active catalyst for hydrodesulfurization and hydrodenitrogenation reactions [38], and metallic rhenium is known as a catalyst with high hydrogenation efficiency [39]. Recently, hydrodeoxygenation of some phenolic compounds have also been performed using different rhenium phases (metal, oxide and sulfide) [40–42]. Moreover, rhenium has a considerably lower price compared to noble metals like Pd, Pt, Ru, Rh and Ir, and it also may have a lower price in the future with the enhanced demand for rhenium compounds and its increased commercial exploitation. Therefore, rhenium sulfide was selected to study its HDO activity and catalytic performance in the conversion of lignin. γ-Alumina, zirconia and desilicated HY zeolite were used as support materials for the rhenium sulfide catalysts in this work. An alkali-assisted depolymerization was also carried out to achieve an enhanced lignin depolymerization, and to study the correlation between depolymerization rate and stabilization rate as a key factor for suppressing char formation in a lignin liquefaction process.NiMo/Al2O3, Re/Al2O3, Re/ZrO2 and Re/HY were examined as catalysts in this work. HY, used as catalyst support, was a mesoporous zeolite obtained by desilication of a commercial Y zeolite (Zeolyst, CBV 780, SiO2/Al2O3 molar ratio: 80) through alkaline treatment in a 0.3 M NaOH solution with mild stirring at 80 °C for 60 min. Then, the sample was filtered, washed with distilled water, and dried at 110 °C overnight. Subsequently, the desilicated zeolite was converted to the protonic form by three successive ion exchanges with a 1 M aqueous NH4Cl solution at 80 °C for 4 h, followed by drying at 110 °C overnight and calcination at 550 °C for 12 h with a heating ramp of 2 °C min−1. As a result of desilication, the SiO2/Al2O3 molar ratio of HY zeolite was decreased from 80 to 36. Supported rhenium catalysts were obtained by incipient wetness impregnation of γ-Al2O3 (Puralox SCCa 150/200, Sasol), ZrO2 (with monoclinic crystalline structure, SZ 31164, NORPRO) and HY with an aqueous solution of NH4ReO4 (Sigma-Aldrich). The NiMo/Al2O3 catalyst was prepared through incipient wetness co-impregnation of γ-Al2O3 with an aqueous solution containing both (NH4)6Mo7O24·4H2O and Ni(NO3)2·6H2O (Sigma-Aldrich). The amount of rhenium loaded on all the supports was approximately 3 wt% (2.8, 2.8 and 2.7 wt% on Al2O3, ZrO2 and HY supports, respectively), and the loading amounts of nickel and molybdenum metals on alumina support were 4.8 and 14.6 wt%, respectively (determined by quantitative XRF). At these metal loading amounts, both rhenium and nickel-molybdenum catalysts gave similar conversions for the HDO of m-cresol (based on initial experiments). Therefore, these metal loading amounts were selected to study the performance of the catalysts in the liquefaction of lignin. After impregnation, the catalysts were dried first at 60 °C (12 h) and then at 110 °C (12 h), with subsequent calcination at 550 °C for 12 h. Prior to reaction, the prepared catalyst was sulfided with dimethyl disulfide (DMDS, ≥ 99 %, Sigma-Aldrich) in the presence of 20 bar hydrogen (99.9 %, AGA) at 340 °C for 4 h in a Parr autoclave reactor.The crystalline structure of the catalysts was determined using X-ray diffraction (XRD) on a Bruker AXSD8 Advance X-ray powder diffractometer with Cu Kα radiation (λ = 1.542 Å). The chemical analysis of catalysts was carried out using an X-ray fluorescence (XRF) instrument (PANalytical Epsilon 3XL). The textural properties of the samples were determined by nitrogen isothermal (−196 °C) adsorption-desorption using a TriStar 3000 instrument. Transmission electron microscopy (TEM) images were acquired with a high angle annular dark field (HAADF) detector using a FEI Titan 80–300 operating at the accelerating voltage of 300 kV. The electronic states of the supported metals were determined by X-ray photoelectron spectroscopy (XPS) measurement using a PerkinElmer PHI 5000 VersaProbe III Scanning XPS Microprobe.The acidity of the catalyst samples was measured by temperature programmed desorption of ammonia (NH3-TPD) and ethylamine (ethylamine-TPD) using an experimental setup consisting of mass flow controllers (MFC, Bronkhorst) for gas mixing, a quartz tube containing the sample in a temperature-controlled furnace and a mass spectrometer (MS, Hiden HPR-20 QUI) for measuring the amount of ammonia or ethylene in the outlet stream. Before ammonia or ethylamine adsorption, the presulfided catalyst was pretreated in Argon at 100 °C for 30 min. Then, it was exposed to 1555 ppm of NH3 or 543 ppm of ethylamine at 100 °C for 2 h. Afterwards, the sample was flushed with argon for 30 min to eliminate physisorbed ammonia/ethylamine. The desorption measurement was performed by heating the sample to 800 °C with a ramp of 10 °C min−1 under an argon flow (20 ml min−1). Thermogravimetric analysis (TGA) of the samples showed that no thermal decomposition occurs up to 800 °C (shown in Fig. S1, Supplementary Information). The samples were heated from 35 to 800 °C with a heating ramp of 10 °C min−1 in a stream of nitrogen gas (30 ml min−1).The hydrodeoxygenation (HDO) of m-cresol (≥ 99 %, Sigma-Aldrich) and reductive liquefaction of kraft lignin (product number: 370959, Sigma-Aldrich) (with 2.1 wt% sulfur content measured by ICP-AES and ICP-SMS) were conducted in a 300 ml Parr autoclave reactor. The elemental and proximate compositions of the kraft lignin sample are presented in Table S1, Supplementary Information. In each experiment, 3 g reactant, 90 ml hexadecane solvent (≥ 99 %, Sigma-Aldrich) and a certain amount of presulfided catalyst were added to the reactor. The lowest catalyst-to-feed ratio applied in this work was 1:3 (at least 1 g solid catalyst) in order to minimize mass transfer limitations for a better comparison of the catalytic performance of the different catalysts. The loaded reactor was sealed, and the air inside it was evacuated by pressurizing/depressurizing the reactor three times with first nitrogen and then hydrogen gas. Afterward, the reactor was pressurized with 30 bar of hydrogen gas and then heated up to the reaction temperature (340 or 400 °C). The reactor pressure was 56–57 and 65–68 bar at the reaction temperatures of 340 and 400 °C, respectively. The reactions were carried out with a stirring rate of 1000 rpm for a duration of 3 h for HDO of m-cresol and 6 h for lignin conversion. In the experiments for HDO of m-cresol, 88 mg DMDS was added to maintain the sulfidation of catalysts during the reaction, and in some experiments for lignin conversion, NaOH (≥ 98 %, Sigma-Aldrich) was added for enhanced depolymerization of lignin via alkali-catalyzed degradation. The liquid composition in HDO of m-cresol was monitored by collecting samples at intervals of 30 min. When the reaction was complete, the reactor was immediately quenched to room temperature and the solid phase was separated from the liquid product by vacuum filtration. The solid residue remaining from lignin conversion reactions was washed with acetone to remove the organics and solvent absorbed on the solid particles. After acetone extraction, the solid fraction (catalyst, char residues and unconverted lignin) was dried at 110 °C and weighed. Subsequently, the solid fraction was washed with dimethyl sulfoxide (DMSO, ≥ 99.9 %, Sigma-Aldrich) to dissolve and remove unconverted lignin. Then, the solids were washed with acetone to remove DMSO and dried at 110 °C overnight. The difference in the weight of solids before and after DMSO extraction was assigned to the amount of unconverted lignin which was almost negligible (below 2 wt% on feed) in all the experiments. The liquid phase products were analyzed by a two-dimensional gas chromatography system (GC × GC, Agilent 7890−5977A). The products were separated by two columns with different polarity (a DB-5 ms column (30 m × 0.25 mm × 0.25 mm) for the first dimension and a BPX-50 column (2.5 m × 0.10 mm × 0.10 mm) for the second dimension), and detected by mass spectrometer (MSD) and flame ionization (FID) detectors for qualitative and quantitative analysis, respectively. The product yields were measured by an external standard calibration method, with calibration curves using several known concentrations (mass) that were related to the FID peak areas (with R2 > 0.99). This calibration was conducted for a number of individual compounds such as toluene, ethylbenzene, propylbenzene, methylcyclohexane, cyclohexane, guaiacol, m-cresol, phenol, propylphenol, naphthalene, methylnaphthalene, tetralin, dimethyltetralin, methylbiphenyl, benzyl phenyl ether, biphenol and phenanthrene. Experiments were repeated 2–3 times to ensure the reproducibility of the data. Fig. 1 presents m-cresol conversion levels and product selectivities with time over sulfided Re/Al2O3 and NiMo/Al2O3 catalysts at 340 °C. Both catalysts exhibited a similar trend for the conversion of m-cresol, giving a complete conversion after 2.5 h reaction. Considering that the number of rhenium atoms loaded on alumina support is almost ten times less than that of molybdenum atoms (the loading amounts of Re and Mo were approximately 3 and 15 wt%, respectively), it could be inferred that rhenium sulfide is more active than molybdenum sulfide as a hydrogenation promoter. After 3 h reaction, the mass yields of methylcyclohexane, methylcyclohexene, ethylcyclopentane and toluene were 63.9, 0.0, 6.2 and 10.9 wt% over NiMo/Al2O3, and 75.1, 0.6, 8.2 and 5.2 wt% over Re/Al2O3, respectively (shown in Fig. 2 ).As can be seen from the products obtained by the conversion of m-cresol, the HDO reaction proceeds through both ring hydrogenation (HYD) and direct deoxygenation (DDO) pathways over both NiMo/Al2O3 and Re/Al2O3 catalysts. In the DDO mechanism, m-cresol is adsorbed through its oxygen atom on the catalyst active site which is a sulfur vacancy, and the double bond on the phenolic ring close to the Caromatic-OH bond is hydrogenated to a single bond. This results in a temporary removal of the electron delocalization effect of the out-of-plane lone pair electron orbital of oxygen onto the phenolic ring π bond orbital and, in turn, a weaker CO bond which can be easily cleaved by dehydration over an adjacent acid site, giving toluene as the final aromatic hydrocarbon product [43–45]. In the ring hydrogenation pathway, the co-planar adsorption of m-cresol on the catalyst surface leads to ring saturation, producing methylcyclohexanol as an intermediate. This is then followed by dehydration to form methylcyclohexene which undergoes subsequent hydrogenation to be converted into methylcyclohexane as the saturated cyclic hydrocarbon product [43,46]. Ring hydrogenation was the dominant HDO pathway over both catalysts, giving high methylcyclohexane-to-toluene molar ratios of 5.5 and 13.6 over NiMo/Al2O3 and Re/Al2O3, respectively. The higher methylcyclohexane-to-toluene ratio over Re/Al2O3 indicates higher hydrogenation activity of this catalyst. In a study for hydrodeoxygenation of 2-ethylphenol over sulfided Mo-based catalysts, it was suggested that ring hydrogenation via co-planar adsorption requires two neighboring sulfur vacancies as active site, while DDO mechanism occurs on a single sulfur vacancy [47]. It is also inferred from the product selectivities over time that toluene, produced via DDO mechanism, does not undergo ring hydrogenation and remains unchanged by the end of the reaction. Another difference between the two examined catalysts is the rate of the hydrogenation of methylcyclohexene which is lower over Re/Al2O3. Methylcyclohexene was almost undetectable during the reaction using NiMo/Al2O3, indicating that this intermediate only exists for a short time before it is rapidly hydrogenated to methylcyclohexane. In contrast, methylcyclohexene was observed in relatively high quantities in the first 2 h of the reaction over Re/Al2O3. This might be ascribed to the slower adsorption of this intermediate on hydrogenation active sites on the surface of Re/Al2O3, more likely due to the higher number of phenolic rings hydrogenated over this catalyst. Ethylcyclopentane was also produced in low yield over both catalysts via acid-catalyzed ring contraction of methylcyclohexane [48].The yields of monocyclic products and char residues obtained by the conversion of kraft lignin over different catalysts are shown in Table 1 . Lignin was almost fully converted (> 98 %) in all the experiments, and other products (not shown in Table 1) are mainly heavy oligomers (non-detectable by GC), tetralins, indenes, naphthalenes, water and gas products. A comparison of the monocyclic product yields of Re/Al2O3 and NiMo/Al2O3 at the reaction temperature of 340 °C (entries 2 and 3, Table 1) reveals a remarkable superiority of rhenium sulfide catalyst; the total monocyclic product yields achieved over Re/Al2O3 and NiMo/Al2O3 were 21.5 and 4.6 wt%, respectively. This is mainly due to the different amounts of char remaining from the conversion of lignin over these two catalysts, with the yields of 40.6 wt% over NiMo/Al2O3 and 11.2 wt% over Re/Al2O3 (the images of char residues remaining from lignin conversion are shown in Fig. S2, Supplementary Information). This significant difference clearly illustrates the high catalytic efficiency of Re/Al2O3 for suppressing char formation which is a major problem in thermochemical processes for conversion of lignin. The typical high char yields from lignin conversion is a result of the low stability and high reactivity of lignin-derived intermediates which undergo condensation reactions to form heavy compounds as solid char residues [49]. Radical coupling, quinone methide and vinyl condensation are some significant condensation mechanisms which lead to high amounts of char remaining from lignin conversion [25]. Low char yield obtained over alumina-supported rhenium sulfide reveals that this catalyst is highly effective for stabilizing the lignin-derived reactive compounds and, in turn, suppressing condensation reactions. This could be due to higher activity of the rhenium sulfide catalyst for hydrogenating free radicals and preventing radical coupling in a reducing atmosphere. The char-suppressing effect of hydrogenation could also be observed by a comparison of the char yield of the alumina-supported hydrogenation promoters (rhenium and nickel-molybdenum sulfides) with that of the pure alumina support. As shown in Table 1, entries 1–3, the highest char yield (48.2 wt%) was obtained using the pure alumina support, indicating a higher condensation rate in the absence of hydrogenation active sites. Meanwhile, almost no monocyclic hydrocarbons were produced over Al2O3 due to the absence of hydrogenation activity.The high efficiency of the rhenium sulfide catalyst should be recognized in light of the fact that the low char yield of 11.2 wt% was obtained in this case using hexadecane as solvent, which is not a good solubilizer of lignin-derived components. Therefore, rhenium sulfide could be effectively used in the absence of the oxygen-containing polar solvents (e.g., alcohols) which are typically used for lignin liquefaction due to their higher solubility for lignin fragments. This makes rhenium sulfide a potential catalyst to be used for co-processing of lignin with hydrocarbon feedstocks in conventional petroleum refinery units. Moreover, considering the resistance of this metal sulfide catalyst to sulfur poisoning, it can be effectively used in hydrotreating of sulfur-containing lignin feedstocks. In addition, although rhenium is more expensive than molybdenum, but it has a lower price compared to the other noble metals with high hydrogenation activity (e.g., Pt, Ru, Ir and Rh). It could also be noticed that the low loading of rhenium on the catalyst support (like 3 wt% in this work compared to the typically high loading amounts of molybdenum (12–15 wt%) in commercial molybdenum-based catalysts) can increase the cost-effectiveness of rhenium-based catalysts. They could also be more economically attractive in the future with the increased commercial exploitation of rhenium due to enhanced demand for rhenium compounds.One characteristic of rhenium which makes it a highly active metal for catalyzing deoxygenation reactions is its high oxophilicity [39]. The oxophilic rhenium species are well known to be efficient for activation of oxy-compounds by strong adsorption of oxygen-containing functional groups to the surface of catalyst. Hence, this facilitated adsorption and strengthened interaction could be a reason for the effective performance of rhenium sulfide for lignin degradation (through cleavage of ether linkages) and HDO of lignin-derived phenolic compounds.The high catalytic activity of rhenium sulfide can also be correlated to the low binding energy shift between the rhenium sulfide phase and rhenium metal. As depicted by XPS analysis, presented in Fig. 3 and Table 2 , the binding energies of the Re 4f7/2 component of the 4f doublet for the ReOx/Al2O3 catalyst are 44.03 and 46.15 eV which are assigned to Re6+ (ReO3) and Re7+ (Re2O7), respectively [50,51]. After sulfidation, this catalyst displayed two Re 4f7/2 contributions with binding energies of 41.36 and 42.57 eV which are attributed to ReS2 species and some oxysulfide species (S-Re-O), respectively [52,53]. As the relative proportion values show, rhenium sulfide on alumina support exists mainly (87 %) as ReS2 with the binding energy close to that of rhenium metal (40.4–40.7 eV) [54,55]. The low binding energy difference between the rhenium sulfide phase and rhenium metal indicates that a high degree of the characteristic of metal is preserved during sulfidation, giving a metal-like nature to the metal-sulfur valence molecular orbitals. As metallic rhenium is believed to be highly effective for activation of H2 molecules [39], this metal-like character of rhenium sulfide species leads to a facilitated uptake of hydrogen and a high rate of the dissociation of molecular H2, giving an enhanced hydrogenation efficiency. The low char yields obtained from the liquefaction of lignin in the presence of rhenium-based catalyst could be associated to the high hydrogenation efficiency of rhenium sulfide; the highly reactive lignin derivatives can be stabilized via rapid hydrogenation, and thus, undesired coupling reactions and repolymerization to char residues can be effectively inhibited. The XPS analysis of the spent alumina-supported rhenium sulfide catalyst used for the HDO of m-cresol shows that the sulfide state of rhenium was maintained during the HDO reaction (presented in Fig. S3). The binding energy of ReS2 species on the spent catalyst was similar to that of the fresh catalyst, while the Re 4f7/2 contribution attributed to oxysulfide species disappeared, indicating the complete sulfidation of these species during the HDO reaction.Rhenium-induced acidity can also play an important role in catalytic performance of rhenium species by improving acid-catalyzed reactions. As revealed by NH3-TPD data presented in Fig. 4 and Table 2, the catalyst acidity was increased by the addition of rhenium species; the total acidity of Al2O3 and ReS2/Al2O3 are 0.364 and 0.451 mmol g−1, respectively. The amount of acidity induced by rhenium sulfide species should be higher than the difference in acid amounts of Al2O3 and ReS2/Al2O3, since the impregnated metal species cover a portion of the acid sites of the support, and the amount of ammonia desorption (in TPD analysis) from support in ReS2/Al2O3 is less than that from alumina support alone. It is also noticeable that, based on the acid strength distribution of these two catalysts, rhenium-induced acidity is mostly of medium strength; the densities of weak, medium and strong acid sites are 0.124, 0.145 and 0.095 mmol g−1 in Al2O3, and 0.129, 0.204 and 0.118 mmol g−1 in ReS2/Al2O3, respectively. This increased acidity can particularly improve the cleavage of ether linkages of lignin, causing an enhanced rate of depolymerization [56]. As a result, lignin fragments can be converted into monomeric compounds before they undergo repolymerization to form heavy solid residues. Moreover, the acidity provided by rhenium species can also improve the dehydration step of HDO reaction, leading to an enhanced deoxygenation and increased hydrocarbon yield [45,57]. According to ethylamine-TPD analysis, the rhenium-induced acidity is mainly Lewis acidity. In ethylamine-TPD, ethylamine is adsorbed on Brønsted acid sites, and the ethylammonium ions (formed via proton transfer) undergo the Hofmann elimination reaction to produce ethylene and ammonia at higher temperatures [58,59]. Therefore, the ethylene detected during ethylamine-TPD is quantified to measure Brønsted acidity. Based on the ethylene desorption profile (shown in Fig. S4), the density of Brønsted acid sites of ReS2/Al2O3 is 0.022 mmol g−1 which constitutes 5% of the total acidity (0.451 mmol g−1, measured by NH3-TPD) of this catalyst, indicating that the catalyst acidity is mainly Lewis type.The reaction pathway and product distribution in a lignin liquefaction process is a strong function of reaction temperature mainly due to the temperature dependence resulting from varying activation energies for the different series and parallel reactions taking place during lignin conversion. The significance of reaction temperature is more realized when it is considered that it greatly affects the rate of repolymerization reactions of lignin derivatives which lead to undesired formation of solid char residues. At low reaction temperatures (usually below 300 °C), low lignin depolymerization occurs due to low thermal cracking and inefficient catalytic degradation (e.g., hydrogenolysis), and instead, the repolymerization of highly reactive lignin-derived compounds yields a high char formation since these compounds cannot be catalytically stabilized at low temperatures [22]. Similarly, at high reaction temperatures (usually above 400 °C), significant char-forming reactions happen as a result of severe carbonization [21,22]. Therefore, the applied reaction temperature should be high enough to provide the activation energies required for both depolymerization of lignin and stabilization of reactive lignin derivatives (via e.g. hydrogenation and alkylation) on one hand, and not too high in order to cause carbonization reactions on the other hand. As mentioned before, Al2O3-supported rhenium sulfide could efficiently suppress char-forming reactions at 340 °C. In order to examine the stabilizing efficiency of this catalyst at an elevated temperature, the reaction temperature was increased to 400 °C while keeping other parameters constant. This caused a reduction in char yield from 11.2 to 8.5 wt% (entries 3 and 4, Table 1), indicating that Re/Al2O3 catalyst could more effectively stabilize lignin fragments at the higher temperature of 400 °C through an enhanced hydrogenation of free radicals. Importantly, this temperature increase caused a remarkably improved deoxygenation efficiency more likely due to the enhanced dehydration activity of the Re/Al2O3 catalyst at elevated temperature; the monocyclic hydrocarbon yield was increased from 7.3 to 16.8 wt%, and the monocyclic phenolic yield was decreased from 14.2 to 0.7 wt% by an increase of temperature from 340 to 400 °C (entries 3 and 4, Table 1). Consequently, the enhanced oxygen removal at 400 °C resulted in a lower monocyclic product yield of 17.5 wt% which was slightly higher (21.5 wt%) at 340 °C. Moreover, HDO reaction selectivity was also affected by the increase of temperature, and the higher temperature of 400 °C favored direct deoxygenation over ring hydrogenation; the monocyclic aromatic hydrocarbon selectivity was increased from 38.7 to 48.3 mol% by increasing temperature from 340 to 400 °C (shown in Fig. 5 ). This could be caused by the decreased availability of hydrogen on the surface of the catalyst at elevated temperature as a result of the reduced hydrogen adsorption due to its exothermic nature [60]. Lower hydrogen availability is favorable for the DDO mechanism which requires less hydrogen consumption compared to the HYD reaction route. This is consistent with several previous studies reporting that HYD and DDO are the dominant reaction pathways taking place at low and high temperatures, respectively [61,62].As can be seen from the yields and selectivities of monocyclic products obtained over Re/Al2O3, Re/ZrO2 and Re/HY, presented in Table 1 (entries 4–6), catalyst support has a significant effect on catalytic performance and reaction pathway. For a better comparison of the catalytic activities of the supports, HY zeolite was desilicated to generate a mesoporous zeolitic structure with lower diffusion limitations of lignin fragments. The textural properties of the supports are shown in Table S2. HDO activity was remarkably affected by the choice of catalyst support, and monocyclic hydrocarbon yield was reduced in the order: Re/Al2O3 > Re/ZrO2 > Re/HY. The monocyclic hydrocarbon yields were 16.8, 11.2 and 8.5 wt%, and monocyclic phenolic yields were 0.7, 10.6 and 14.9 wt% over Re/Al2O3, Re/ZrO2 and Re/HY catalysts, respectively, indicating that the use of alumina as catalyst support led to the highest deoxygenation efficiency. As a result of the enhanced oxygen removal over Re/Al2O3, this catalyst gave a lower mass yield of total monocyclic compounds compared to Re/ZrO2 and Re/HY catalysts. It can be seen from the data shown in Fig. 5 that HDO reaction selectivity was not largely influenced by catalyst support, and the monocyclic aromatic hydrocarbon selectivity was similar (44.7–48.3 mol%) over Al2O3-, ZrO2- and HY-supported rhenium catalysts. Similar to the HDO activity trend of the catalysts, the stabilizing efficiency was also decreased in the order: Re/Al2O3 > Re/ZrO2 > Re/HY, giving the char yields of 8.5, 10.7 and 11.8 wt%, respectively.To study the effect of the addition of a zeolitic catalyst to Re/Al2O3, a combination of Re/Al2O3 (1 g) and HY (1 g) was used as the catalytic system for the conversion of kraft lignin at 400 °C. The addition of zeolite had a negative effect and resulted in a reduction of monocyclic product yield from 17.5 to 13.8 wt% (entries 4 and 7, Table 1). This could be due to the condensation of lignin-derived fragments catalyzed by zeolite acid sites in the absence of a hydrogenation promoter, converting both hydrocarbons and oxygenates into catalytic coke deposited inside the zeolite channels [63].TEM images, shown in Fig. 6 , illustrate that rhenium species are well dispersed on all the supports as spherical particles. This is in agreement with previous studies reporting a high dispersion of rhenium particles on different catalyst supports [38,45,64]. The spherical structure of rhenium sulfide species on a γ-alumina support was also observed in a study by Quartararo et al. [65]. However, Escalona et al. [38] reported the presence of rhenium sulfide as both layered crystallites and spherical particles (with different ratios) on a γ-alumina support, suggesting that the ratio between the two structures depends on the sulfiding condition. The same group later showed that ReS2 species supported on different materials had layered structure which was transformed to spherical structure (with lower S/Re ratio) when exposed to the electron beam for 15 min, as a result of the desulfurization to metallic Re under the beam [52]. However, this structural change under electron beam was not observed in our work, and the rhenium sulfide species appeared as spherical particles from the beginning of the exposure to the beam. The histograms of size distribution, presented in Fig. 6, indicate that the mean particle diameter of rhenium sulfide species is 1.1, 1.4 and 0.9 nm on Al2O3, HY and ZrO2 supports, respectively. Besides, the two main lattice fringes in the TEM image of ReS2/Al2O3 have interplanar spacings of about 0.20 and 0.14 nm, which correspond to the (400) and (440) reflections of the γ-alumina phase, respectively (shown in Figs. S5 and S6a). The difference in the catalytic performance of the supports could be correlated with their different acidic properties measured by NH3-TPD analysis. The data presented in Table 2 and Fig. 4 illustrate that the acid site density of the catalysts decreased in the order: Re/Al2O3 (0.451 mmol g−1) > Re/ZrO2 (0.236 mmol g−1) > Re/ HY (0.214 mmol g−1). The higher acidity of Re/Al2O3 promotes the dehydration step of the HDO reaction to remove the oxygen atom of lignin-derived phenolics as water. Hence, it is supposed that the promoted dehydration, which is an acid-catalyzed reaction, leads to the high hydrocarbon yield achieved over the Re/Al2O3 catalyst. Moreover, as revealed by XPS data, shown in Fig. 3 and Table 2, catalyst support also affects the chemical state of the supported rhenium sulfide species, which this, in turn, can influence the reaction rate and pathway. The binding energy of ReS2 species loaded on different supports decreased in the order: ReS2/ZrO2 (41.94 eV) > ReS2/HY (41.62 eV) > ReS2/Al2O3 (41.36 eV). Therefore, rhenium sulfide loaded on alumina support has the minimum binding energy difference with that of metallic rhenium, and in turn, exhibits the most metal-like behavior. Besides, based on the relative proportion values of Re 4f7/2 components, the alumina support leads to the highest degree of sulfidation of rhenium oxide species (87 %) followed by zirconia (84 %) and HY zeolite (75 %) supports. Furthermore, the atomic ratios of S/Re on the different supports reduced in the order: ReS2/ZrO2 (3.1) > ReS2/HY (2.4) > ReS2/Al2O3 (2.2), indicating that rhenium loaded on alumina support has the minimum attached sulfur ligands, resulting in the most sulfur-deficient sulfide phase and, in turn, the highest number of sulfur vacancies which can contribute to hydrogenation/hydrogenolysis reactions.Base-catalyzed cleavage of lignin linkages, particularly aryl-alkyl ether bonds, is a well-studied approach for lignin depolymerization. A wide variety of low-cost and commercially available catalytic reagents such as NaOH, LiOH and KOH have been applied for alkaline degradation of lignin [3]. As an example for the mechanism of base-catalyzed degradation of lignin, the cleavage of β-O-4 linkages, as the most dominant aryl-alkyl ether bonds in lignin structure, occurs by the polarization promoted by a base catalyst [66,67]. Using NaOH as base catalyst, it is proposed that the sodium cation polarizes the β-O-4 ether linkage via formation of a cation adduct with lignin. As a result, the oxygen atom of the ether bond gains an increased negative partial charge, and in turn, less energy is required for heterolytic cleavage of this bond. This leads to the formation of a sodium phenolate along with a carbenium ion transition state which is subsequently neutralized by the hydroxide ion.In this study, NaOH was added to the reaction system for an increased depolymerization degree and a higher yield of monomeric products. However, the ratio of the amount of Re/Al2O3 to the amount of added NaOH was critical for the fate of the converted lignin as either the desired target monomeric products or undesired solid char residues. The effect of this ratio is shown in Table 1 (entries 8–10). The addition of 1 g NaOH resulted in a large amount of char formation from lignin conversion (44.3 wt%), giving a monocyclic product yield of 16.3 wt% and monocyclic hydrocarbon yield of 11.9 wt% which were less than those obtained in the absence of NaOH (17.5 and 16.8 wt%, respectively). This is due to the high amount of additional NaOH which resulted in a high rate of base-catalyzed degradation and, in turn, a large number of lignin fragments produced in a short period of time at the beginning of the reaction. The amount of Re/Al2O3 catalyst (1 g) in the reaction medium was insufficient in order to stabilize the rapidly derived intermediates, so that they underwent thermal condensation to repolymerize into char residues. The high rate of condensation at the beginning of the reaction was confirmed by carrying out another experiment with similar amounts of Re/Al2O3 (1 g) and NaOH (1 g) for a total reaction time of 20 min; the yield of char generated in the first 20 min was 40.9 wt% which is 92 % of the char produced in the 6 h reaction experiment. As expected, an increase of the Re/Al2O3-to-NaOH ratio, by using 2 g Re/Al2O3 and 1 g NaOH led to a lower char yield of 23.1 wt% and, in turn, an increased monocyclic product yield of 20.9 wt%. This was further improved by a higher Re/Al2O3-to-NaOH ratio, and the combination of 2 g Re/Al2O3 and 0.5 g NaOH exhibited an even better catalytic performance, giving a high monocyclic product yield of 24.6 wt% (monocyclic saturated and aromatic hydrocarbon yields of 13.5 and 11.1 wt%, respectively) and a low char yield of 11.3 wt%. To examine whether this improvement was caused by a well-optimized Re/Al2O3-to-NaOH ratio or the use of an increased amount of Re/Al2O3, another experiment was performed using 2 g Re/Al2O3 with no addition of NaOH. This resulted in a monocyclic product yield of 18.6 wt% which was slightly higher than that obtained over 1 g Re/Al2O3 (17.5 wt%) and considerably lower than that of a catalytic system including 2 g Re/Al2O3 and 0.5 g NaOH (24.6 wt%), indicating that the optimum Re/Al2O3-to-NaOH ratio was the main reason to achieve a high monomer production. However, the char formation increased from 5.4 to 11.3 wt% when adding 0.5 g NaOH. The significance of the stabilizing effect of Re/Al2O3 could also be realized by the high char yield of 47.0 wt% obtained in a NaOH-catalyzed depolymerization of lignin in the absence of Re/Al2O3 (entry 11, Table 1).A comparison of the yields of monocyclic products and char residues obtained using NaOH, Re/Al2O3 and the combination of NaOH and Re/Al2O3 illustrates that the highest lignin conversion efficiency could be achieved in the presence of both NaOH and Re/Al2O3 (at optimum amounts). NaOH results in a high depolymerization of lignin to the intermediates which undergo condensation in the absence of Re/Al2O3, leading to a low yield of monocyclic products (4.2 wt%) (entry 11, Table 1). A higher monocyclic product yield of 18.6 wt% was obtained over Re/Al2O3 since this catalyst could effectively stabilize lignin fragments and suppress char-forming reactions (entry 12, Table 1). However, the highest monocyclic product yield of 24.6 wt% was achieved using a combination of NaOH and Re/Al2O3 due to the high depolymerization rate of lignin via base-catalyzed degradation on one hand, and the efficient stabilization of lignin-depolymerized fragments (suppression of condensation reactions) over Re/Al2O3 on the other hand (entry 10, Table 1). It could also be inferred from the results obtained at different Re/Al2O3-to-NaOH ratios that the correlation between the rate of lignin degradation and the rate of stabilization of lignin-depolymerized fragments is a key factor for suppressing char formation in a lignin liquefaction process. The rates of lignin depolymerization and subsequent stabilization of lignin derivatives should be well balanced to ensure that lignin derivatives can be stabilized by the applied catalytic system before they undergo repolymerization to form a large amount of char.It is also noteworthy that no phenolics were detected in the liquid product using the combination of 2 g Re/Al2O3 and 0.5 g NaOH, indicating the complete HDO of phenolic compounds to hydrocarbons. Using this catalytic system, complete conversion of lignin was observed, and monocyclic saturated hydrocarbons, monocyclic aromatic hydrocarbons, tetralins, indenes and naphthalenes were the main GC-detectable organic products with the yields of 13.5, 11.1, 2.1, 1.7 and 1.2 wt%, respectively (shown in Table 3 ). Meanwhile, the monocyclic hydrocarbon distribution presented in Fig. 7 show that methylcyclohexane (16.5 wt%), (1-methylethyl)cyclohexane (7.9 wt%) and 1-ethyl-2-methylcyclohexane (7.1 wt%) were the most dominant monocyclic saturated hydrocarbons, and 1-ethyl-4-methylbenzene (6.9 wt%), ethylbenzene (5.7 wt%) and toluene (4.4 wt%) were the major monocyclic aromatic hydrocarbons. The monocyclic hydrocarbon products were composed of 23.8 wt% C7-C9 arenes, 21.4 wt% C10-C14 arenes, 46.5 wt% C6-C9 cycloalkanes and 8.3 wt% C10-C13 cycloalkanes. All GC-detectable liquid products are listed in Table S3. The molecular weight distribution of the liquid products is also shown in Fig. S7. Scheme 1 presents a proposed network of the major series and parallel reaction pathways taking place in the catalytic liquefaction and hydrodeoxygenation of kraft lignin. First, lignin polymer is fragmented into lower weight oligomers by cleavage of CC and COC linkages through thermal cracking and NaOH-catalyzed degradation. These smaller lignin fragments can diffuse into the channels of the solid hydrotreating catalyst (e.g., ReS2/Al2O3) and be converted into monocyclic phenolics via hydrogenolysis. Monocyclic phenolics can undergo either direct deoxygenation or ring hydrogenation to produce monocyclic aromatic or saturated hydrocarbons, respectively. The HDO reaction selectivity could be controlled by several parameters including solvent type (polar or non-polar; solvent polarity affects solvent-reactant interactions and, in turn, the orientation of reactant adsorption on the catalyst surface), catalyst properties (oxophilicity, catalytic performance of hydrogenation active sites, hydrogen dissociation efficiency and catalyst channel size), reaction temperature (affecting hydrogenation activity and hydrogen availability on the surface of catalyst) and reaction time. In the case where free radicals of oligomers and phenolic derivatives are not well stabilized by hydrogenation, irreversible radical coupling reactions lead to rapid condensation which forms solid char residues. Therefore, the high-yield production of monocyclic hydrocarbons from the conversion of kraft lignin, using the catalytic combination of NaOH and ReS2/Al2O3, occurs through four major steps: (i) depolymerization of lignin to oligomers via thermal cracking and alkali-catalyzed degradation; (ii) hydrogenolysis of oligomers to monocyclic phenolics over ReS2/Al2O3; (iii) stabilization of the free radicals intermediates by ReS2-catalyzed hydrogenation; (iv) hydrodeoxygenation of monocyclic phenolics to monocyclic hydrocarbons.ReS2/Al2O3 was a highly efficient catalyst for high-yield production of monocyclic hydrocarbons in the reductive liquefaction of a sulfur-containing lignin due to: (i) high stabilizing efficiency via hydrogenation of the free radicals of lignin-depolymerized fragments, resulting in significant suppression of char-forming condensation reactions; (ii) high hydrodeoxygenation activity leading to an efficient oxygen removal from lignin-derived phenolic compounds; (iii) resistance to sulfur poisoning. Compared to NiMo/Al2O3 as a conventional sulfide catalyst, ReS2/Al2O3 led to a significantly lower char yield and higher monocyclic product yield; in the reductive liquefaction of kraft lignin at 340 °C, the alumina-supported nickel-molybdenum and rhenium sulfide catalysts resulted in char yields of 40.6 and 11.2 wt%, total monocyclic product yields of 4.6 and 21.5 wt%, and monocyclic hydrocarbon yields of 4.6 and 7.3 wt%, respectively. Char suppression over ReS2/Al2O3 was also observed at an elevated temperature of 400 °C, giving a low char yield of 8.5 wt%. Moreover, it was shown that HDO activity is greatly affected by the choice of catalyst support; Al2O3-, ZrO2- and HY-supported ReS2 catalysts resulted in considerably different monocyclic hydrocarbon yields of 16.8, 11.2 and 8.5 wt%, respectively. This could be ascribed to different rates of dehydration and hydrogenation reactions which are affected by the acid property of the support and the sulfidation degree of the supported phase, respectively. Oxophilicity, sufficient acidity, metal-like behavior of rhenium sulfide and high dispersion of the supported phase are the significant characteristics of ReS2/Al2O3, making this catalyst highly effective for reductive conversion of lignin into monocyclic hydrocarbons. The alkali-assisted depolymerization of lignin by an addition of NaOH clearly illustrated the significance of ReS2/Al2O3-to-NaOH ratio which needs to be well optimized for a positive effect of NaOH addition. Otherwise, the addition of a depolymerization promoter (e.g., NaOH) higher than its optimum amount leads to an insufficient stabilization of the lignin-depolymerized fragments, and in turn, shifts the reaction pathway towards a rapid condensation and high char formation. The ReS2/Al2O3-to-NaOH ratio of 4 g/g led to a high monocyclic hydrocarbon yield of 24.6 wt% and a low char yield of 11.3 wt%. Pouya Sirous-Rezaei: Conceptualization, Methodology, Investigation, Validation, Writing - original draft. Derek Creaser: Writing - review & editing, Supervision. Louise Olsson: Writing - review & editing, Supervision, Resources, Funding acquisition.There are no conflicts of interest to declare.The Swedish Energy Agency (P47511-1), Formas (2017-01392) and Area of Advance Energy at Chalmers are greatly acknowledged for their financial support. We would also like to acknowledge the Chalmers Material Characterization (CMAL) facilities for STEM and XPS measurements.Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.apcatb.2021.120449.The following is Supplementary data to this article:

Thermochemical processing of lignin ends up with a major problem which is the high yield of char remained from lignin conversion, causing low yields of desired products. The ReS2/Al2O3 catalyst, used in this work, exhibited a high char-suppressing potential and high hydrodeoxygenation efficiency in the reductive liquefaction of kraft lignin. Compared to NiMo/Al2O3, as a conventional sulfide catalyst, ReS2/Al2O3 showed significantly better catalytic performance with 72.4 % lower char yield, due to its high efficiency in stabilizing the lignin-depolymerized fragments. The remarkable catalytic performance of ReS2/Al2O3 is attributed to its high oxophilicity, the metal-like behavior of rhenium sulfide and sufficient acidity. The effects of reaction temperature and different catalyst supports (Al2O3, ZrO2 and desilicated HY zeolite) were also studied. In an alkali (NaOH)-assisted depolymerization of lignin, it was revealed that ReS2/Al2O3-to-NaOH (stabilization-to-depolymerization) ratio plays a crucial role in determining the reaction pathway toward either solid char residues or liquid monomeric products.

The increasing environmental pollution and the depletion of nonrenewable energy have raised awareness towards the development of hydrogen storage systems as a viable source of clean energy. However, the suitability for on-board application has been a challenge to realizing the much-anticipated hydrogen economy [1]. MgH2 is one of the most promising candidates for hydrogen storage and has been extensively investigated over the last few decades. This is because of its large gravimetric and volumetric hydrogen capacities of about 7.6 wt% and 110 g/L, respectively. MgH2 also has the benefits of natural abundance, low cost, and non-toxicity [2,3]. However, the sluggish sorption rates and high operating temperature of MgH2 ranging between 300 and 400 °C which result from high kinetic barriers and stable thermodynamics (dehydrogenation enthalpy ΔHf ≥ 75 kJ/mol), prevent its mobile application [4,5]. To overcome these barriers, tremendous efforts have been given to nano-structuring [6–8], alloying [9–14], and catalysis [15–19] in the past few years with significant progress made on the sorption properties. However, hydrogen capacity loss by the formation of a multicomponent alloy and the introduction of confined porous materials that result from nano-structuring and alloying [6,12] has drawn attention towards catalytic doping as one of the most effective ways of improving the hydrogen storage properties of MgH2. The use of catalysts such as transition metals [20,21], transition metal oxides [19,22,23], carbides [18,24], hydrides [25,26], etc., have been shown to improve the kinetics of MgH2.Among the various catalysts, TiO2 has recently attracted great interest due to its particle size and tunable Ti valence state (via oxygen vacancy creation) which can enhance the hydrogen storage properties of MgH2 [27,28]. For instance, MgH2 doped with 5 mol% of rutile, anatase, and P25 TiO2 synthesized by high-energy ball milling were investigated to enhance the hydrogenation properties of Mg [27]. From the results, the rutile TiO2 doped composite showed the fastest absorption kinetics and highest capacity; this was attributed to the formation of an ultrafine nanocomposite MgH2-TiO2. Pandey et al. [29] also observed that MgH2 catalyzed by 7 and 50 nm TiO2 exhibited the optimum catalytic effect for hydrogen desorption and absorption respectively among the particle sizes of 7, 25, 50, 100, and 250 nm. It was stressed that the TiO2 could partially be reduced at temperatures below 340 °C to form defective TiO2-x with oxygen vacancies during the sorption process of MgH2. Furthermore, nanocrystalline TiO2 supported on carbon (TiO2@C) has also shown good catalytic performance in the hydrogen storage reaction of MgH2 [30]. The addition of 10 wt% of the catalyst into MgH2 reduced the onset dehydrogenation temperature to 205 °C with the release of 6.5 wt% H2 within 7 min at 300 °C and reabsorption of 6.6 wt% H2 within 10 min at 140 °C. The mechanism behind the improvement was based on the weakening/breaking of the Mg-H bond by TiO2 as obtained from DFT calculations which agreed with the experimental results. Some recent investigations have revealed that hydrothermally synthesized TiO2-based catalysts such as the 2D graphene-like TiO2 nanosheet, and graphene-supported TiO2 nanoparticles (TiO2@rGO), could improve the sorption properties of MgH2 [31,32]. The perovskite oxides of titanium also demonstrated some level of efficiency towards improving the sorption properties of MgH2. For example, Zhang et.al reported the catalytic performance of Na2Ti3O7 nanotubes, with a diameter of 10 nm, which could facilitate the hydrogen de/absorption kinetics of MgH2 by providing a lot of hydrogen diffusion channels [33]. With 5 wt% of the catalyst, MgH2 could desorb 6.5 wt% H2 at 300 °C in 6 min and absorb 4.1 wt% H2 at 150 °C in 10 min. Some other examples include MgH2 catalyzed with Li2TiO3 [17], SrTiO3 [34], BaTiO3 [35], and NiTiO3 [36]. On a general note, it was resolved that fairly stable metal oxides with a large number of possible structural defects and high valence states possess high catalytic efficiency on the sorption reaction of MgH2 [22,37,38].Given the above investigations, it is reasonable to examine the catalytic effect of light metal hydride (NaH) pre-activated TiO2 in the improvement of MgH2 hydrogen storage performance. The choice of NaH as a reducing agent for TiO2 is based on its strong reducing ability which can help boost vacancy concentration, and defect sites in TiO2 under a mild preparatory condition; this was confirmed in our previous works [39,40] by using Na/NaCl induced-oxygen vacancy in TiO2 for hydrogenation and water-gas shift reactions. High-energy milling of TiO2 with NaH under room temperature yields black TiO2 powder (TiO2-x) which is reportedly characterized by surface disorders, surface defects, and oxygen vacancies [41,42]. Introducing 5 wt% of this powder reduces the operating temperature of MgH2 to ∼185 °C, with room temperature absorption of 4.5 wt% H2 in 45 min. 2.5 wt% of the catalyst also enables MgH2 to undergo 100 cycles of de/absorption at 300 °C. To the best of our knowledge, no investigation has been conducted before now on this branch of knowledge.Pure MgH2 (98%), pure NaH (95%), and anatase TiO2 (99.8% metal basis 25 nm particle size) powders were commercially purchased from Alfa-Aesar, Sigma Aldrich, and Macklin chemical respectively. The powders were used without further treatment. All samples were prepared under an argon atmosphere in the glovebox with a circulative purification system (O2 < 10 ppm, H2O < 0.1 ppm) to avoid the influence of oxygen and moisture. NaH doped TiO2 catalyst was prepared by ball milling NaH with TiO2 separately in mole ratios of 0, 0.5, 1, and 2 to 1 for 3 hr. After that, a preselected amount of each catalyst obtained was added into fresh MgH2 powder and ball-milled. Pristine MgH2 was also ball-milled separately with and without TiO2. Ball milling of all the MgH2 composites lasted for 16 hr with batch weight kept at 2 g. All of the milling processes were performed using Retsch PM 400 at room temperature with a rotation speed of 200 RPM. The mixtures were sealed in 150 ml stainless steel vials in a glovebox with a ball-to-powder weight ratio (BPR) of about 80:1. The milling process was interrupted for 2.5 min after every 10 min of rotation to dissipate accumulated heat.Powder X-ray diffraction (XRD) measurements were conducted using an X'Pert3 Materials Research Diffractometer (Malvern Panalytical) with Cu Kα radiation (λ = 0.154 nm) at 40 kV and 40 mA. Samples were measured into the steel sample holders and covered with Kapton to avoid contamination during the measurement. Each measurement was done at a scan speed of 2°/min over diffraction angles of 10° to 90°. The microstructure and morphology of the samples were investigated using Hitachi S-4800 scanning electron microscopy (SEM) equipped with an energy-dispersive X-ray spectroscopy (EDS) analysis unit and an FEI Tecnai G2 F20 S-TWIN transmission electron microscopy (TEM). For TEM analysis, the samples were dispersed in hexane, sonicated, and drop cast on a copper grid. Image processing was performed using Digital Micrograph (Gatan) software. X-ray photoelectron spectroscopy (XPS; Thermo Fisher Scientific, ESCALAB 250Xi, Al-Kα = 1486.6 eV) technique was used to analyze the surface state of the catalyst. The binding energy was calibrated using C-C binding energy at 284.4 eV as a reference.The thermal decomposition properties of the samples were first investigated by using a homemade temperature-programmed desorption system equipped with a mass spectrometer (HPR20, Hiden) (TPD-MS). About 15–20 mg of the samples were loaded into an air-tight sample holder and sealed to the reactor in the glovebox. The analysis was carried out between room temperature and 400 °C at a heating rate of 2 °C/min under 20 mL/min argon flow.The volumetric desorption of the samples was carried out using Gas Reaction Controller (Advanced Material Corporation, USA). 120–150 mg of each sample was tested. Samples were heated up from room temperature to 400 °C with a heating rate of 2 °C/min under 0.001 bar of H2. However, room temperature absorption was conducted at 10, 30, and 50 bars of H2 backpressure.Thermodynamic and kinetic de/re-hydrogenation behaviors of samples were evaluated by using a conventional Hy-Energy PCT pro-2000 pressure-composition-isotherm (PCI) analyzer. 200–250 mg of each sample was loaded into a standard autoclave steel reactor. Isothermal desorption was investigated at 260, and 290 °C under 0.001 bar of H2pressure, while isothermal absorption was investigated at 50, 100, 200, 230, and 260 °C under 30 bars of H2 pressure. The composite's reversibility was evaluated at 300 °C under 0.001 and 30–50 bars of H2 pressure for de/absorption. PCI desorption measurement was performed at 300, 320, and 340 °C. The thermodynamic property was determined by using the Van't Hoff equation [43], which is expressed as a function of the equilibrium pressures recorded during PCI measurements. (1) In ( P H 2 / P θ ) = − Δ H / RT + Δ S / R Where PH2, ΔH, and ΔS are the hydrogen equilibrium pressure, enthalpy, and entropy change, respectively.The apparent activation energy of each sample under investigation was determined using the Kissinger method [44] through the mass spectra data obtained from TPD-MS. Samples were heated from room temperature to 400 °C with heating rates of 2, 6, 8, and 10 °C/min under 20 mL/min of argon flow. The method is as described in equation 2. (2) In ( β / T p 2 ) = − E a / R T p + A Where β is the heating rate, Tp 2 is the peak temperature of desorption given by the result of TPD-MS, R is the gas constant, A is a linear constant and Ea is the activation energy calculated from the slope value of the Kissinger plot.The optimum ratio of NaH to TiO2 with the best catalytic performance was determined by TPD-MS (Fig. S1), which indicates that NaH doped TiO2 in a 1:1 mole ratio (designated as NaTiOxH) has the best catalytic effects. Following that, the dehydrogenation properties of as-milled MgH2, MgH2–5 wt% TiO2, and MgH2-ywt% NaTiOxH (y = 2.5, 5, and 10) composites were measured by TPD-MS and the results are summarized in Fig. 1 a. The results show that 10 wt% NaTiOxH catalyzed MgH2 starts to desorb hydrogen from ∼174 °C and reaches its peak at 237 °C; which is ∼100 °C lower than the as-milled MgH2. However, reducing the amount of NaTiOxH from 10 to 5 and 2.5 wt% only influences the dehydrogenation peak slightly.To clearly show the catalytic effects of the different catalysts and doping amounts, temperature-programmed volumetric desorption of the prepared samples was measured and plotted in Fig. 1(b). It shows that 2.5, 5, and 10 wt% NaTiOxH catalyzed MgH2 begin to release H2 from temperatures below 200 °C, while they desorb ∼6.9, 7.2, and 6.2 wt% H2 at ∼260 °C; finally, the composites release off ∼7.5, 7.4, and 6.5 wt% H2 at ∼320 °C. However, as-milled MgH2 starts to dehydrogenate at ∼ 295 °C which is ∼100 °C higher than these catalyzed samples, and it liberates ∼7.5 wt% H2 at ∼375 °C. Taking the kinetics and H2 capacity into consideration, the 5 wt% NaTiOxH catalyzed MgH2 was selected for further investigations.The hydrogenation properties of MgH2–5 wt% NaTiOxH were measured under varying conditions as shown in Fig. 2 . At room temperature (Fig. 2a), the dehydrogenated MgH2–5wt% NaTiOxH absorbs ∼4.5 and more than 5.0 wt% H2 under 50 bars of H2 pressure within the first 45 and 120 min, respectively. Meanwhile, under 10 and 30 bars of H2 pressure, the composite respectively absorbs ∼4.1 and 5.5 wt% H2 within the first 180 min; ∼5.5 and 6.0 wt% H2 in 6 hr; ∼6.0 and 6.4 wt% H2 (∼87.7% of the total) in 12 hr (shown in Fig. S2). The observed fluctuation of absorption curves between 30 and 50 bars at the apex could be ascribed to an uncontrollable variation in environmental temperature. Furthermore, a moderate temperature absorption measurement of the composite at 30 bars of H2 pressure (Fig. 2b) shows an absorption capacity of about ∼4.4 and 4.5 wt% H2 within the first 30 min at 50 °C and 100 °C, respectively; while as-milled MgH2 could barely absorb at these temperatures. This enhanced absorption kinetics further confirms the catalytic efficiency of NaTiOxH catalyst on MgH2.Subsequently, a comparative measurement of isothermal de/re-hydrogenation at high temperatures of both the non-catalyzed and catalyzed MgH2 was conducted at four (4) different constant temperatures of 200, 230, 260, and 290 °C under 0.001 bar and 30 bars of H2, respectively (Fig. 3 ). Figure 3(a) shows that the composite releases ∼7.2 wt% H2 within the first 15 min at 290 °C, and ∼6.9 wt% in 60 min at 260 °C. However, as-milled MgH2 releases only ∼3.1 wt% H2 after 120 min at 290 °C, with no significant release at 260 °C. Furthermore, as shown in Fig. 3(b), the composite absorbs ∼6.6 wt% H2 within the first 120 s at 260 °C and ∼6.9 wt% after 20 min at 230 °C, while as-milled MgH2 only absorbs ∼6.0 and 6.1 wt% H2 within the same time range at 230 °C and 260 °C, respectively. Lastly, the composite charges ∼6.0 wt% H2 after 30 min at 200 °C while the as-milled sample absorbs less.A reversibility test was carried out to investigate the stability of the de/re-hydrogenation kinetics. In pursuit of a high hydrogen capacity with reasonable de/re-hydrogenation kinetics, MgH2–2.5 wt% NaTiOxH was chosen for cycling measurement across 100 cycles at 300 °C with a total of 475 hr (Fig. 4 ). Dehydrogenations were measured under 0.001 bar of H2 while hydrogenations were measured under 30, 40, and 50 bars of H2, respectively. The data profiles of the 1st, 50th, and 100th cycles of de/absorption (Fig. S3 and S4) show that the composite remains fairly intact with only a slight variation in the hydrogen capacity. Aside from the drop in kinetics, after 100 cycles, the hydrogen desorption capacity remained at ∼6.1 wt%, equivalent to ∼84% capacity retention and 0.012 wt% hydrogen loss per cycle. The kinetic relaxation observed could be attributed to the agglomeration of Mg/MgH2 particles and their separation from the catalyst during cycling. The zero-point slight variation noticeable between the 50th and 100th cycle absorptions could be attributed to the H2 backpressure increase from 40 to 50 bars.The dehydrogenation kinetic improvement of MgH2–5 wt% NaTiOxH was investigated by applying the Kissinger model to calculate the apparent activation energy (Ea). First, the mass spectra data of as-milled and catalyzed MgH2 were collected at heating rates of 2, 6, 8, and 10 °C/min as shown in Fig. 5 (a and b). The as-milled MgH2 shows clearly two-step desorption behavior similar to reported studies [36,45,46]. This was attributed to either the formation of metastable high-pressure γ-MgH2 or the non-uniformity of its grain/particle sizes. As shown in Fig. 5(c), the Kissinger plots of as-milled and 5 wt% NaTiOxH catalyzed MgH2 indicate that ball milling pristine MgH2 could reduce its Ea from the reported value of ∼180 kJ/mol [34,45,47,48] to ∼101 (±4) kJ/mol, and further reduction to ∼57 (±1) kJ/mol by adding NaTiOxH catalyst. The PCI desorption measurement of the composite at 300, 320, and 340 °C exhibits a distinct plateau region at each isotherm as shown in Fig. 5(d). The Van't Hoff plot of equilibrium pressure against temperature (inset) provides the dehydrogenation enthalpy change of ∼77 (±1.5) kJ/mol-H2. This indicates that NaTiOxH does not have any thermodynamic improving effect on MgH2.A comparative tabulation of the room temperature absorption capacity, activation energy, and reversibility of MgH2 catalyzed by a few representative TiO2-based catalysts is shown in Table 1 . This confirms the improved H2 storage performance of MgH2 by adding NaTiOxH catalyst.X-ray diffraction (XRD), transmission electron microscopy (TEM), selected area electron diffraction (SAED), and scanning electron microscopy (SEM) equipped with energy-dispersive X-ray spectroscopy (EDS) analysis unit were used to evaluate the phase composition, morphology, and crystallography of the synthesized NaTiOxH catalyst. X-ray photoelectron spectroscopic (XPS) investigation was also conducted to examine the chemical status of the component species consisted in the catalyst as shown in Fig. 6 . The XRD profile of NaTiOxH (Fig. 6a) displays the diffraction peaks corresponding to TiO2. The reduced intensity and broadness of some of the peaks have previously been attributed to an amorphous layer of defective TiO2-x nanoparticles formed near the surface [52]. In addition, a careful resolution of the peaks (Table S1) shows the emergence of a few crystalline phases of Ti3O5 and Na2Ti3O7 near the TiO2. The presence of these crystalline phases and the formation of black TiO2-x powder suggest that a reaction occurred between NaH and TiO2 during ball milling [42,53,54]. The SAED pattern of the catalyst (Fig. 6b) exhibits a typical point-ring diffraction characteristic of polycrystalline materials. The agreement in the calculated d-spacing values (Table S1) of indexed Ti3O5, and Na2Ti3O7 between the SAED and XRD patterns of the catalyst (shown in Fig. 6a and b), confirms their in-situ formation. In addition, the analysis of selected areas on the HRTEM image of NaTiOxH (Fig. 6c) shows the lattice fringes corresponding to Ti3O5 (110), and Na2Ti3O7 (101 and 103). The Fast Fourier Transformed (FFT) image (inset) reveals a rod-like diffraction pattern of the crystalline Na2Ti3O7 on the (101) plane. More TEM images of the catalyst that show the distribution and lattice fringes of Na/TiOx species on different sides could be seen in Fig. S5 and S6. For the morphology, the SEM micrograph of pristine TiO2 anatase (Fig. 6d) shows spherical nanoparticles (∼23 nm) with reduced size and increased surface area that is noticeable after milling with NaH (Fig. 6e). Bright patches on the particles’ surface could be attributed to contaminations on exposure to air before/during measurement. Furthermore, sodium 1s XPS spectrum of the catalyst (Fig. 6f) reveals its existence as Na-O-Ti, and Na-O/OH positioned at 1070.5 and 1072.1 eV, respectively [55]. The energy peak at 1069.2 eV is attributed to the Ti-Auger effect (Ti LMM) [56,57]. Titanium 2p XPS of the catalyst (Fig. 6g) shows that it exists in four (4) different valence states of +4, +3, +2, and 0; as distinct from the spectra of pure TiO2 (Fig. S7). This is due to the formation of defects and oxygen vacancies after milling with NaH; a phenomenon that could generate trapped-in electron densities around vacant 3d orbitals of the corresponding adjacent Ti atoms [49,52,58,59]. The appearance of the peak corresponding to Ti3+ at 457.5 eV could be attributed to the Ti3O5 and/or its related species with other possible oxyhydrides [60,61]. Ti2+ peak at 455.1 eV could be due to TiO species probably covered up in the amorphous layer of the defective TiO2-x [28,61,62], while Ti0 peak at ∼452.5 eV is considered to be the characteristic peak of Ti metal [28,60,63]. The 1 s XPS spectrum of oxygen (Fig. 6h) identifies its existence as O-Ti/Na (530.0 eV), H-O-Ti/Na, and/or peroxides (531.9 eV) with the binding energy slightly shifting to the higher region as compared to oxygen 1s spectra of pure TiO2 sample (Fig. S8). An adsorbed contaminant on exposure of the sample to air before measurement accounts for the peak at 532.4 eV, while the peak at 536.8 eV is attributed to the Na-Auger effect (Na KLL) [64–66]. Summarily, ball milling NaH with TiO2 liberates defective TiO2-x species which is characterized by reduced valences of titanium, and oxygen vacancies. A few crystalline phases of Ti3O5 and Na2Ti3O7 also emerge in situ during the reaction.As revealed in Sections 3.1 and 3.2, NaH doped TiO2 in a 1:1 mole ratio (NaTiOxH) exhibits excellent catalytic performance on the hydrogen storage properties of MgH2. To unravel the mechanism behind this performance, XRD patterns of the as-milled and de/re-hydrogenated samples before and after cycling were collected as shown in Fig. 7 (a). The patterns show characteristic peaks of MgH2, Mg, and MgO. The MgO peak could arise due to oxygen contamination before/during the XRD measurement [67] or the presence of MgO characteristic of the oxide additives loaded MgH2 [28]. Aside from these regular phases of MgH2/Mg/MgO, the phase stability of Ti3O5, Na2Ti3O7, and defective TiO2-x after 100 cycles show a probable reason behind the improved de/absorption behaviors of MgH2. Meanwhile, the SAED image of the as-milled composite (Fig. 7b) shows the plane corresponding to Ti3O5 and Na2Ti3O7 species. The TEM image of the as-milled composite reveals the contacting catalyst nanoparticles around the MgH2 particles (Fig. 7c), while the HRTEM images of the as-milled and hydrogenated composites reveal the corresponding lattice fringes of MgH2, MgO, Na2Ti3O7, and TiO2-x species (Fig. 7d and e). The calculated d-spacing values of the resolved lattice fringes as detected by SAED and TEM agree with both the XRD pattern and the standard values of d-spacing as shown in Table S2. More TEM images of the composites with some lattice fringes on different sides are shown in Fig. S9 and S10. Moreover, from the morphological perspective, the SEM micrograph of pristine MgH2 (Fig. 7f) shows micro-particles with a size distribution >30 µm. High energy milling reduces the size below 10 µm (Fig. 7g) with the size distribution profile shown in Fig. S11. The addition of NaTiOxH catalyst (Fig. 7h) further reduces the size of MgH2 coupled with an increased contact which could help promote its kinetics of de/re-hydrogenation [34]. The SEM micrograph of the dehydrogenated sample (Fig. S12) shows fine sponge-like Mg particles with bright patches on the surface attributed to MgO. A slight expansion of the particles could be observed on hydrogenation (Fig. S13). This expansion is believed to promote contact between the particles of Mg/MgH2 and NaTiOxH which could facilitate the reversibility process as observed in Fig. 4. Likewise, EDS mapping of the composite (Fig. 7i) also reveals the bright spots of Na, Ti, and O species well dispersed around the network of MgH2; a phenomenon that could positively influence the electron dynamics around MgH2 for an improved hydrogen storage performance. Furthermore, titanium 2p XPS spectral of the as-milled sample (Fig. 7j) shows its existence in four (4) different states similar to Fig. 6(g) above. However, the addition of MgH2 causes a significant electronic effect around Ti3+ at 457.0 eV, and Ti2+ at 454.8 eV as compared to the Ti spectra of NaTiOxH catalyst in Fig. 6(g). This kind of shift in binding energy (∼1 eV difference) was previously attributed to the reduction in ionic contribution in the respective titanium chemical bond formation; the unstable ionic character could arise due to factors such as lattice distortion, hybridization, and crystal field stabilization effects [68–71]. On dehydrogenation, the compositional changes around the valences due to H2 release result in a slight shift to the higher binding energy region. Interestingly, the Ti0 state at 452.5 eV in the as-milled sample which corresponds to titanium metal disappears. Considering the high chemical reactivity between titanium and oxygen, the reason for this disappearance could be attributed to the oxidation of the titanium metal into higher state sub-oxides even as more oxygen atoms are likely to be present after dehydrogenation [41,54,72–74]. After re-hydrogenation, a similar pattern to the as-milled sample in Fig. 7(j) above re-emerges (Fig. S14). This indicates probably the occurrence of a redox reaction via H2 insertion and removal; a process that causes some sort of vacancy creation and elimination between the valence and conduction bands of Ti-O bonds as previously reported by the following studies [17,75]. The Oxygen 1s data plot of the as-milled sample (Fig. S15) shows the bond states corresponding to phases of Mg-O, Na/Ti-O, hydroxides, and/or peroxides [65]. The binding energy shift of the corresponding O-based species (while Mg-O binding energy peak remains) is also consistent with the redox process as shown in Fig. S16. It should be noted at this point that the Na2Ti3O7 species formed in-situ remains after 100 cycles of de/re-hydrogenation from XRD analysis (Fig. 7a); hence, a logical conclusion could be reached that a ‘topotactic’ reaction probably occurs between Mg-H and this catalytic species similar to the reported investigations on TiO2 surface topotactic reactions [73,76]. This reaction probably generates nonstoichiometric layers around the Na-doped TiO6 octahedrons which provide multiple diffusion channels that enhance the Mg-H bond de/association [17,33]. In addition, the disappearance and re-appearance of Ti3O5 species in the respective XRD patterns of the dehydrogenated and re-hydrogenated composites, and the unstable valences of Ti3+/Ti2+/Ti0 species in the composites (consistent with H2 insertion and removal), evidently confirm that the redox process around these defects facilitates hydrogen diffusion in Mg/MgH2 [17,32,75]. This is schematically illustrated in Fig. 8 . However, apart from the multivalent titanium states, and the Ti3O5/Na2Ti3O7 species formed in situ, some other components of the defective TiO2-x could have played their respective roles towards improving the electronic conductivity around the MgH2 bond; this creates the gap for further investigations.This work presents NaH doped nanocrystalline TiO2 NaTiOxH as an effective catalyst for the hydrogen storage properties of MgH2. The introduction of 2.5 wt% of the catalyst stabilized the reversibility of Mg/MgH2 up to 100 cycles with ∼6.1 wt% H2 retained afterward. Interestingly, 5 wt% of the catalyst could influence a remarkable absorption of ∼4.5 wt% H2 in 45 min at room temperature under 50 bars of H2 pressure. This composite also absorbed ∼5.0 wt% H2 in 60 min at 50 °C under 30 bars of H2 pressure. In addition to these improvements, the desorption analysis revealed that 5 wt% NaTiOxH catalyzed MgH2 could start to release H2 from ∼185 °C and desorb ∼7.2 wt% H2 within 15 min at ∼290 °C; thanks to the apparent activation energy of desorption calculated to be ∼57 kJ/mol which is 44 kJ/mol below as-milled MgH2, and 123 kJ/mol below pristine MgH2. However, the observed dehydrogenation enthalpy of ∼77 (±1.5) kJ/mol-H2 indicates that NaH only acted as a reducing agent for TiO2 without any positive influence on the thermodynamic property of MgH2. The results obtained from XPS measurement revealed that NaTiOxH is an effective catalyst for MgH2 possibly due to the existence of reduced titanium valences (Ti<4+) which showed partial/full reversibility via hydrogen insertion and removal, due to vacancy creation and elimination. Additional information from XRD, TEM, SEM, and EDS complementarily revealed an intimate contact between the homogeneously dispersed NaTiOxH and MgH2 particles. It was also observed that the presence of catalytically active Ti3O5 and “rod-like” Na2Ti3O7 formed in-situ with some other possible multivalent titanium sub-oxides in the defective black TiO2-x powder could enhance the hydrogen storage performance of MgH2 by providing multiple diffusion channels during the de/absorption.The authors declare no competing interest. Figure S1. Temperature programmed desorption curves of MgH2 catalyzed with 10wt% NaH @ TiO2 at ratio 0.5, 1, and 2:1. Figure S2. Room-temperature absorption curves of MgH2–5 wt% NaTiOxH in 12 hr at 10, and 30 bars of hydrogen pressure. Figure S3. 1st, 50th, and 100th desorption cycles of MgH2–2.5 wt% NaTiOxH at 300 °C and 0.001 bar of H2 pressure. Figure S4. 1st, 50th, and 100th absorption cycles of MgH2–2.5 wt% NaTiOxH at 300 °C and 30–50 bars of H2 pressure. Table S1. Comparisons between calculated d-spacing of phases in the catalyst and standards. Figure S5. TEM image of NaTiOxH catalyst. Figure S6. HRTEM images of NaTiOxH catalyst. Figure S7. XPS profile of Titanium 2p in pure TiO2 sample. Figure S8. XPS profile of Oxygen 1s in pure TiO2 sample. Table S2. Comparisons between calculated d-spacing of phases in the composite and standards. Figure S9. HRTEM images of as-milled MgH2–5 w% NaTiOxH. Figure S10. HRTEM images of hydrogenated MgH2–5 w% NaTiOxH. Figure S11. Size distribution profile of as-milled MgH2. Figure S12. SEM image of dehydrogenated MgH2–5 wt% NaTiOxH. Figure S13. SEM image of hydrogenated MgH2–5 wt% NaTiOxH. Figure S14. XPS profile of Ti 2p in MgH2–5wt% NaTiOxH at different phases. Figure S15. XPS profile of Oxygen 1s spectra in MgH2–5wt% NaTiOxH. Figure S16. XPS profile of O 1s in MgH2–5wt% NaTiOxH at different phases.The authors acknowledge the Project supported by the National Key R&D Program of China (2019YFE0103600, 2018YFB1502101), the Key R&D Program of Shandong Province, China (2020CXGC010402), the National Natural Science Foundation of China (51801197), the Liaoning Revitalization Talents Program (XLYC2002076), the Dalian High-level Talents Program (2019RD09), the Youth Innovation Promotion Association CAS (2019189) and K.C. Wong Education Foundation (GJTD-2018–06).

This paper presents the catalytic effect of NaH doped nanocrystalline TiO2 (designated as NaTiOxH) in the improvement of MgH2 hydrogen storage properties. The catalyst preparation involves ball milling NaH with TiO2 for 3 hr. The addition of 5 wt% NaTiOxH powder into MgH2 reduces its operating temperature to ∼185 °C, which is ∼110 °C lower than the additive-free as-milled MgH2. The composite remarkably desorbs ∼7.2 wt% H2 within 15 min at ∼290 °C and reabsorbs ∼4.5 wt% H2 in 45 min at room temperature under 50 bar H2. MgH2 dehydrogenation is activated at 57 kJ/mol by the catalyst. More importantly, the addition of 2.5 wt% NaTiOxH catalyst aids MgH2 to reversibly produce ∼6.1 wt% H2 upon 100 cycles within 475 hr at 300 °C. Microstructural investigation into the catalyzed MgH2 composite reveals a firm contact existing between NaTiOxH and MgH2 particles. Meanwhile, the NaTiOxH catalyst consists of catalytically active Ti3O5, and “rod-like” Na2Ti3O7 species liberated in-situ during preparation; these active species could provide multiple hydrogen diffusion pathways for an improved MgH2 sorption process. Furthermore, the elemental characterization identifies the reduced valence states of titanium (Ti<4+) which show some sort of reversibility consistent with H2 insertion and removal. This phenomenon is believed to enhance the mobility of Mg/MgH2 electrons by the creation and elimination of oxygen vacancies in the defective (TiO2-x) catalyst. Our findings have therefore moved MgH2 closer to practical applications.

For hydrogen to become competitive compared to fossil fuels, cost effective and sustainable catalytic materials are needed for low-temperature water-electrolysis. The oxygen evolution reaction (OER) is the main limiting step of the overall water-splitting process, as it presents a large overpotential in comparison to the thermodynamic limit of 1.229 V vs. RHE. This is the main driving force for the current research on electrolyzers for the OER. Well-studied RuO2 and IrO2 present good efficiency, however, sustainability requires the replacement of these critical elements by materials containing Earth-abundant elements [1,2]. In the last decades, an increased interest appeared for mixed hydroxides of Ni and Fe for OER application, as described in the recent review of Gao and Yan [3]. These include layered double hydroxides, which structurally consist of positively-charged, brucite-like layers with edge-sharing M(OH)6 octahedrons (M = Ni, Fe), charge-balanced by interlayer anions plus water molecules. These Ni-Fe layered double hydroxides (LDH) were reported to present low overpotential and low Tafel slope for OER in alkaline media [4,5], which are very important catalytic indicators.Although the particular platelet like structure of LDH is often mentioned as an advantage leading to a large surface area of the catalyst [6,7], it is not yet clear why this specific material is efficient towards OER and whether it is due to the elemental composition or to its particular structure. A better understanding of the OER reaction mechanisms is necessary to highlight the characteristics allowing faster kinetics in order to obtain an efficient design of this type of catalysts. The impact of several factors on the OER catalytic activity of Ni-Fe LDH has been evaluated, such as the Ni/Fe ratio of the cationic layer [8,9], the nature of the intercalated anion [10], or the effect of delamination of the LDH [6,11]. However, the wide range of experimental conditions and material analysis procedures reported, the lack of information to fully replicate the published works and, consequently, the scattering of results obtained in different works leads to difficult comparison of the published results. As a result, newcomers in the field can easily doubt on the main factor(s) behind the good efficiency claimed for these materials.It is worth to mention that current alkaline electrolyzers industrially used are often Ni-based electrodes, at which surface the reaction takes place [2]. On these electrodes, a strong layer of hydroxides with different structures and level of hydration is formed upon cycling. Some groups studied the effect of different amounts of Fe in Ni hydroxides on the efficiency towards the OER, but without ever mentioning the name of LDH structure [12]. Others reported the interesting properties of combining Ni and Fe elements in electrodeposited [12], sputtered [12], or chemically bulk-synthesized hydroxides[1].In the case of Ni-Fe in LDH structure, the observation we made is that the literature rarely considers and discuss the importance of both the ratio and the crystallinity of the catalyst in a single work, or at least not explicitly, which is the objective of this work.As regards to the ratio of metallic cations, Oliver-Tolentino et al. reported the bulk preparation of Ni-Fe LDH of MII/MIII ratios 1.5 and 2, obtaining better efficiency towards OER with ratio 1.5, containing more Fe, which is considered to allow the activation of Ni centers by partial-charge transfer mechanism [8]. Zhou et al. prepared Ni-Fe samples with different ratios, from pure Ni hydroxide to Fe oxide [9]. Of all these compositions, we consider that only the sample with MII/MIII ratio of 2 can be considered to have a LDH structure and it is the one that presented the highest efficiency on their work. Later, Görlin et al. studied nanosized Ni-Fe oxyhydroxides catalysts of different metallic ratios and obtained an optimal activity towards OER for a MII/MIII ratio of 0.8 [13]. They suggested that the good OER efficiency of the catalyst was related to the distortion of the matrix around the metallic active sites [13].These two last works open the discussion on the role of the structure and the atomic ordering of the material. It is well known that, for any catalyst, the arrangement of the atoms and structure of the material play an important role in its electrochemical behavior. In particular, it is commonly accepted that a low degree of crystallinity can lead to high amount of defects which are considered to be good active sites [14]. Hall et al., when reviewing Ni(OH)2 materials [15], verified that the reasons for this increased efficiency are still uncertain, being not clear whether the electron vacancies are induced by impurities or by high level of disorder. Electrodeposited Ni-Fe hydroxides [16] and sprayed Ni-Fe oxides [17] with amorphous structures are claimed to be very good catalysts thanks to the rough surfaces, high structural disorder and improved charges transfer rate resulting from the low order range nature of the material. Regarding the Ni-Fe LDH, the crystallinity of these materials is still poorly studied. Only three papers are reported in the review of Diogini and Strasser in 2016 under this topic [18]. Trotochaud et al. prepared amorphous electrodeposited Ni-Fe oxyhydroxide subjected to electrochemical aging, leading to an increase of long range order perpendicular to the metal cation sheets [19]. In the electrochemical tests, they obtained no clear difference of OER efficiency with or without aging, which led them to the conclusion that the long range order had no impact on the activity of the material and that structural defects did not enhance the activity [19]. This conclusion is in line with the work of Song et al., who prepared highly crystalline LDH to then exfoliate them to obtain single sheets (no stacking of the metal hydroxides sheets) [11]. Although the ordering within the 2D layer was not mentioned, they got a noticeable increase in efficiency, which was not entirely due to the increase in electrochemical surface area of the material and which may be due to a higher amount of edges giving a different electronic configuration of the active sites, similarly to defects. Later, the work of Xu et al. focused on the effect of the intercalated anion and the crystallinity of the LDH on the efficiency towards OER [20]. They used a Ni/Fe ratio of 3 and observed that the higher crystallinity led to a decrease of efficiency. Nevertheless, the smaller size of the lengthly hydrothermally treated particles also induced a higher surface area which may be the cause of the improvement. In the work of Xiong et al., highly crystalline LDH were prepared and then reduced to create defects through oxygen vacancies allowing better performance [21]. Thus, based on this literature review, it is possible to wonder to which extent is the LDH ordered structure, with its characteristic platelet morphology, relevant towards OER, when compared to a disordered Ni/Fe oxyhydroxide.To help answer this question, the present work gives insights on the importance of the ratio and crystallinity of the Ni-Fe LDH catalyst for its efficiency towards the OER. The approach followed to extract the electrochemical performance parameters can be divided into three main parts. First, the experimental setup and measurements carried out were validated using RuO2 as benchmark. This intends to prove that the conditions in which the measurements were done are line with the literature. Second, NiFe LDHs prepared with different degrees of crystallinity and Ni/Fe ratios were characterized using a method proposed by McCrory et al [22,23]. This part aims to generate data that can be directly compared with other works performed under the same conditions, thereby reducing differences in performance resulting from the breath of experimental conditions used for the preparation of the electrodes and not of the electrocatalyst properties per se. Third, the data obtained, namely overpotential and Tafel slopes, were not just reported but critically analyzed: in the case of overpotential values, statistical analysis was performed in order to identify trends beyond random fluctuations, while in the case of Tafel slopes, O2 measurements using microelectrodes were performed in order to unambiguously identify the region of O2 evolution and evaluate whether Tafel slopes could be extracted or not. Altogether, this is an original approach and it is expected to shed light on the properties that can actually affect the electrocatalyic properties of NiFe LDH family towards OER, with implications on the design of electrolyzers at industrial level.Herein, Ni-Fe LDH intercalated with carbonates were synthesized with different degrees of crystallinity through a simple hydrothermal method, for different Ni/Fe ratios. The carbonate intercalated hydroxide was chosen as it is the most stable form of Ni-Fe LDH and, hence, would present less degradation by anion-exchange in electrolytes containing potential impurities, making it an “application-friendly” catalyst. Their efficiency as catalysts for the OER was evaluated and compared with standard RuO2 as a benchmark. The Tafel slopes and overpotentials were correlated with the structural features of the synthesized nanostructured Ni-Fe. Finally, the difficulty in comparing the electrochemical results with published data is discussed.Ni-Fe/CO3 LDH were synthesized using co-precipitation method in aqueous solution to obtain materials with a Ni/Fe ratio of 2, 3 and 4 [8,24]. The LDH were prepared by dropwise addition of an aqueous solution (25 mL; pH = 12.5) of NaOH (1.17м) and Na2CO3 (0.34 м) to an aqueous solution (25 mL) of nitrate metallic salts (0.5м), under vigorous stirring at room temperature. The metallic salt solution was prepared with the desired Ni/Fe ratio. After complete addition, the resulting brownish slurry was stirred for two hours (final pH = 9). Half of the slurry volume is kept aside and called Ni/Fex-AsPrep (x = Ni/Fe ratio). The rest of the slurry was thermally treated in an autoclave at 120 °C for 24 h and named Ni/Fex-HT. All six samples were rinsed with DI water and centrifuged 3 times to eliminate any salt residues.The LDH slurries were dried overnight in oven at 80 °C and ground to a fine powder. The powder X-Ray diffraction was recorded using a PANalytical Xpert Pro instrument with Cu Kα radiation (λ = 1.5418 Å) and graphite monochromator. Phase analysis was performed using the PDF-4 + 2019 database from the International Center for Diffraction Data. The profile matching of the obtained pattern was performed with the FullProf software, using a cell of space group R-3 m (166).The morphology of the materials was observed by Scanning Electron Microscopy with a Hitachi S-4100 system using an electron beam energy of 25 keV. The sampling was made by drop-casting a suspension of the LDH powder on a Si wafer. LDH samples were also analyzed by HR-TEM, using a transmission electron microscope energy-filtered TEM EF 200 kV, JEOL brand, model 2200FS, high-resolution electron gun Schottky emission (SE), omega type energy filter column spectrometry with electron energy loss EELS.The particles sizes and Zeta potentials were measured with a Malvern ZetaSizer Nano ZS apparatus and LDH dispersions prepared from dried powders and sonicated at least 5 min in water to ensure good particles dispersion.Attenuated total reflectance infrared spectra of dry LDH powders were collected with a Bruker Optics tensor 27 Fourier Transform-IR spectrometer, equipped with a Golden Gate ATR accessory plate. The spectra were collected at room temperature in ambient air, and 128 scans were averaged for each sample.Atomic Absorption Spectroscopy was performed on the samples dissolved in HCl (37%) using an Avanta apparatus from GBC Scientific equipment with an air-acetylene flame to measure the amount of Fe (Lamp: 248.3 nm at 6 mA, with slit of 0.2 nm) and Ni (Lamp: 232.0 nm at 5 mA, with slit of 0.2 nm).A glassy carbon rotating disk electrode (GC RDE) with 3 mm diameter was used as supporting electrode after being polished successively with SiC papers (P2500 and P4000) and suspensions of alumina powder (1 µm and 0.3 µm particle size), rinsed and sonicated 2 min in DI water after each polishing and dried in air. Aqueous suspensions of catalysts (0.5 mg mL−1) were prepared from the LDH powder and sonicated 30 min. The glassy carbon RDE were modified by drop-casting the suspension (11 µL) and let dry 30 min in air, giving a final catalyst loading of 28 µg cm−2. The RDE was then covered with a thin film of ion conductive polymer by drop-casting a Nafion solution (11 µL of 5 wt% Nafion® perfluorinated resin solution from Sigma-Aldrich diluted 100 times in ethanol) and let to dry 5 min at room temperature to avoid the detachment of the catalyst. The electrode was then installed as prepared in the electrochemical cell. A powder of ruthenium (IV) oxide (Alfa Aesar), deposited with the same procedure on the electrode, was used as a benchmarked electrocatalyst to evaluate the performance of the synthesized Ni-Fe LDH.The electrochemical experiments were performed using an Autolab PGSTAT 302 N potentiostat with the GPES software. The measurements were carried out at room temperature, inside a Faraday cage, in a cell with three-electrode configuration, with the GC RDE as working electrode, a saturated calomel electrode (SCE) as reference and a Pt wire as counter electrode. The testing electrolyte was 0.1 M KOH electrolyte. Before each experiment oxygen was bubbled in the cell for at least 20 min and the RDE rotated at 1600 rpm driven by an Autolab rotator and motor controller. The electrochemical characterization followed McCrory et al. [22,23]. A conditioning stage was performed consisting of 20 cyclic voltammetry sweeps in the 0–0.8 V vs. SCE potential range at a scan rate of 10 mV s−1. This step allowed the stabilization of the catalyst and Nafion layers, verified by at least 5 identical last scans. Then, three cycles were measured at 5 mV s−1 between 0 and 0.8 V vs. SCE to extract the figures of merit: overpotential and Tafel slope. The stability of the modified electrode was tested by chrono-potentiometry for 2 h at 10 mA cm−2. All polarization potentials reported are relative to the reversible hydrogen electrode (RHE) and current densities per geometric area (0.196 cm2). The small catalyst loading and the fact of being impregnated in Nafion, prevented the use of XRD or XPS to verify compositional and structural changes after the experiments, as discussed in other works [25].Electrochemical data (overpotential values; n = 5 per each material) was submitted to statistical analysis. Data normality and homoscedasticity were previously tested using Shapiro-Wilk and the Spearman tests (p < 0.05), respectively. A two-way analysis of variance was then used to compare the statistical OER differences among synthetized materials, considering the factors “treatment type” and “Ni/Fe ratio”, followed by the Tukey’s multiple comparison test, whenever significant differences were observed (p < 0.05). The statistical analysis was performed with the software Prism version 8.Ni-Fe/CO3 LDH were synthesized by co-precipitation with a Ni/Fe ratio of 2, 3 and 4. Part of the powders were further subjected to a hydrothermal treatment (HT). As such, six samples were produced, three resulting directly from the synthesis, and named Ni/FeX-AsPrep, and three coming from the HT and named Ni/FeX-HT (where X = Ni/Fe ratio)The aspect of the prepared materials is observed by scanning electron microscope (SEM) (Fig. 1 ). The as-prepared samples present aggregates of poorly defined shapes, while the heat-treated samples display individual platelets with approximate hexagonal shape, as reported for several LDH-type materials [26–28]. More particularly, the sample Ni/Fe4-HT has the aspect of a desert rose, probably due to the higher content of Ni, as this morphology is characteristic of some Ni(OH)2 materials [15].The average particle size (Z-Ave) of the studied LDH dispersed in deionized water, which is presented in Table 1 , varied between 242 nm and 758 nm. These values should be considered more as an order of magnitude than as the exact particle sizes, because the particles have platelet-like shape and not the spherical form required by the dynamic light scattering theory [29]. Moreover, the measured entities seem to have been mainly aggregates composed of smaller particles, appearing on SEM images with widths between 50 nm and 250 nm. Table1 also presents the zeta potential (ZP) of the particles. The values ranged between + 40 and + 50 mV, allowing for the stabilization of a dilute suspension of particles in deionized water due to electrostatic repulsions (concentration < 0.1 mg mL−1), which is important for preparing well dispersed suspensions of LDH for the electrochemical measurements.The attenuated total reflectance infrared (ATR-FTIR) measurements (Fig. 2 ) confirm the presence of CO3 2– in the LDH samples, with a strong absorption band at 1350 cm−1. The large band around 3500 cm−1 corresponds to O-H stretching and the one at 1630 cm−1 to the H-O-H deformation, showing qualitatively the hydration of the LDH samples. Bands under 1000 cm−1 correspond to bonds between oxygen and metallic atoms forming the hydroxide layers [30].The results of the chemical analysis of the dissolved materials through atomic absorption spectroscopy (AAS) are displayed in Table2 . The molar ratios of the synthesized materials correspond to the expected ones, except for the sample Ni/Fe2-HT. The solution of Ni/Fe2-HT contained some undissolved particles remaining, corresponding to an oxide phase, which was hence not analyzed with the rest of the solution, leading to a lower amount of Fe measured than for Ni/Fe2-AsPrep. This observation proves that the LDH phase obtained in the case of the Ni/Fe2-HT has a higher Ni/Fe ratio than expected. Fig. 3 a displays the X-ray diffraction patterns for the three compositions investigated in this work, both as-prepared and after hydrothermal treatment. The pattern is matched with the peaks corresponding to the rhombohedral structure of Iron-Nickel Carbonate Hydroxide Hydrate of formula Ni6Fe2(OH)16(CO3)·4H2O (computed pattern JCPDS 01–082–8040). The peaks (003), (006) and (009) are characteristic of the layered structure of the 3R LDH polytype, from which can be extracted the basal spacing of 7.7 ± 0.15 Å, typical of a carbonate-intercalated LDH [28]. The sample Ni/Fe2-HT has a pattern with more defined peaks in the region 2θ = 35° to 60°. This region corresponds to the reflections of the interlayer, which let us suppose that the molecules forming the interlayer are more ordered in this sample.For each sample, the average crystallite size in the a direction is estimated by the Scherrer equation using the full width at high maximum of the (110) peak, determined by fitting together (110) and (113), often overlapped. The obtained crystallite sizes are ranging between 8 and 33 nm (Fig. 3b). It is worth to note that these values are a low-end estimate of the actual crystallite sizes, as the broadening of the diffraction peak can be due to local imperfections in the lattice such as strain and chemical heterogeneities. As expected, the crystallite size increased after hydrothermal treatment, for all the Ni-Fe LDH compositions considered in this work. Moreover, bigger crystallite sizes were obtained for samples with lower Ni/Fe ratio, which suggest a link between the amount of MIII cations in the structure and the propensity of the hydroxide layers to stack more regularly. In the case of well-defined hexagonal platelets for Ni/Fe2-HT and Ni/Fe3-HT visible on SEM pictures, the actual crystal sizes correspond to the width of the platelets, which is ranging between 80 and 120 nm, meaning 3 to 15 times the estimated crystallite sizes. Hence, the values estimated through the Scherrer equation are not to be considered as the actual size of the LDH crystal but as a parameter to assess the quality of the long-range order of the structure.To explore further the crystallinity of the samples, the profile matching of the diffraction patterns was performed using the FullProf software using a R–3 m symmetry to extract the cell parameters. Both parameters a and c were found to increase with the Ni/Fe ratio (Fig. 3c). This is consistent with the size of the cations, as the bivalent nickel is 0.08 Å bigger than the trivalent iron, according to the Shannon tables [31]. The c parameter increases with the Ni/Fe ratio, which is also consistent with the fact that the layers are less positively charged and hence, the attraction between layer and interlayer is smaller, leading to a less “compact” stacking.The XRD pattern of the sample Ni/Fe2-HT presents additional peaks which do not correspond to the Ni-Fe carbonate hydroxide structure (Fig. 3a). Such pattern has also been observed in other works synthesizing Ni-Fe LDH by hydrothermal technique [24,32] and were assigned to the spinel structure of the NiFe2O4 phase (JCPDS: 00–054-0964). The hydrothermal treatment in autoclave, carried out with the aim to enhance the growth of the LDH particles, promoted the segregation of an iron-rich phase. A similar work, where the material is thermally treated under lower conditions of temperature (50 °C) and pressure does not present the NiFe2O4 phase [8], which shows that this segregation is due to the higher temperature applied to the sample. It is worth of mention that the spinel structure is consistent with the presence of insoluble particles found when dissolving the Ni/Fe2-HT sample for AAS measurements was attempted (cf. Table 2).Transmission electron microscopy (TEM) was used to more clearly identify some features evidenced in SEM and XRD analyses. TEM images depicted in Fig. 4 reveal that the obtained Ni-Fe LDHs present a plate-like morphology, with some hexagonal shapes being identified, especially for samples subjected to hydrothermal treatment (Ni/Fe2-HT, Ni/Fe3-HT, Ni/Fe4-HT). Moreover, particle size of individual particles increases after hydrothermal treatment.The sample Ni/Fe2-HT, which revealed a secondary phase by XRD (recall Fig. 3), actually presents a few particles which have different shape and density (highlighted with an orange circle in Fig. 4). This could be associated with the NiFe2O4 formed after hydrothermal treatment. However, it must be mentioned that due to the overlapping of these particles with LDH particles, diffraction analysis could not be performed to unambiguously identify this secondary phase.Cyclic voltammograms (CV) obtained for the neat glassy carbon electrode, the RuO2 benchmark and the LDH samples are displayed in Fig. 5 after ohmic drop correction of the data. The glassy carbon substrate presents no current in the studied potential range. The electrochemical procedure was first validated by experiments using commercial RuO2 powder as a benchmark. Working electrodes were prepared following the procedure reported by Jung et al. with commercial RuO2 [22], with a catalyst loading of 800 µg cm−2. We obtained a similar Tafel slope (67 mVdec-1) and an overpotential of 290 ± 11 mV, lower than the reference value (the overpotential at 10 mA cm−2 was 380 ± 20 mV and the Tafel slope was 65 mVdec-1) [22]. Following these tests, the LDH loading was reduced to 28 µg cm−2 in order to obtain a stable film, not possible with the 800 µg cm−2 loading.The CVs of the different prepared electrodes (Fig. 5) exhibit an increasing catalytic current for potentials more positive than 1.5 V vs. RHE, corresponding to the onset of the OER. In comparison, the electrode prepared with RuO2 showed a catalytic current lower than the LDH samples, leading to an overpotential of 500 mV at 10 mA cm−2. The curves of the LDH are very similar to one another, however, the current slope corresponding to the OER was higher for AsPrep samples, leading to lower overpotentials at 10 mA cm−2 compared to the hydrothermally treated samples. In the case of LDH with a Ni/Fe ratio of 4, an additional wave appeared at 1.49 V (Ni/Fe4–AsPrep) and 1.52 V (Ni/Fe4-HT), which was attributed to the NiII/NiIII oxidation [8]. In samples with lower Ni/Fe ratio, this feature is less visible due to the lower Ni content and to its overlapping with the OER peak. Confirming the change of the Ni state at this potential, a change of color from light to dark brown was detected, more perceptible in the materials with higher Ni. In the reverse scan, the NiIII/NiII reduction peak appeared in each Ni/Fe LDH sample, from 1.43 V to 1.40 V, in the order Ni/Fe2-HT > Ni/Fe2-AsPrep > Ni/Fe3-HT > Ni/Fe3-AsPrep > Ni/Fe4-HT > Ni/Fe4–AsPrep. This order may illustrate the availability of Ni atom in each LDH for changing its oxidation state upon polarization.In the E vs. log(i) plots of the cyclic voltammetry data obtained for each of our samples (Fig. 6 a), different regions are visible. The main linear region appears at low currents (0.01 to 1 mA cm−2) and is identified with red lines in the graph. It is attributed to the oxidation of NiII to NiIII, with a low Tafel slope, between 20 and 30 mVdec-1, highlighting the fast reaction kinetics. To confirm that OER was taking place at higher potentials, the voltammograms were repeated with an O2 micro-sensor close to the electrode surface for detecting the potential at which O2 was formed. For these experiments the glassy carbon electrode did not rotate and was facing up. The micro-sensor was a 10 μm platinum disk polarized at a potential were O2 is reduced with limiting control (in this experiment −0.8 to −1V vs SCE). The idea is that the current measured by the microelectrode will be constant and proportional to the concentration of dissolved O2 in the bulk solution but should immediately increase as soon as O2 starts to be generated in the glassy carbon electrode with catalyst. The result is presented in Fig. 6 b) and shows the superposition of the current measured in the GC electrode with LDH catalyst and the current measured by the O2 sensing microelectrode. The × axis depicts the potential at the GC. The O2 reduction current measured at the microelectrode is negligible until the GC potential reaches 0.5 V vs SCE. Then, for more positive potentials, the current increases rapidly until the saturation of the measuring device (1nA) is reached. This experiment was performed with all LDH samples and confirmed that O2 was produced at potentials higher than the linear region of the Tafel plot. In these plots, the region corresponding to the OER does not present a Tafelian behavior, making difficult its determination with confidence. This supports the suggestion of McCrory et al. about the possible change of reaction mechanisms with change of potential [23]. Based on these findings it was decided to not use in this work the Tafel slope as a figure of merit to assess the efficiency of the catalyst. Fig. 6 The overpotential values measured on the ohmic drop-corrected plots at 10 mA cm−2 are reported in Fig. 7 a. Each electrode was also submitted to a stability test to observe the evolution of the overpotential while applying a current of 10 mA cm−2 for two hours (Fig. 7b). No important degradation of the electrodes has been observed as the increase of the overpotential was kept under 50 mV for all electrodes. The mean overpotentials varied between 330 mV and 410 mV. These values are higher than the reported for the best materials [4,20,21], but it is important to note that in this study a low catalyst loading was used and no conductive carbon material was added to the ink. Furthermore, the aim of this paper is to compare samples in the same conditions, as the comparison with materials from other publications using the overpotential is not accurate, owing to the number of experimental parameters that can influence the results. Qualitatively, two main findings can be highlighted: (i) the overpotential decreased in the AsPrep samples with increasing Ni/Fe ratio and (ii) the overpotential of the Ni/FeX–HT samples was larger for any of three Ni/Fe ratios surveyed, which seems to indicate that the higher crystallinity due to the hydrothermal treatment leads to lower efficiency. Fig. 7 The observations performed in the previous paragraph are based on the comparison of the mean values of the figure of merit chosen to describe the efficiency of the electrocatalysts (cf. Fig. 7a). An analysis of variance (ANOVA) of the results of the overpotential was performed using the 2-way ANOVA test to evaluate the effect of the hydrothermal treatment and of the Ni/Fe ratio. To the best of our knowledge, no study on NiFe LDH for OER catalysis takes into consideration the importance of samples replication and the data variability. In the present study, five replicates of electrode for each LDH were tested, from which we extract the overpotential value. The assumptions of data normality and homoscedasticity were successfully verified through Shapiro-Wilk and Spearman’s test, respectively, before performing the 2-way ANOVA tests. The ANOVA statistical analysis highlighted the effective impact on the thermal treatment on the catalyst efficiency: a hydrothermally treated LDH sample is more crystalline and present a significantly higher overpotential than its non-treated counterpart (F(1,24) = 22.48, p < 0.0001), while the Ni/Fe ratio caused no statistically relevant differences in measured OER values (F(2,24) = 2.29, p = 0.1225) (pls. cf. Table S1).The Tukey’s multiple comparison test showed that there is difference in the overpotential values measured between samples of Ni/Fe4 LDH AsPrep vs. HT (pls. cf. Table S1). No significant overpotential differences were found in the comparison AsPrep vs. HT for the other two ratios, neither between samples with similar hydrothermal treatment and different ratios (pls. cf. Table S1).This study raises awareness about results reported in the literature taking conclusion on the impact of metal ratios or crystallinity of the sample without mentioning repeatability of the electrochemical measurements. The preparation of the working electrode can have as much impact as the difference of Ni/Fe ratio in the ranges studied in this work.Firstly, the electrochemical characterization presented in section 3 highlights that the layered double hydroxides prepared in this work present a lower overpotential than the RuO2 reference material, justifying once again the efforts in better understanding and developing NiFe electrocatalyst LDHs.Then, when comparing samples within AsPrep or HT group separately, we observe that the increase of Ni/Fe ratio implies an increase in the cell parameters a and c linked with a decrease of the positive charge of the LDH layer due to less divalent Fe atoms. Many references [8,9,13], report a better efficiency for lower Ni/Fe ratios (1.5, 2 and 0.8 respectively), as if a higher concentration of Ni would be unfavorable for the catalyst efficiency. However, from the results of this study, the changes in composition and structure do not introduce significant electrocatalytic differences.On the other hand, the statistical analysis of the electrochemical data shows that using a more crystalline material on the electrode results in a lower electrocatalytic performance. In our work, it was seen for NiFe4-AsPrep vs. NiFe4-HT but less clearly for lower Ni/Fe ratios. These results are consistent with a work reported by Xu et al. in 2015, which shows, through the comparison of overpotentials, that a LDH material with a Ni/Fe ratio of 3 presents lower efficiency when it is hydrothermally treated, mentioning that lower crystallinity provides less confined active sites [20]. However, the discussion does not make the distinction between the atoms ordering and the size of the LDH particles inducing a higher surface area, which could be the cause for the improvement. Another work, from Gao et al. also concludes that the amorphous nature of the material makes it more flexible and hence more stable over time to electrochemical processes in comparison to crystalline materials [16], although their material is preconditioned for a long time to achieve the higher catalytic efficiency, which eventually may lead to a rearrangement of the material in more stable phases. The interest of defects for electrocatalysis is also discussed in the work of Trotochaud et al., in 2014, which mentions that the increase of efficiency of a β-(NiOH)2 is due to the inclusion of Fe impurities, and not to the more ordered structure than in the α-(NiOH)2 [19]. They conclude that the long-range order in the material seems unimportant. Görlin et al. discussed this point in a work on nanosized Ni-Fe oxyhydroxides catalysts of different metallic ratios and highlighted that the higher OER efficiency of the catalyst is related to the distortion of matrix around the metallic active sites [13]. Hence, from the results obtained and what is seen in the literature, it seems that more disordered material is more efficient, in terms of overpotential.Mostly, the present study reinforces that the importance of the LDH phase in the NiFe mixed hydroxide lies in the fact that this structure allows a “meta-stable” phase of the hydroxide, preventing phase segregation, but the long-range order or the platelet-like morphology does not seem necessary and would even be detrimental for the electrochemical efficiency of the material. In parallel, the effect of the metallic cation ratio is not clear. More fundamental studies would help determine if the discussion of the influence of the Ni/Fe ratio is overrated as a single factor or if it is a combination of the presence of Fe with a distorted β-Ni2(OH) structure that can be the right direction for the development of better NiFe-based electrocatalysts.From an application-based point of view, the statement that the hydrothermal treatment is detrimental for the catalytic efficiency is interesting as it allows to remove a time and energy-intensive step from the catalyst production process.On a more general note, there are several recent works in the literature reporting promising eletrocatalysts for different reactions [33–35]. The strategy followed in this work could be extended in a general way to the design of similar materials.This work exposes views on the currently highly studied NiFe layered double hydroxides for efficient catalysis of the OER. Both the Ni/Fe ratio and the crystallinity of the synthesized material are investigated to highlight their influence on the electrocatalytic activity of the Ni/Fe LDH. No evidence was found for any impact of the Ni/Fe ratio in the efficiency of the OER but the hydrothermal treatment performed to obtain a higher crystallinity of the catalysts leads to a decrease of their efficiency. In conclusion, this work renders insights on the structure and Ni/Fe ratio, which are relevant for the design of NiFe LDHs. Future works will involve the use of other catalyst supports to increase the loading of catalyst and test the material closer to the application conditions, namely using larger current densities. Data availability The raw/processed data required to reproduce these findings cannot be shared at this time as the data also forms part of an ongoing study.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.Thanks are due to FCT/MCTES for the financial support to CICECO-Aveiro Institute of Materials (UIDB/50011/2020; UIDP/50011/2020) and CESAM (UIDP/50017/2020 + UIDB/50017/2020), through national funds. We thank also the European Commission funding the project NANOBARRIER (Reference N°280759) through the programme FP7-NMP. This study was also carried out in the framework of the NANOGREEN R&D project (CIRCNA/BRB/0291/2019) funded by national funds (OE), through FCT. Roberto Martins and Alexandre Bastos funded by national funds (OE), through FCT – Fundação para a Ciência e a Tecnologia, I.P., in the scope of the framework contract foreseen in the numbers 4, 5 and 6 of the article 23, of the Decree-Law 57/2016, of August 29, changed by Law 57/2017, of July 19 (CEECIND/01329/2017), and in frame of SMARTAQUA project, which is funded by the Foundation for Science and Technology in Portugal (FCT), the Research Council of Norway (RCN), Malta Council for Science and Technology (MCST), and co-funded by European Union’s Horizon 2020 research and innovation program under the framework of ERA-NET Cofund MarTERA (Maritime and Marine Technologies for a new Era).Supplementary data to this article can be found online at https://doi.org/10.1016/j.matdes.2021.110188.The following are the Supplementary data to this article: Supplementary data 1

The Oxygen Evolution Reaction (OER), half-reaction of the water-splitting process for hydrogen production, suffers from sluggish kinetics. NiFe materials appeared as interesting catalytic materials for this reaction in alkaline electrolyzers and have been studied, particularly in the form of NiFe Layered Double Hydroxides (LDH). However, the impact of the specificity of the atomic arrangement in the LDH and of its composition on the catalytic efficiency of the material are still unknown. Herein, LDH are synthesized with Ni/Fe ratios from 2 to 4 and different levels of crystallinity to assess their electrocatalytic behavior in 0.1 M KOH. Statistical analysis of the electrochemical results allows to highlight that, while no effect from the atomic ratio is observed, an increase in the crystallinity of the LDH seem detrimental to the catalytic efficiency.

Benzophenone reduction is a common process in chemical industry. Its main product, diphenylmethanol (also known as benzhydrol), has an important role in perfume and pharmaceutical manufacturing. Many different benzophenone reduction methods have already been reported such as photoelectrochemical method [1], electrocatalysis [2,3], simple reduction using reductive compounds [4–6], or catalytic hydrogenation [7–15]. From these methods, catalysis seems to be the most promising because of its lower chemical need, high conversion, and good selectivity. Furthermore, according to the paper of Cirtiu et al. [2], catalytic hydrogenation is 10 times faster than electrocatalysis; thus, its investigation seems to be the most beneficial. Depending on the reaction conditions, over-reduction of benzophenone to diphenylmethane is possible (Fig. 1 ).Carbon-based catalysts are often used in heterogeneous catalysis due to their high specific surface area, good adsorption, and chemical stability [16–18]. Activated carbon might be the most used carbon material (e.g. in granulated form under the trademark name of Norit) in catalysis, but carbon nanotubes also seem to be promising and available because of their constantly decreasing price. Carbon-supported catalysts have already been tested in many different hydrogenation reactions with high conversion and selectivity [19–25]. However, many publications evidence that carbon nanotubes in other hydrogenation reactions can be even better support than the commonly used activated carbons [26–28]. According to other researchers, the presence of covalently bonded nitrogen renders catalysts to be selectively inhibited to produce an intermediate or a product with high selectivity [29,30]. Based on this observation, one can expect N-doped carbon nanotubes to be selective in hydrogenation processes. In this paper we compare the efficiency and selectivity of palladium-decorated activated carbon (Norit) and carbon nanotubes in benzophenone hydrogenation.For catalyst synthesis anhydrous palladium(II) acetate (Fluorochem Ltd.) were used. The hydrogenation tests were carried out using benzophenone (99%, Alfa Aesar), hydrogen (4.5, Messer), and tetrahydrofuran (≥99%, VWR Chemicals). As catalyst support, activated charcoal Norit (Norit RBAA-3, rod, Sigma-Aldrich) was used in grinded, powder form. For GC (gas chromatography) calibration the following compounds were used: benzophenone (99%, Alfa Aesar), benzhydrol (99%, Acros Organics), dicyclohexyl ketone (98%, Sigma-Aldrich), dicyclohexylmethanol (98%, Alfa Aesar), diphenylmethane (≥99%, Alfa Aesar), cyclohexyl phenyl ketone (98%, Sigma-Aldrich), and cyclohexyl(phenyl)methanol (99%, Sigma-Aldrich). Reaction samples for gas chromatography coupled with mass spectrometry (GC-MS) analysis were diluted using methanol (99.8%, GC grade, Merck). Acetophenone (≥99%, Sigma-Aldrich) was used as an internal standard for GC-MS analysis.Nitrogen-doped BCNTs, which was used for the Pd containing catalyst preparation, were produced by the catalytic chemical vapor deposition (CCVD) method. MgO with 5 wt% nickel was used as catalyst in CCVD synthesis at 973 K. The carbon source (butylamine) was fed into the reactor by a syringe pump (6 mL h−1), where it vaporized and was carried into the catalyst bed with nitrogen gas (100 mL min−1). Because the carbon source is a nitrogen containing substance, this method results in BCNTs. The nickel was removed from the nanotubes by hydrochloric acid. The purity of BCNTs was checked by using of thermogravimetric analysis, the carbon content was 95.9 wt%.The carbon-based material (Norit or BCNT, 1.9 g) was dispersed in distilled water by a Hielscher Ultrasound tip homogenizer. Into the dispersion, aqueous solution of Pd(OAc)2 (211 mg in 10 mL water) was added, and the mixture was sonicated for 10 min. After evaporating the water using a vacuum rotary evaporator, the residue was dried at 378 K overnight. The catalysts were activated in a hydrogen flow at 673 K, 30 min. The efficiency of the reduction was checked by high resolution transmission electron microscopy (HRTEM) and X-ray diffractometry (XRD). The theoretical metal content was 5 wt%, which was checked and supported by inductively coupled plasma - optical emission spectrometry (ICP-OES).The morphology and size of the noble metal particles on the support surface were examined by HRTEM (FEI Tecnai G2) operating at an accelerating voltage of 20 kV. The sample preparation was carried out by dropping from the aqueous suspension onto 300 Mesh Cu grid (Ted Pella Inc.) The size of the nanoparticles was measured by using the ImageJ software. Based on the size of the scalebars of the HRTEM images, 100 nanoparticles were measured randomly. The size distribution histograms were created by OriginPro 8.The confirmation of metallic phases of the noble metal was checked by a Bruker Advance D8 X-ray diffractometer (Cu-Kα source, 40 kV and 40 mA generator settings), in parallel beam geometry (Göbel mirror) and with Vantec1 detector.To measure the surface area of the catalyst, Brunauer–Emmett–Teller (BET) method was used. The nitrogen adsorption measurements were carried out using a TriStar 3000 type instrument on 77 K.The nitrogen content of the carbon supports was measured by Vario Macro CHNS element analyzer. The certificated standard material was sulphanilamide (N: 16.25%, C: 41.81%, S18.62%, H: 4.65%, Elementar Analysensysteme GmbH). The carrier gas was helium (99.9990%), while oxygen (99.995%) was used for oxidation.The incorporated nitrogen forms were characterized by the X-ray photoelectron spectroscopy method with SPECS XPS equipped with a PHOIBOS 150 MCD analyzer (MgKα and AlKα).The turn-over number (TON) was calculated using eq. (1), for both catalysts to compare their efficiency. (1) T O N = n b e n z h y d r o l n P d where nbenzhydrol is the amount of the product (mol), while nPd stands for the amount of Pd (mol). The TON value is suitable to compare catalysts with different metal content because it correlates the amount of the product to a unit of catalytically active metal [31].The catalytic hydrogenation was carried out in a stainless steel reactor (200 mL total volume) equipped with a heating jacket (Büchi Uster Picoclave system). The load volume was 150 mL. The concentration of the benzophenone solution was 10 mM. Tetrahydrofuran was used as the solvent. The pressure of hydrogen was constant 20 bar during the tests. Catalyst loading was constant, 0.1 g in all tests. After starting the hydrogenation, samples were withdrawn from the reactor at 5, 10, 15, 20, 30, 40, 60, 80, 120, 180 and 240 min and analyzed using GC-MS technique. Conversion of benzophenone (X %) and selectivity for benzhydrol (S %) were calculated according to eqs. (2) and (3), respectively, where n is the amount of substance in moles. (The stoichiometric ratio is 1:1 in all the possible reduction steps.) (2) X % = n b e n z o p h e n o n e , c o n s u m e d n b e n z o p h e n o n e , i n i t i a l × 100 % (3) S % = n b e n z h y d r o l n p r o d u c t s × 100 % The GC-MS is a suitable method for both qualitative and quantitative determination of the possible reaction products. The progress of the hydrogenation was followed by a Shimadzu GCMS-QP2020 mass spectrometer-coupled gas chromatograph. Chromatographic separation was performed using a Stabilwax-MS capillary column (30 m length × 0.25 mm i.d., 0.25 μm film thickness) from Restek Corp. The column temperature program was as follows: 130 °C (1 min), 130–250 °C (10 °C min−1), 250 °C (2 min). The injector and the detector were set at 250 °C and 230 °C, respectively. Helium was used as the carrier gas at 0.86 mL min−1 column flow rate. The electron ionization mode was performed with 70 eV electron energy. MS cut time was 3 min. 1 μL of the sample was injected by a Shimadzu AOC-6000 autosampler. The split was set at 1:10. Before injection, 40 μL of the reaction mixture sample plus 100 μL of the internal standard solution (acetophenone in methanol) was diluted to 1000 μL with methanol.The nitrogen content of the two carbon structures was measured by the CHNS (carbon, hydrogen, nitrogen, sulfur) method. According to the results, the BCNTs contained 6.19% of nitrogen, whereas the Norit had only 0.54% of nitrogen. This large difference had a significant effect on the outcome of the hydrogenation reactions [32].The metal content of the catalysts was checked by the ICP (inductively coupled plasma) method. In case of the Norit-based catalyst, 2.94 w/w%, meanwhile in case of the BCNT-based 3.53 w/w% Pd content were determined.According to the specific surface area measurements, the BET surface areas of 5 wt% Pd/Norit and 5 wt% Pd/BCNT were 487.47 m2/g and 146.62 m2/g, respectively.The confirmation of metallic phases of the noble metal was checked by the XRD method. The diffractogram of the Pd/Norit catalyst shows reflexions of (111), (200), and (220) phases at 39.9°, 46.5°, and 68.0° two theta degrees, respectively, confirming the presence of the elemental Pd phase (Fig. 2 A). The C(002) and C(010) phases are also present in the XRD pattern. On the XRD pattern of the Pd/BCNT catalyst, the reflexion peaks of the elemental Pd are visible (Fig. 2 D). Characteristic reflections of carbon (002, 010) were also found. Moreover, the Ni(011) peak is present at 44.8° two theta degree. This can be explained by the synthesis method of the carbon nanotube, during which nickel as a catalyst metal was used to grow the BCNTs. These nickel nanoparticles are closed inside of the carbon nanotubes; thus they are not removable by chemical purification (SI Fig. 1). In this sense, the nickel particles are not available for catalysis.The morphology of the catalysts was examined by HRTEM. In the HRTEM picture of the Pd-Norit catalyst, Pd nanoparticles with small size and homogenous dispersibility on the surface of the activated carbon layer can be seen (Fig. 2 B). According to the size measurements, most of the nanoparticles are in the range of 3–10 nm. The average size is 5.3 ± 3.2 nm (Fig. 2 C). In case of the Pd-BCNT the average size of Pd particles was 8.1 ± 1.8 nm (Fig. 2 D).The particle size is influenced by the force of interaction between the adsorbent and the adsorbed metal and their ions. More adsorption interaction mechanisms play role at the same time during the anchoring of catalytically active metals and their ions. Namely ion-exchange, electrostatic interaction, complexation, and physical adsorption provide the adsorption of metal ions on CNT surfaces. N-BCNTs have more defect sites than their single-walled or multiwalled counterparts, and because ofthe incorporated nitrogen atoms they also have special adsorption points which are excellent spots for catalytically active metal particles. The N-doped carbon nanotubes are easier to oxidize than their non-doped counterparts owing their structure; thus on the surface of N-CNTs many oxygen contained surface groups are located, which are formed during the synthesis. The mentioned functional groups mainly hydroxyl and carboxyl groups, which play role in the metal ions adsorption. Moreover, the adsorption mechanism is also affected by the complex formation between palladium and the incorporated nitrogen atoms in case of N-BCNT. The nitrogen incorporation weakens the π–π interaction between the neighboring carbon atoms; thus the metal ions (palladium) can establish donor–acceptor interaction with the π-system of N-BCNTs. These stable adsorption interactions help to avoid the excessive crystal growth; thereby small nanoparticles are formed.The carbon nanotube–based catalyst was also examined by the XPS method (Supplementary Information I).Compositions of the reaction mixtures after a given reaction time at different temperatures are shown in Fig. 3 . Besides benzophenone, benzhydrol, and diphenylmethane, other possible reaction products and intermediates, namely dicyclohexyl ketone, dicyclohexylmethanol, cycohexyl(phenyl)methanol, and cyclohexyl phenyl ketone, were produced usually either in trace amount, or their sum of yields was below 0.5%. At 283 and 293 K, conversion of benzophenone was faster using the BCNT than using the Norit-supported catalyst. This trend changed at 323 K, while at 313 K the speed of conversion was roughly equal with the two catalysts. As it is concluded later on from the Arrhenius plots, the conversion rate for the Norit-supported catalyst is more temperature dependent than that of the BCNT-supported catalyst. With Norit, depending on the temperature, 14.3–98.4% of the benzophenone was converted in 240 min. Using BCNTs, the conversion was in the range of 20.8–96.3% (Fig. 4 ). Regarding the product distribution, there is a marked difference in the selectivity of the two types of support. While BCNT is highly selective (98.5–99.3%) toward the formation of benzhydrol at all the studied temperatures, only 38.0–76.6% selectivity for benzhydrol could be achieved using the Norit-based catalyst (Fig. 4). With BCNTs, the concentration of benzhydrol is continuously increasing throughout the studied time period (240 min) independent of the reaction temperature. In case of Norit, the same trend can be seen at lower temperatures (283 K and 293 K), while at 313 K and 323 K it reaches a maximum at about 180 min and 80 min, respectively. Diphenylmethane concentration remained low for the BCNT runs (0.9–1.5% selectivity). Using Norit, it was much higher (23.4–62.0% selectivity).Kinetics of the reactions was also studied. Plots of benzophenone concentration over time showed first-order kinetics (Fig. 5 ). The reaction rate constants (k) were determined by applying non-linear regression using eq. (4). (4) c b e n z o p h e n o n e = c b e n z o p h e n o n e , i n i t i a l × e − k × t The results are summarized in Table 1 .The lower rate constants observed at 283 and 293 K for Norit becomes higher as reaction temperature is increased to 313 and 323 K. Clearly, the rate constant for Norit exhibits steeper temperature dependence, than BCNT. Activation energies for both catalysts were determined based on the Arrhenius plots (Fig. 6 ). Using linear regression, activation energies of 64.0 ± 3.0 kJ/mol for Pd/Norit and 45.2 ± 3.6 kJ/mol for Pd/BCNT catalysts were calculated. R 2 values were 0.995 and 0.987, respectively.The TON was calculated using eq. (1), for both catalysts to compare their efficiency (Table 2 ). As it can be seen the TON values of the BCNT catalyst is continuously increasing by the temperature. However, in case of the Norit catalyst the TON values start decreasing at around 313 K. This phenomenon is caused by the over-hydrogenation effect of the Norit catalyst, as it started to produce increased amount of diphenylmethane (only 38% benzhydrol selectivity at 323 K). Furthermore, despite the lower reaction rate of the BCNT-based catalyst than the Norit, it still produced more benzhydrol.In this research, two Pd-contained carbon-based catalysts were characterized and used in benzophenone hydrogenation. Both the Norit (activated carbon) and the nitrogen-doped carbon nanotube based catalysts were tested on four different temperatures for kinetic calculations. At lower temperatures (283 K and 293 K) the BCNT and at higher temperatures (313 K and 323 K) the Norit-based catalyst had higher reaction rate. At the highest temperature applied (323 K) both catalysts provided high benzophenone conversion (BCNT: 96.3%, Norit: 98.4%). However, in case of the Norit the selectivity was relatively low (76.6%), while using BCNT the result achieved was excellent (99.3%). The high benzhydrol selectivity might be explained by the presence of covalently bonded nitrogen atoms in the catalyst (BCNT: 6.19 w/w%, Norit 0.54 w/w%) that can inhibit the over-hydrogenation process; thereby BCNTs are better catalyst supports for benzhydrol production than the commonly used activated carbon supported catalysts. Furthermore, in spite of the approximately three times lower BET surface are of the BCNT-based catalyst than the Norit, the BCNT-based catalyst seemed to be more efficient than the commonly used activated carbon–based one.This research was supported by the European Union and the Hungarian State, co-financed by the European Regional Development Fund in the framework of the GINOP-2.3.4-15-2016-00004 project, aimed to promote the cooperation between the higher education and the industry. Á. Prekob: Conceptualization, Investigation, Writing - original draft. L. Vanyorek: Conceptualization, Supervision, Writing - review & editing. Z. Fejes: Conceptualization, Validation, Supervision, Writing - review & editing.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.The following is the Supplementary data to this article: Multimedia component 1 Multimedia component 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.mtchem.2020.100409.

Catalytic activity of palladium catalysts with two different types of carbon support, Norit (an activated carbon), and bamboo-shaped carbon nanotubes (BCNT) have been tested for benzophenone hydrogenation. The selectivity toward the two possible reaction products (benzhydrol and diphenylmethane) can be directed by the catalyst support. It has been found that the Norit support preferred the over-hydrogenation of benzhydrol to diphenylmethane. The BCNT support proved to be much more selective and resulted as much as 99.3% benzhydrol selectivity at 96.3% benzophenone conversion. The high benzhydrol selectivity might be explained by the presence of covalently bonded nitrogen atoms in the catalyst (BCNT: 6.19 w/w%, Norit 0.54 w/w%) that can inhibit the over-hydrogenation process, thereby BCNTs are better catalyst supports for benzhydrol production than the commonly used activated carbon–supported catalysts.

Hydrogen generation from photoelectrochemical (PEC) water-splitting provides a promising route to store the intermittent solar energy, and serves as a sustainable and environmentally-benign alternative to the existing hydrogen production technologies. The essential component of these PEC water-splitting systems is the electrocatalyst that efficiently expedites the kinetics of electrode reactions [1], including oxygen evolution reaction (OER) at anode and hydrogen evolution reaction (HER) at cathode. The state-of-art electrocatalysts are based on noble metals, such as Pt for HER [2], and IrOx for OER [3], but their scarcity and high cost limit their practical application on a global scale. Consequently, great efforts have been dedicated to developing highly efficient and robust electrocatalysts, which are low-cost and can be prepared and integrated into the PEC water-splitting systems by simple and scalable methodologies. For example, Earth-abundant MoS2 [4,5], nickel-iron hydroxides [6,7], metal phosphides [8,9], cobalt phosphate [10], calcium iron oxides [11] can be prepared and integrated onto the photoelectrodes for PEC water-splitting simply by solution based approaches, such as (photo-)electrodeposition or spin-coating.Hydrogen generation can also be assisted from the reforming of renewable organics and chemical waste [12–20]. In particular, hydrogen generation from photocatalytic reforming of plastics, such as polyethylene terephthalate (PET), synthesized by condensation polymerization of ethylene glycol (EG) with terephthalic acid (TA), provides several distinct advantages over water-splitting, including less energy required for hydrogen generation, and mitigation of the environmental threat posed by plastics waste. To date, several electrocatalytic and photo-(electro-)catalytic systems have been developed for the depolymerisation or reforming of plastics into hydrogen. For example, Hori et al. reported an electrocatalytic system, consisting of Pt/C anode and Pt/C cathode, for conversion of plastic waste into hydrogen at 200 °C [16]. Jiang et al. developed solar thermo-coupled electrochemical process using nickel as the electrocatalyst for the depolymerization of polypropylene into methane and hydrogen [17]. Ag2+ ions was found to be capable of mediating the electrochemical oxidation of aliphatic polymers [21]. On the other hand, efficient hydrogen production from reforming of plastics can also be achieved photocatalytically by using Pt modified TiO2 nanoparticles [18], toxic CdS/CdOx quantum dots [19], and nickel phosphide modified C3N4 [20] as photocatalysts. Nonetheless, these developed systems suffered from wide distribution of products from the oxidation of plastics [16–21]. The wide product distribution due to the low selectivity requires complicated separation steps and therefore imposes an additional cost for whole hydrogen generation process. Consequently, to establish an efficient and economically attractive PEC or photocatlytic platforms for hydrogen generation from reforming of plastics, the development of the electrocatalysts that can efficiently catalyze plastics oxidation with high selectivity towards specific chemicals, especially for valuable platform chemicals (e.g., formic acid), is, therefore, of great importance.Nickel phosphides (NiPx) and their alloys have been discovered as efficient electrocatalysts for the electrochemical reactions of importance [22–31]. For example, their unique surface structure, consisting of both proton-accepting sites and hydride-accepting sites, promotes the catalysis of HER [22,28,29]. Besides, under appropriate anodic conditions, their surface can be in-situ converted into active nickel oxyhydroxide species responsible for OER [23,29] and electrocatalytic oxidation of organics [27,31]. To date, several synthetic strategies have been established for the preparation of NiPx, but most of them required costly, environmentally harmful, and energy-intensive conditions [22–27,29–31], which not only prevents the large-scale production of NiPx, but also impedes direct integration of Ni-P into the PEC devices for hydrogen generation.In the present contribution, we report on the facile electrosynthesis of nickel-phosphorous alloy nanoparticulate thin film (nanoNi-P) and its applications towards overall water-splitting and reforming of EG and PET plastics. Through the detailed investigation on the effects of electrosynthetic conditions, the factors influencing the chemical composition, surface morphology, and the resultant HER activity of nanoNi-P are elucidated. nanoNi-P and its composite with carbon nanotubes with high HER activity can be directly deposited onto the semi-conducting and conducting substrates with high surface roughness less than half minute at room temperature by our developed electrosynthetic approach. In addition, we also show that the prepared nanoNi-P is an excellent pre-catalyst not only for OER, but also for the selective oxidative conversion of EG and PET plastics into formic acid. It is the first time that high selectivity (∼100 %) towards of the oxidation of EG and PET have been realized using electrocatalyst solely made of Earth-abundant materials. Finally, efficient and selective generation of hydrogen and formic acid from PEC reforming of PET plastics was demonstrated using an Earth-abundant PEC platform based on nanoNi-P modified TiO2 nanorods photoanode and nanoNi-P based cathode. The production of formic acid and hydrogen from photoelectrocatalytic reforming of EG and PET at ambient conditions sets a sharp contrast to the conventional production of formic acid and hydrogen via an energy-intensive and high-pressure processes involving the use of fossil fuel-based reactants. This work also paves a path for developing artificial leaf for simultaneous environmental mitigation and photosynthesis of renewable fuels and valued chemicals.Nickel-phosphorous alloy nanoparticulate thin film modified electrodes (nanoNi-P) were prepared by electrochemical deposition in a deaerated plating solution containing NiCl2 (0.2 M), NaH2PO2, and NH4Cl (0.25 M) at a specific applied current density (Idep) for 27 s. The electrodeposition was carried out using a CHI 760E potentiostat (CH Instruments, Inc., USA) connected with a conventional three-electrode electrochemical cell with screen-printed carbon electrode (diameter: 3 mm; Zensor R&D, Taiwan) or carbon paper (TPG-H-60, Alfa Aesar) working electrode, Ag/AgCl (sat’d KCl) reference electrode, and Pt foil (1 cm × 4 cm) counter electrode. Prior to the electrodeposition, carbon paper was cleaned sequentially in nitric acid (65 %, Honeywell Fluka™), ethanol (99.5 %, ECHO), and de-ionized water (DIW) under ultrasonication for 10 min. The synthetic parameters, including Idep and the concentration of NaH2PO2 (Chypophosphite), were adjusted to tune the chemical composition, surface morphology, and the overall electrocatalytic acitivity of the resultant nanoNi-P.The nanocomposite of carbon nanotubes with Ni-P alloy nanospheres (CNT/nanoNi-P) was electrodeposited onto the SPCE and carbon paper (exposed surface area: ∼1.0 cm2) in a deaerated plating solution containing NiCl2 (0.2 M), NaH2PO2 (0.04 M), NH4Cl (0.25 M), 2,2'-azino-bis(3-ethylbenzothiazoline-6-sulphonic acid) (0.5 mg mL−1), and multiwall carbon nanotubes (CNTs, 0.25 mg mL−1) at Idep of -20,000 μA cm-2 for 27 s. Note that the plating solution was subjected to ultrasonication at least for 30 min to disperse CNTs prior to the electrodeposition.TiO2 nanorod photoanode, designated as nanoTiO2, was prepared by the direct growing TiO2 nanorods onto the fluorine-doped tin oxide coated glass substrate (FTO) using hydrothermal method. Briefly, FTO substrates were cleaned at 70 °C for 30 min in an aqueous ammonia-hydrogen peroxide solution. Thereafter, the cleaned FTO substrate was placed inclined against the wall of a Teflon liner (volume: 23 mL) with its conducting side facing down. Thereafter, the Teflon liner was filled with a precursor solution containing 15.73 mL HCl (6.0 M) and 0.27 mL titanium (IV) isopropoxide, and was placed into a stainless steel autoclave. Finally, the sealed autoclave was heated in an oven at 150 °C for 3 h. After being rinsed with DIW and dried under N2 purge, the samples were annealed in air at 500 °C for 1 h.nanoNi-P modified nanoTiO2 photoanode, designated as nanoTiO2|nanoNi-P, was prepared by subjecting the prepared nanoTiO2 photoanode (exposed surface area: 6.0 cm2) to the electrodeposition process at Idep of -20,000 μA cm−2 for 1 or 27 s in a deaerated plating solution containing NiCl2 (0.2 M), NaH2PO2 (0.04 M), and NH4Cl (0.25 M).Characterization on surface morphology and film composition of the prepared modified electrodes were performed using characterized using scanning electron microscope (SEM, Hitachi SU-8010) equipped with energy-dispersive X-ray spectroscopy (EDS). High-resolution transmission electron microscopy (HRTEM) and EDS elemental mapping were performed using a JEM-2100F Transmission Electron Microscope (JEOL Ltd., Japan) to analyze the chemical composition, elemental distribution, and structure of CV-activated nanoNi-P particle and nanoTiO2|nanoNi-P, which were detached from electrode substrates under ultrasonication. A Horiba Jobin Yvon JY 2000−2 ICP optical emission spectrometer was used to quantify the amount of nickel species in the nanoNi-P, CNT/nanoNi-P, and nanoTiO2|nanoNi-P electrodes. X-ray photoelectron spectra (XPS) of the nanoNi-P modified electrodes were measured using a PHI 5000 VersaProbe X-ray spectrometer with Al X-ray beam as excitation source. The binding energies (BE) shown in the XPS were corrected by referencing the C 1s peak to 284.6 eV. Ni K-edge X-ray-absorption near-edge structure (XANES) and K-edge extended X-ray-absorption fine-structure (EXAFS) spectra of the nanoNi-P samples were measured at Taiwan Light Source (TLS) beamline 17C1 in National Synchrotron Radiation Research Center (NSRRC). Soft X-ray absorption spectra (sXAS) were recorded at TLS beamline 20A1 in NSRRC. Ni L-edge and O K-edge absorption spectra of the samples were measured in both total-electron-yield (TEY) and total-fluorescence-yield (TFY) modes. Raman spectra of nanoNi-P samples were measured using a Raman spectrometer (Protrustech Corporation Ltd.) with 532 nm laser with a light power of 300 mW.The activity of the nanoNi-P and CNT/nanoNi-P modified electrodes towards HER were analyzed utilizing a potentiostat (CHI 760, CH Instruments, Inc., USA) connected with a nitrogen-purged two-compartment, separated with an anion exchange membrane (AEM-025, Johnson Matthey, UK), three-electrode electrochemical cell with Hg/HgO (1 M NaOH) reference electrode and Pt foil (1 cm × 4 cm) counter electrode. IR drop was compensated for all the electrochemical experiments, and all potentials are referenced to the reversible hydrogen electrode (RHE) with Eq. (1): (1)   E V v s . RHE = E V v s . Hg / HgO + 0.140 + 0.059 × pH The electrochemically active surface area (ECSA) of the prepared nanoNi-P modified electrodes was calculated by firstly determining the double-layer capacitance (Cdl) utilizing cyclic voltammetry, followed by dividing the determined Cdl value with Cdl of pure nickel thin film (∼16 μF cm−2) [32].Two kinds of turnover frequencies (TOF) were used to evaluate the activity of nanoNi-P modified electrodes, including one based on the loading amount of nickel species (TOFNi) and the other one based on surface sites (TOFECSA). TOFNi, determined by using Eq. (2), was used to evaluate the overall activity, whereas TOFECSA, determined by using Eq. (3), was used to access the intrinsic activity [22,33]. (2) TO F Ni = number of hydrogen turnover per geometric surface area number of loaded nickel atom per geometric surface (3) TO F ECSA = number of hydrogen turnover per geometric surface area number of surface site per real surface area The number of loaded nickel atom per geometric surface area (NNi) was determined by ICP-OES, whereas the number of hydrogen turnover per geometric surface area (#H2), corresponding to specific current density, and the number of the surface site per real surface area (#surface site) were calculated by using Eqs. (4) and (5), respectively: (4) # H 2 = ( j mA c m 2 ) × ( 1 C s -1 1000 mA ) × ( 1 mol e − 96485 C ) × ( 1 mol H 2 2 mol e − ) × ( 6 .022 × 1 0 23 H 2 molecules 1 mol H 2 ) = 3.12 × 10 15 H 2 /s c m 2 per mA c m 2 (5) # surface site = 4 atoms/unit cell 43.76 × 10 − 24 c m 3 /unit cell 2 3 = 2.029 × 10 15 atoms c m real -2 ( metallic nickel ) Subsequently, a plot of current density (j) vs. overpotential (η) can be converted into that of TOF vs. η by Eqs. (6) and (7): (6) TO F Ni = ( 3.12 × 10 15 H 2 /s c m 2 per mA c m 2 ) × | j | N Ni (7) TO F ECSA = ( 3.12 × 10 15 H 2 /s c m 2 per mA c m 2 ) × | j | # surface site × A ECSA Note that as the exact hydrogen binding site is unknown, the number of the surface atom was used as #surface site in the TOFECSA calculation instead. Besides, as the number of the surface atom for Ni ( 2.029 × 10 15 atoms c m real -2 ), Ni2P ( 2.001 × 10 15 atoms c m real -2 ), and Ni3P ( 2.023 × 10 15 atoms c m real -2 ) are similar, and considering the low P content of the nanoNi-P prepared in this work, we used the number of the surface atom for Ni, i.e., 2.029 × 10 15 atoms c m real -2 , in the TOFECSA calculation.As the nanoNi-P modified electrode, prepared with Idep= -20000 μA cm−2 and Chypophosphite = 0.02 M, exhibited the best overall HER activity (vide infra), it was selected as the pre-catalyst for OER and electrochemical reforming of EG and PET. In addition, prior to its application for the reactions as above-mentioned, it was activated by cyclic voltammetry (CV) between 1.2–1.65 V vs. RHE at a scan rate of 50 mV s-1 for 225 cycles. Catalytic properties of the CV-activated nanoNi-P modified electrodes towards OER and electrochemical reforming of EG and PET were performed utilizing a CHI 760 potentiostat (CH Instruments, Inc., USA) connected with a nitrogen-purged three-electrode electrochemical cell consisting of Hg/HgO (1 M NaOH) reference electrode and Pt foil (1 cm × 4 cm) counter electrode. PET lysate for PET reforming was prepared by soaking 4.0 g of PET flakes, obtained from cutting commercially available PET bottles, in 100 mL KOH solution (2.0 M) in a sealed vial at 80 °C for 24 h, and subsequent diluting the resultant solution with an equal amount of DIW. The prepared PET lysate was used for the reforming process directly without further purification or filtration. The amount of PET dissolved in the PET lysate, determined by subtracting the solid content in the PET lysate from the amount of PET flakes added, was found to be 0.55 ± 0.03 g, corresponding to the solubility of 2.73 ± 0.16 g L-1. As the exact molecular weight of PET in commercially available PET bottle is unknown, the concentration of PET repeating unit (C10H8O4) in PET lysate (1.42 ± 0.08 mM), determining by dividing the solubility of PET in PET lysate with the molecular weight of PET repeating unit (C10H8O4), was used as the basis for the calculation of conversion rate in the (photo-)electrochemical PET reforming.Characterization on the PEC properties of nanoTiO2 and nanoTiO2|nanoNi-P were carried out utilizing a MultiPalmSens4 potentiostat (PalmSens BV, Netherlands) in a home-made electrochemical cell containing deaerated KOH solution (1.0 M, pH 14.0) under light irradiation (AM 1.5 G 100 mW cm−2) using an XES-40S2-CE solar light simulator (SAN-EI Electric). Prior to the photoelectrochemical measurements, nanoTiO2|nanoNi-P was activated by potential cycling at 50 mV s−1 between 1.2–1.65 V vs. RHE for 225 cycles.Electrochemical impedance spectroscopy (EIS) analyses of nanoTiO2 and nanoTiO2|nanoNi-P were carried out in the PET lysate under light illumination using MultiPalmSens4 potentiostat equipped with EIS spectrum analyser. The applied potential, AC amplitude, and frequency range were open-circuit potential, 10 mV, and 1 MHz to 0.1 Hz, respectively. The obtained EIS data were fitted with the equivalent circuit [34] using built-in ZView® software to retrieve the parameters associated with the kinetics of interfacial charge transfer.Quantification of formate generated from the (photo-)electrochemical reforming of EG and PET was performed using a 883 Basic IC plus ion chromatography (Metrohm) equipped with Metrosep Organic Acids Guard (4.6 × 50 mm) and Metrosep Organic Acids column (7.8 × 250 mm). Note that as a small amount of formate (0.67 ± 0.04 μmole) was found in the PET lysate, presumably resulted from the decomposition of impurity in the commercially available PET bottle, prior to the reforming experiments, the difference in the amount of formate in PET lysate before and after reforming experiments was reported as the actual amount of formate generated from (photo-)electrochemical PET reforming.Quantification of hydrogen generated from PEC reforming of PET plastics was carried out by headspace gas analysis using an Agilent 7890A Series gas chromatography equipped with a thermal conductivity detector. The GC oven holding the 5 Å molecular sieve column was set at 40 °C.The nanoNi-P modified electrodes were prepared by electrochemical deposition under various applied current densities (Idep) for 27 s (see Experimental Section for the details). The composition and the amount of nickel species in the prepared nanoNi-P modified electrode were summarized in Table 1 . Fig. 1 shows the XPS spectra of the nanoNi-P modified electrodes prepared with different Idep. Features in XPS spectra confirm the formation of Ni-P alloy regardless of Idep, including (i) peaks in Ni 2p3/2 region at binding energy (BE) of 851.8 and 853.2 eV corresponding to metallic nickel and Niδ+, respectively, in nickel phosphide [35,36], and (ii) peaks in P 2p region at BEs of 129.1 and 130.0 eV corresponding to Pδ− in nickel phosphide [35,36]. The existence of small peaks at BEs of 855.8, 132.8, and 530.8 eV is indicative of the presence of NiOx and POx species, presumably formed by the oxidation of surface NiPx upon exposure to air [37], on the electrode surface. The analyses on the composition depth profile of all the prepared samples (Figure S1 and Figure S2) indicate that P atoms were well-dispersed in metallic nickel matrix, and the P content of the prepared nanoNi-P modified electrodes decreased as Idep was increased. Figure S3 shows the chronopotentiograms recorded during the electrochemical preparation of the nanoNi-P modified electrodes. As revealed, the required potential shifted to more negative potential as Idep was increased. The negative shift in the deposition potential promoted the hydrogen evolution reaction (HER), which consequently decreased the Faraday efficiency for the deposition of nanoNi-P (Table 1). In addition, in contrast to the electrochemical deposition of nickel (Eq. 8), the electrochemical reduction of hypophosphite ions into P° (Eq. 9) requires the transfer of protons. As a result, the depletion of proton in the vicinity of electrode surface due to HER discouraged the electrochemical reduction of P, resulting in lower P content of the nanoNi-P modified electrode prepared at higher Idep (Figures S1-S2 and Table 1). It is interesting to note that the amount of nickel in the nanoNi-P modified electrode prepared with lower Idep exceeded the theoretical value based on Faraday’s law. This additional amount of metallic nickel was deposited most likely via electroless deposition as the reduction of Ni2+ ions to metallic nickel by hypophosphite (Eqs. (10) and (11)) is thermodynamically favourable [38]. Although the deposition rate of electroless deposition is low at room temperature [39], the relative long electrodeposition duration (Figure S3) required for the preparation of the nanoNi-P electrode at Idep = -12.5 μA cm-2 allowed the noticeable amount of deposits. (8) N i 2+ + 2 e − → Ni E 0 = − 0.25 V v s . N H E (9) H 2 P O 2 − + 2 H + + e − → P + 2 H 2 O E 0 = − 0.248 V v s . N H E (10) N i 2+ + H 2 P O 2 − + H 2 O → Ni + H 2 P O 3 − + 2 H + E r e a c t i o n 0 = 0 .254 V (11) 2N i 2+ + H 2 P O 2 − + 2 H 2 O → 2 Ni + H 2 P O 4 − + 4 H + E r e a c t i o n 0 = 0 .514 V Figure S4 shows the SEM images of the nanoNi-P modified electrodes prepared at different Idep. As revealed, all the nanoNi-P modified electrodes consisted of the aggregates of particles with the decreasing size with increasing Idep. For example, nanoNi-P, prepared with Idep of -12.5 μA cm−2, comprises the aggregates of particles with a size of 0.5∼ 1.0 μm, but that, prepared with Idep of -200,000 μA cm−2, consists of the aggregates of nanoparticles with a size of ∼200 nm. The decrease in particle size at higher Idep could be attributed to the competing HER, as the hydrogen bubbling on the as-deposited Ni-P nuclei would limit the access of the plating ions and discourage the growth of Ni-P particles. The analyses on the electrochemically available surface area (ECSA), shown in Table 1 and Figure S5, reveal that the nanoNi-P modified electrode, prepared with higher Idep, exhibited higher ECSA, which is in agreement with SEM results (Figure S4) that the decrease in nanoNi-P particle size increased the surface area.The prepared nanoNi-P modified electrodes were subjected to the electrochemical characterizations on their HER activity, and the results are shown in Fig. 2 . The HER activity indexes, including exchange current density (i0), Tafel slope, the required overpotential (η) to achieve a geometric current density (Jgeo) of -10 mA cm−2, and TOFNi and TOFECSA at η= −200 mV, were included in Table 1. As revealed, the nanoNi-P modified electrode, prepared with Idep of -20,000 μA cm−2, exhibited the best HER activity. In terms of overall activity based on TOFNi, (Fig. 2 b and Table 1), it exhibited a TOFNi of 187.2 ± 7.2 h-1 at η= −200 mV, which is about 1.3, 3.7, 36.1, and 80.0 times higher than those prepared with Idep of -200,000, -2,000, -200, and -12.5 μA cm−2, respectively. In terms of intrinsic activity based on the electrochemically available surface sites, i.e., TOFECSA, (Fig. 2 c and Table 1), it exhibited a TOFECSA of 7.40 ± 0.29 s-1 at η= −200 mV, which is about 2.0, 3.0, 26.6, and 23.7 times higher than those prepared with Idep of -200,000, -2,000, -200, and -12.5 μA cm−2, respectively. The above results suggest that the superior electrocatalytic activity of nanoNi-P, prepared with Idep of -20,000 μA cm−2, was not simply resulted from the differences in ECSA or loading amount of nickel species. Tafel analyses (Fig. 2d) show that nanoNi-P, prepared with higher Idep, exhibited lower Tafel slope, which indicates that Idep had significant influences on the HER mechanism. Figure S6 shows the X-ray-absorption near-edge spectra (XANES) of the Ni K edge. As indicated, the Ni K edge position of nanoNi-P is slightly more positive than that of nickel foil, which is in line with the XPS results that Ni-P alloy formation induced an electron transfer from Ni to P to form Niδ+ species. The presence of negatively-charged P atoms (Pδ−) is beneficial as they can act as proton-acceptor sites to facilitate HER [28]. The bonding environment of nickel atoms was also studied by Fourier-transformed extended X-ray-absorption fine-structure (EXAFS) at Ni edge, and results along with the fitted parameters are shown in Fig. 3 and Table S1, respectively. The nanoNi-P modified electrodes, prepared with smaller Idep, was found to exhibit a higher coordination number in Ni-P bonding (CNNi-P) but a lower coordination number in Ni-Ni bonding (CNNi-Ni). The increasing CNNi-P/CNNi-Ni ratio by decreasing Idep indicates that the incorporation of more P atoms in the nickel crystal lattice decreased the concentration of isolated nickel sites (hydride-acceptor sites) and discouraged the first electron transfer to water molecules (Volmer step) during HER [40], resulting in higher Tafel slope and decreased activity (Fig. 2). In contrast, the nanoNi-P modified electrodes, prepared with significantly high Idep (i.e., -200,000 μA cm−2), exhibited zero CNNi-P, which suggests insufficient proton-acceptor sites available for HER, resulting in lower intrinsic activity than that prepared with Idep = -20,000 μA cm-2. These results are in agreement with the previous report that the highest HER activity of Ni-P catalyst relied on the coexistence and collaboration of hydride-acceptor sites (isolated nickel atoms) and proton-acceptor sites (negatively-charged P atoms) [28].The above findings indicate that Idep is crucial in determining the P content and bonding environment of the nanoNi-P modified electrodes. To further investigate the effect of P content on the HER activity, we prepared nanoNi-P modified electrodes with optimized Idep (i.e., -20,000 μA cm−2) in a plating solution containing hypophosphite of various concentrations, and the film composition along with the amount of nickel species in the prepared nanoNi-P modified electrodes are summarized in Table 2 . P content of the prepared modified electrodeswas found to increase with increasing hypophosphite concentration (Chypophosphite) in the plating solution, suggesting that the P content of the nanoNi-P modified electrode can be controlled simply by tuning Chypophosphite. The SEM characterizations (Figure S7) reveal that the nanoNi-P modified electrode, prepared without hypophosphite, consists of aggregates of particles with irregular shape and smooth surface. In contrast, those prepared with hypophosphite exhibited aggregates of particles with spherical shape and rough surface. These difference in surface morphology resulted in the variation of ECSA (Figure S8). Fig. 4 shows the electrochemical characterization on the HER activity of the nanoNi-P modified electrodes prepared with various Chypophosphite. The HER activity indexes, including i0, Tafel slope, η required to reach Jgeo of -10 mA cm−2, and TOFNi and TOFECSA at η= −200 mV, were included in Table 2. As revealed from Fig. 4a-c and Table 2, there was an optimal P content to achieve the highest HER activity. In terms of the overall activity, the nanoNi-P modified electrode, prepared with Chypophosphite = 0.04 M, exhibited the highest TOFNi of 216.0 ± 3.6 h-1 at η= −200 mV. Nonetheless, in terms of the intrinsic activity, the nanoNi-P modified electrode, prepared with Chypophosphite = 0.2 M, exhibited the highest TOFECSA of 7.40 ± 0.29 s-1 at η= −200 mV. The analyses of XANES (Figure S9) and EXAFS (Figures S10 and Table S2) spectra reveal that the nanoNi-P modified electrode prepared with Chypophosphite = 0.04 M exhibited no Ni-P bonding. The lack of sufficient proton-acceptor sites for HER would therefore be the main cause resulting in lower intrinsic activity than that prepared with Chypophosphite = 0.2 M. Tafel analyses (Fig. 4d) show that the nanoNi-P modified electrode, prepared in the presence of hypophosphite, exhibited lower Tafel slope than that prepared in the absence of hypophosphite, which suggests P played a role in determining the mechanism of HER, and the incorporation of suitable amount of P atoms as the proton-acceptor sites can facilitate the HER kinetics. Since the nanoNi-P modified electrode, prepared with Idep= -20,000 μA cm−2 and Chypophosphite = 0.04 M, exhibited the highest overall HER activity, it was designated as nanoNi-Pop and selected for further investigation.To further improve the overall HER activity, we prepared carbon nanotube-supported Ni-P alloy nanospheres (CNT/nanoNi-Pop) modified electrodes by the electrodeposition of nanoNi-P in the presence of CNTs under optimal electrodeposition condition (i.e., Idep= -20,000 μA cm−2 and Chypophosphite = 0.04 M). As revealed in Fig. 5 a, the CNT/nanoNi-Pop modified electrode consisted of Ni-P nanospheres with a size of ∼150 nm embedded in between entangled CNTs. Characterization of electrocatalytic activity (Fig. 5b-c) indicates that the CNT/nanoNi-P modified electrode exhibited significantly higher HER activity than nanoNi-Pop, though the loading amount of nickel species are similar (2.62 ± 0.02 vs. 2.64 ± 0.02 μmole cm-2). Notably, the CNT/nanoNi-Pop modified electrode required η of -149.7 ± 6.1 and -206.3 ± 9.2 mV to reach jgeo of -10 and -100 mA cm-2, respectively, which were about 20 mV smaller than nanoNi-Pop. Furthermore, at η= −200 mV, the CNT/nanoNi-Pop modified electrode exhibited a TOFNi value of 431.2 ± 59.6 h-1 that is about two times higher than nanoNi-Pop. Finally, as revealed in Fig. 5 d and Figure S11, the CNT/nanoNi-Pop modified electrode also exhibited high stability at both Jgeo of -10 and -100 mA cm-2. The excellent activity and stability of the CNT/nanoNi-P modified electrode place itself among the most active earth-abundant HER catalysts in alkaline aqueous media (Table S3).Encouraged by its excellent HER activity, the applications of the nanoNi-Pop modified electrode for OER and electrochemical oxidation of EG were further investigated. In addition, the nanoNi-P modified electrode prepared without hypophosphite, designated as nanoNi for simplicity, was also included for comparison. Moreover, prior to the application for the OER and electrochemical reforming of EG, both modified electrodes were subjected to a CV-activation process (see Experimental Section for the details), and the CVs recorded during the activation process are shown in Figure S12. As revealed, the charge under the redox peaks remained constant during the last 10 cycles of the CV-activation, suggesting that the growth of oxy-hydroxide layer on the surfaces of nanoNi-Pop and nanoNi completed after the CV-activation process. Fig. 6 a-b show TEY sXAS of nanoNi-Pop and nanoNi before and after CV-activation. Additional signals at 854.2 eV in the Ni L3-edge region (Fig. 6a), reflecting the formation of Ni3+ [41,42], and at 528.3 eV in O K-edge region (Fig. 6b), resulted from the hybridization of O(2p)-Ni(3d)eg [41,42], were observed after CV-activation, which confirms that NiOOH species formed on both nanoNi-Pop and nanoNi after CV-activation process. Besides, the intensity ratio of the double-peak features at 854.2 and 852.1 eV (I854.2/I852.1) in Ni L3-edge region, an index of the oxidation state of the nickel atoms [41], was found to be influenced by P incorporation. nanoNi after CV-activation, designated as nanoNi(CV) for simplicity, exhibited a higher I854.2/I852.1 value (0.37 vs. 0.31) than nanoNi-Pop after CV-activation, designated as nanoNi-Pop(CV), which implies that nickel species in nanoNi(CV) had higher average oxidation state than those in nanoNi-Pop(CV). Nonetheless, analysis of TFY sXAS spectra of NiL-edge (Fig. 6c) reveals that no change in peak position nor the appearance of any new peak after CV-activation for both samples. As TFY is a bulk-sensitive probe and has higher probing depth than TEY mode [43,44], the observed different features in TFY and TEY modes suggest that CV-activation only induced a very thin oxyhydroxide layer on the surface of both nanoNi-Pop(CV) and nanoNi(CV). For example, as revealed from the TEM analyses (Figure S13), the thickness of oxyhydroxide layer formed on nanoNi-Pop(CV) was <10 nm. Fig. 6d shows the CVs of the nanoNi-Pop(CV) and nanoNi(CV) electrodes in KOH (1.0 M) solution. As revealed, both nanoNi-Pop(CV) and nanoNi(CV) electrodes exhibited CV features characteristic to the redox reaction of Ni(OH)2/NiOOH (Eq. (12)): (12) Ni OH 2 + O H − ⇌ NiOOH + e − + H 2 O (13) NiOOH + C H 2 OHC H 2 OH ⇌ Ni OH 2 + product In addition, both nanoNi-Pop(CV) and nanoNi(CV) exhibited multiple cathodic peaks during the reverse scan, which could be ascribed to the various routes of reductive transformation of NiOOH to Ni(OH)2 [45,46]. Due to the difference in the energetics of the reductive transformation of β-NiOOH and γ-NiOOH, these reductive transformations occurred at different potentials, and cathodic peaks r1, r2, and r3 are assigned to the transformations of β-NiOOH to β-Ni(OH)2, γ-NiOOH to β-Ni(OH)2, and γ-NiOOH to α-Ni(OH)2, respectively [45,46]. In other words, the multiple cathodic peaks observed in Fig. 6d indicate that β-NiOOH and γ-NiOOH coexisted in both nanoNi-Pop(CV) and nanoNi(CV). The molar ratio of β-NiOOH to γ-NiOOH, determined by dividing the charge under the cathodic wave r1 to that under cathodic waves r2, and r3 (Figure S14), was found to be 2.74 and 1.27 for nanoNi-Pop(CV) and nanoNi(CV), respectively. This finding suggests that the β-NiOOH content of nanoNi-Pop(CV) was higher than that of nanoNi(CV), and P played a role in suppressing further transformation of β-NiOOH into γ-NiOOH. Additional analyses on the formation of oxyhydroxide on the nanoNi-Pop(CV) and nanoNi(CV) electrodes using Raman spectroscopy were attempted, but the results (Figure S15) didn’t show any features characteristics to these oxyhydroxides, which can be attributed to the fact that the oxyhydroxide layer is too thin (Figure S13) to be observed by Raman spectroscopy. Figures S16 shows electrochemical characterizations on the OER activity of both nanoNi-Pop(CV) and nanoNi(CV) electrodes in KOH solution (1.0 M). It can be found that nanoNi-Pop(CV) exhibited significant higher activity, in terms of η, than nanoNi(CV). Notably, the nanoNi-Pop(CV) modified electrode required η of ∼400 mV to achieve jgeo of -100 mA cm−2, which is ∼ 60 mV smaller than nanoNi(CV) (Figures S16a-b). The Tafel analyses (Figure S16c) reveal that nanoNi-Pop(CV) exhibited a significantly lower Tafel slope than nanoNi(CV) (44 vs. 61 mV dec-1), which implies that the mechanism of OER was different at these two catalysts, and OER proceeded in a more feasible fashion at nanoNi-Pop(CV). Fig. 7 shows the electrochemical characterizations on the activity of both nanoNi-Pop(CV) and nanoNi(CV) electrodes towards electrochemical oxidation of EG in KOH solution (1.0 M). In the absence of EG, both nanoNi-Pop(CV) and nanoNi(CV) electrodes exhibited CV features characteristic to Ni(OH)2/NiOOH redox reactions. Besides, upon addition of EG, both nanoNi-Pop(CV) and nanoNi(CV) electrodes exhibited additional features, including a notable enhancement in the peak current of o1 wave and the decrement in the peak current of r1 wave. These changes in CV features indicate that the oxidation of EG at both nanoNi-Pop(CV) and nanoNi(CV) electrodes involved electrochemical formation of active NiOOH species (Eq. (12)) and a follow-up reaction between EG and NiOOH (Eq. (13)). Moreover, as compared with the extent of decrement in the cathodic current of r2 and r3 peaks in the presence of EG (Fig. 7a and b), the significantly pronounced decrement in the cathodic current of r1 peak in the presence of EG was noticed. As discussed in the previous section, the cathodic peak r1 is resulted from the reductive transformation of β-NiOOH to β-Ni(OH)2, and the pronounced decrease in the current of peak r1 in the presence of EG suggests that β-NiOOH would be the active species responsible for the electrocatalysis of EG. Furthermore, nanoNi-Pop(CV) exhibited higher catalytic current responses to EG than nanoNi(CV). Notably, nanoNi-Pop(CV) exhibited a Jgeo of 38.3 mA cm−2 at 1.5 V vs. RHE in EG solution (0.1 M), which is about 1.6 times higher than nanoNi(CV) (∼24.6 mA cm−2). The higher electrocatalytic activity of nanoNi-Pop(CV) than nanoNi(CV) would be attributed to the higher amount of active β-NiOOH sites available for EG oxidation. This result suggests that P played a beneficial role in reserving active β-NiOOH species for electrochemical EG oxidation. Product analyses after 1-h electrolyses at 1.5 V vs. RHE (Fig. 7 c and d) reveal that the electrocatalytic EG oxidation at both nanoNi-Pop(CV) and nanoNi(CV) electrodes mainly generated formate (Faradic efficiency: ∼100 %), which suggests that EG oxidation at both electrodes not only involves an electrocatalytic scheme consisting of the electrochemical generation of active β-NiOOH species (Eq. (12)) and its ensuing redox reactions with ethylene glycol and intermediates, but also involve the CC bond cleavage to generate formate as the sole product (see Scheme 1 ) [47,48]. Besides, the rate of formate generation (Rformate) at nanoNi-Pop(CV) was about 1.3 times higher than nanoNi(CV) (244.6 ± 15.0 μmole cm−2 h−1 vs. 182.9 ± 26.0 μmole cm−2 h−1). These results suggest that β-NiOOH content affected the kinetics of the electrochemical EG oxidation, but didn’t exhibit any effect on the reaction pathway of EG oxidation. To the best of our knowledge, it is the first time that high conversion of EG into formate with high selectivity has been achieved using precious metal-free electrocatalysts. Note that as Rformate at nanoNi-Pop(CV) was much higher than that at nanoNi(CV) electrode, the pH drop, due to the dissociation of formic acid, nearby the surface of nanoNi-Pop(CV) electrode was therefore more pronounced. The pronounced pH drop would result in the positive shift in the redox potential of Ni2+/Ni3+ redox couple, and decrease in the overpotential available for the electrochemical EG oxidation during the controlled-potential electrolysis [15,49], which in turn gradually reduces the catalytic current and thus causes curving of charge transient of nanoNi-Pop(CV) electrode (line ii in Fig. 7c).Further application of nanoNi-Pop(CV) towards photoelectrochemical (PEC) EG oxidation was explored by integrating nanoNi-Pop(CV) onto the TiO2 nanorods (nanoTiO2) photoanode. Initial attempt was performed by electrodepositing nanoNi-Pop onto nanoTiO2 photoanode with charge passage (C) of 0.54 C cm−2, and subsequently CV-activating the resultant electrode (see Experimental Section for the details). The results of SEM (Figure S17) and XRD (Figure S18) analyses confirm the prepared photoanode, designated as nanoTiO2|nanoNi-Pop(CV) (C = 0.54 C cm−2), consisted of Ni-P submicron-sized spheres decorated rultile TiO2 nanorods. However, as revealed in Figure S19 and Fig. 8 g, the nanoTiO2|nanoNi-Pop(CV) (C = 0.54 C cm−2) photoanode exhibited inferior PEC performance than the bare nanoTiO2 photoanode. The poor performance can be attributed to the large size of nanoNi-Pop(CV), which short-circuits photo-induced charges between the neighbouring TiO2 nanorods and disables the hole transfer to EG. To improve the PEC performance of the nanoNi-Pop(CV) modified nanoTiO2 photoanode by reducing the size of nanoNi-Pop(CV), we reduced the duration of electrodeposition process from 27 s to 1 s, which means that only 0.02 C cm−2 charge passed for the deposition of nanoNi-Pop. The SEM and EDS analyses (Figures S20) reveal that nanoTiO2|nanoNi-Pop(CV) (C = 0.02 C cm−2) consisted of nanoNi-Pop(CV) nanoparticles with size of 10∼20 nm uniformly decorated onto nanoTiO2. Fig. 8a-f show the analyses of HR-TEM and EDS elemental mapping for nanoTiO2|nanoNi-Pop(CV) (C = 0.02 C cm−2). The clear lattice fringes with a lattice spacing of 0.206 nm (Fig. 8 b) corresponds to the (011) lattice plane of metallic nickel (PDF no. 45-1027). The SAED pattern, shown in the inset of Fig. 8b, consists of the bright spots and weak diffuse ring. The bright spots indicate that the prepared rutile TiO2 nanorod is single crystal, whereas the diffuse ring is indexed to the (011) plane of low-crystalline nickel. The results of EDS elemental mapping (Fig. 8c-f) suggest that the surface of nanoTiO2 is uniformly covered with nanoNi-Pop(CV). Fig. 8g-h show the PEC characterizations of the bare nanoTiO2 and nanoTiO2|nanoNi-Pop(CV) (C = 0.02 C cm−2) photoanodes. As revealed, nanoTiO2|nanoNi-Pop(CV) (C = 0.02 C cm−2) exhibited superior PEC performance than bare nanoTiO2. To begin with, nanoTiO2|nanoNi-Pop(CV) (C = 0.02 C cm−2) exhibited a lower photocurrent onset than bare nanoTiO2 (0.24 vs. 0.26 V vs. RHE). In addition, the photocurrent response of nanoTiO2|nanoNi-Pop(CV) (C = 0.02 C cm−2) was found to be about 5 times higher than bare nanoTiO2 (∼0.23 mA cm−2 vs. ∼46.0 μA cm−2) at 0.4 V vs. RHE. Finally, nanoTiO2|nanoNi-Pop(CV) (C = 0.02 C cm−2) exhibited Rformate of 2.1 ± 0.2 μmole cm−2 h−1 with Faradaic efficiency (FEformate) of 57.2 ± 3.1 % at 0.5 V vs. RHE, whereas bare nanoTiO2 only showed Rformate of 0.3 ± 0.1 μmole cm−2 h−1 and FEformate of 27.0 ± 4.8 % at the same applied potential. This remarkable enhancement in PEC performance by surface modification of nanoNi-Pop(CV) further confirms the crucial role of nanoNi-Pop(CV) in catalyzing EG oxidation. Note that PEC EG oxidation by nanoTiO2|nanoNi-Pop(CV) (C = 0.02 C cm−2) resulted in lower FEformatethan electrochemical EG reforming by nanoNi-Pop(CV), which could be attributed to the fact that the surface of nanoTiO2 is not fully covered with nanoNi-Pop(CV) due to the small loading amount of nanoNi-Pop(CV) (95.0 ± 3.0 nmole cm−2), and the direct hole transfer from nanoNi-Pop(CV)-free TiO2 surface to EG is possible, resulting in the lower FEformate.As the EG is one of the monomers constituting PET, the direct application of nanoNi-Pop(CV) to the reforming of PET was also attempted and investigated. Figures S21 shows electrochemical characterizations on the activity of the nanoNi-Pop(CV) electrode towards PET reforming in KOH solution (1.0 M). It can be found from Figure S21a that nanoNi-Pop(CV) exhibited higher catalytic current in PET lysate than in KOH solution, indicating nanoNi-Pop(CV) had promising activity towards electrochemical PET reforming. Besides, product analyses after 2-h controlled-potential electrolyses indicate that nanoNi-Pop(CV) can also selectively reform PET into formate with Rformate of 50.7 ± 6.7 μmole cm−2 h−1 and FEformate of 103.9 ± 4.7 %. The PET conversion rate, based on the amount of dissolved PET repeating unit (C10H8O4), was found to be 16.8 ± 2.2 %. Note that nanoNi-Pop(CV) exhibited negligible activity in electrocatalysing the oxidation of terephthalic acid (TA) and exhibited high selectivity (FEformate: 95.7 ± 2.7 %) towards the generation of formate from the electrochemical EG oxidation in the presence of TA (Figure S22a). These findings not only indicate the oxidation of terephthalate unit of PET was unlikely involved in the PET reforming, but also suggest the PET reforming at nanoNi-Pop(CV) would start with the electrocatalytic oxidation of EG unit at β-NiOOH active sites and follow-up CC bond cleavage to release formate. The high activity of nanoNi-Pop(CV) towards PET reforming would be therefore attributed to the beneficial role of P in the reserving high amount of active β-NiOOH species. It is also important to note that the presence of TA slightly reduced the activity of nanoNi-Pop(CV) towards the electrocatalytic EG oxidation. For example, Rformate from the electrocatalytic oxidation of EG (0.1 M) in the presence of TA (0.1 M) was 203.3 ± 26.0 μmole cm−2 h−1, about 1.2 times lower than that obtained in the absence of TA. The apparent decrease in the activity can be attributed to the fact that the presence of TA (0.1 M) induced a drop in bulk pH (from 14.01 to 13.90), which in turn shifts in the redox potential of Ni2+/Ni3+ redox couple to the positive side (Figure S22b), and decreases the overpotential available for the electrocatalytic EG oxidation during the controlled-potential electrolysis. This result also suggests that the release of TA during the PET reforming would be one of the main reasons causing the gradual decrease in the catalytic current during the controlled-potential electrolysis of PET (Figure S21b). On the other hand, due to the limited solubility of PET (2.73 ± 0.16 g L−1), PET concentration available for reforming is very low (i.e., in the order of μM). During PET reforming, the thickness of the diffusion layer nearby electrode surface would increase significantly with this low PET concentration, resulting in the fast decrease in the concentration gradient and thus contributing to the loss in catalytic current.Encouraged by the promising activity of nanoNi-Pop(CV) and CNT/nanoNi-Pop towards PET reforming and HER, respectively, we subsequently examined their applicability towards PEC PET reforming. For this purpose, a nanoTiO2|nanoNi-Pop(CV) (C = 0.02 C cm−2)//CNT-nanoNi-Pop two-electrode PEC device, consisting of nanoTiO2|nanoNi-Pop(CV) (C = 0.02 C cm−2) photoanode and CNT/nanoNi-Pop cathode, was established, and its PEC performance was characterized and shown in Fig. 9 b. For comparison, PEC characterization of another two-electrode PEC device based on nanoTiO2 photoanode and Pt foil, designated as nanoTiO2//Pt, was also included in Fig. 9 b. Note that to prepare cathode with the same geometry surface area to that of nanoTiO2|nanoNi-Pop(CV) photoanode, CNT/nanoNi-Pop cathode was prepared by electrodepositing CNT/nanoNi-Pop onto the carbon paper instead of SPCE (see Experimental Section for the details). The SEM results (Figure S23a) reveal that the surface morphology of CNT/nanoNi-Pop deposited on the carbon paper was similar to that deposited on the SPCE substrate (Fig. 5a), suggesting that the proposed deposition methodology is suitable for both flat substrate and substrate with 3D porous structure. Electrochemical analysis, shown in Figures S23b-c and Fig. 9a, reveals that CNT/nanoNi-Pop modified carbon paper required an additional η = 40∼60 mV to achieve the same Jgeo with Pt foil in the current density range between -0.1 to -70 mA cm−2, but showed comparable activity at Jgeo ≥ -90 mA cm−2. Particularly, CNT/nanoNi-Pop modified carbon paper required η of ∼ −180 mV to achieve a Jgeo= -100 mA cm−2, which is about 30 mV lower than Pt foil. In addition, as compared with bare nanoTiO2, nanoTiO2|nanoNi-Pop(CV) (C = 0.02 C cm−2) exhibited enhanced photoelectrocatalytic activity towards PET reforming. For example, nanoTiO2|nanoNi-Pop(CV) (C = 0.02 C cm−2) showed a photocurrent density of ∼310 μA cm−2, which is about 1.6 times higher than bare nanoTiO2. The EIS analyses, shown in Table S4 and Figure S24, indicate that nanoTiO2|nanoNi-Pop(CV) (C = 0.02 C cm−2) exhibited smaller interfacial charge transfer resistance (709.3 vs. 1104.0 Ω) than nanoTiO2. Consequently, the enhancement in photocurrent can be mainly attributed to the improved kinetics of interfacial charge transfer by nanoNi-Pop(CV). In typical 4-h PEC PET reforming at an external bias of 0.5 V, nanoTiO2|nanoNi-Pop(CV) (C = 0.02 C cm−2)//CNT-nanoNi-Pop generated 6.4 ± 1.4 μmol formate (FEformate: 57.1 ± 1.7 %) and 12.5 ± 1.1 μmol H2(FEhydrogen: 76.8 ± 7.8 %), whereas nanoTiO2//Pt produced 1.0 ± 0.1 μmol formate (FEformate: 14.7 ± 2.5 %) and 7.1 ± 2.4 μmol H2(FEhydrogen: 65.7 ± 11.7 %). The PET conversion rate, based on the amount of dissolved PET repeating unit (C10H8O4), for nanoTiO2|nanoNi-Pop(CV) (C = 0.02 C cm−2)//CNT-nanoNi-Pop and nanoTiO2//Pt were found to be 14.9 ± 3.3 and 2.4 ± 0.2 %, respectively. The significantly enhanced PET conversion and product generation by nanoTiO2|nanoNi-Pop(CV) (C = 0.02 C cm−2)//CNT-nanoNi-Pop indicate that the overall PEC performance was limited by the oxidative conversion of PET at nanoTiO2|nanoNi-Pop(CV) (C = 0.02 C cm−2) photoanode, which further confirms the crucial role of nanoNi-Pop(CV). Note that a gradual decrease in photocurrent for both devices (Fig. 9b) was observed during the photoelectrochemical PET reforming. The apparent loss in photocurrent could be attributed to the (i) the pH drop induced by the release and dissociation of formic acid and TA during the reforming of PET, and (ii) decrease in the PET concentration gradient by the increase in the thickness of diffusion layer. To stabilize the photocurrent by minimizing the accumulation of acids and stabilizing the thickness of diffusion layer, the development of flow-type photoelectrochemical system for PET reforming is currently under investigation in our lab.nanoNi-P with various P content and morphologies have been successfully synthesized, using a facile and simple electrodeposition method, and their applications in catalyzing HER, OER, and reforming of EG and PET were explored and investigated. P content of nanoNi-P was found to play an crucial role in regulating the relative amount of hydride-acceptor sites and proton-acceptor sites, and thus determining the intrinsic activity of nanoNi-P towards HER. An ultrathin layer of nickel-oxyhydroxide formed onto nanoNi-P after the CV-activation process, and both β-NiOOH and γ-NiOOH was found to coexist in this thin oxyhydroxide layer. Nonetheless, the presence of P in nanoNi-Pop(CV) played an important role in suppressing further transformation of β-NiOOH into γ-NiOOH, and nanoNi-Pop(CV) showed higher β-NiOOH cotent than nanoNi(CV). β-NiOOH was the main active species responsible for the (photo-)electroforming of EG and PET, rendering higher activity of nanoNi-Pop(CV) over nanoNi(CV). Efficient and selective generation of hydrogen and formate from PEC PET reforming was successfully realized using an Earth-abundant nanoTiO2|nanoNi-Pop(CV) (C = 0.02 C cm−2)//CNT-nanoNi-Pop PEC device. Our work opens a sustainable avenue for simultaneous mitigation of plastic pollution and photosynthesis of renewable fuel and valued chemicals. Chia-Yu Lin: Conceptualization, Writing - original draft, Funding acquisition, Supervision. Shih-Ching Huang: Investigation, Validation, Writing - original draft. Yan-Gu Lin: Funding acquisition, Writing - review & editing. Liang-Ching Hsu: Investigation, Validation. Chih-Ting Yi: Investigation, Validation.The authors report no declarations of interest.We gratefully acknowledge the Ministry of Science and Technology, Taiwan for the financial support (Grant number 110-2218-E-006-016-, 109-2218-E-006-023-, 108-2112-M-213-002-MY3). The research was supported in part by Higher Education Sprout Project, Ministry of Education to the Headquarters of University Advancement at National Cheng Kung University (NCKU). We would also like to thank Dr. Shu-Chih Haw and Prof. Jih-Jen Wu for their technical support. The assistance in HR-TEM analyses from Center for Micro/Nano Science and Technology of National Cheng Kung University was also acknowledged.Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.apcatb.2021.120351.The following is Supplementary data to this article:

Photoelectrochemical reforming of plastic waste offers an environmentally-benign and sustainable route for hydrogen generation. Nonetheless, little attention was paid to develop electrocatalysts that can efficiently and selectively catalyze oxidative transformation of valueless plastic wastes into valued chemicals. Herein, we report on facile electrosynthesis of nickel-phosphorus nanospheres (nanoNi-P), and their versatility in catalyzing hydrogen generation, water oxidation, and reforming of polyethylene terephthalate (PET). Notably, composite of nanoNi-P with carbon nanotubes (CNT/nanoNi-P) requires −180 mV overpotential to drive hydrogen generation at -100 mA cm−2. Besides, CV-activated nanoNi-P (nanoNi-P(CV)) was shown to be capable of reforming PET into formate with high selectivity (Faradic efficiency= ∼100 %). Efficient and selective generation of hydrogen and formate from PET reforming is realized utilizing an Earth-abundant photoelectrochemical platform based on nanoNi-P(CV)-modified TiO2 nanorods photoanode and CNT/nanoNi-P cathode. This work paves a path for developing artificial leaf for simultaneous environmental mitigation and photosynthesis of renewable fuels and valued chemicals.

Porous carbons have been widely used as catalyst supports for the preparation of heterogeneous catalysts because of their high surface area, high chemical and thermal stabilities, diverse surface properties, and tunable macroscopic shape [1–6]. However, the main limitation of carbon materials as support for metal catalysts is their poor stability under oxidation conditions. Nowadays, owing to their attractive performance, porous carbons have been increasingly used as catalyst supports in mild oxidative catalytic reactions that do not deteriorate their structure, including fuel cell [7,8], oxidation of volatile organic compounds (VOC) [9,10], decomposition of NOx [11,12], oxidative dehydrogenation [13,14] and so on. Nevertheless, oxidation-resistant carbon supports are in high demand for the increasing variety of catalytic reactions. A series of methods have been proposed for improving the oxidation resistance of carbon materials, such as graphitization [15], purification [16], surface coating [17] or modification [18], composite fabrication [19] and others (Fig. S1). Purification can slightly improve the oxidation resistance of porous carbon. However, other methods result in serious porosity degradation.Since its experimental discovery in 2004 [20], graphene has quickly attracted tremendous attention and has become one of the most explored nanomaterials in the scientific world. It is possible to prepare graphene-based porous carbon with exceptional properties and a vast application potential from graphene oxide (GO) colloids by chemical reduction [21], intercalation with pillar molecules [22] or nanoparticles [23], thermal exfoliation [24], and KOH activation [25], among them KOH activation method gives rise to porous graphene with the highest porosity. The structure of porous graphene can be further tuned by high-temperature heating under Ar atmosphere. Heat treatment of porous graphene at the temperature region of 1873 ~ 2273 K can partly recover its crystallinity and electrical conductivity by sacrificing a certain amount of porosity according to our previous study [26]. To the best of our knowledge, heat-treated porous graphene has the highest graphitization degree among the high surface area porous carbons except single wall carbon nanotubes (SWCNTs). High quality SWCNTs are still too expensive for a wide variety of applications. The gasification of carbon atoms at the edge of graphene unit through oxidation is 102 to 103 times faster than that on the basal planes [27]. The heat-treated porous graphene with partially graphitized structure should have higher oxidation resistance than conventional porous carbons due to its less prismatic edges and more basal planes, being a promising catalyst support for reactions under oxidative conditions. It is well known that metal catalysts can promote the gasification of carbons under an oxidative atmosphere. Therefore, the development of porous carbon with sufficient oxidation resistance, especially in the presence of metal is necessary. This paper reports the excellent oxidation resistance of heat-treated porous graphene as a metal catalyst support.The GO colloid was prepared using natural graphite (Madagascar graphite from Madagascar) as described elsewhere [28]. The prepared GO suspension was mixed with KOH at a weight ratio of KOH/C = 10, followed by a unidirectional freeze-drying method [29,30]. The obtained GO monolith mixed with KOH was heated up to 573 K at a heating rate of 1.5 K min−1 to prepare the porous graphene monolith-KOH mixture, which was heated up to 1073 K at a heating rate of 30 K min−1 and maintained at this temperature for 1 h. The entire process was conducted under a 400 mL min−1 Ar flow. The mixture was washed with distilled water up to pH = 7 of the supernatant, and then soaked in 1 mol L−1 HCl solution at 353 K for 24 h to completely remove the residual KOH. The obtained monolith was rewashed with distilled water and dried in a vacuum oven at 393 K for 24 h. Finally, the monolith was ground into a powder for further use. The as-prepared sample is denoted as porous graphene (PG). A pitch-based activated carbon fiber ACF-A20 (abbreviated as ACF, AD'ALL Co., Ltd) was studied for comparison. ACF-A20 used here is a metal free porous carbon which is extremely suitable for the research that needs to eliminate the interference from metal impurities.The samples were placed in a graphite-resistance furnace and heated to the target temperature at a heating rate of 20 K min−1 and then maintained at the target temperature for 30 min before cooling down. The entire process was conducted under a pure argon flow (1 L min−1). Thermal treating within the temperature region of 1873 ~ 2473 K enables partial graphitization of PG without degrading all the porosity according to our previous study [26]. We need porous graphene with both high surface area and high graphitization degree as catalyst support, then the heating temperature of 1873 and 2273 K are selected in present work. The yields of the PG and ACF after heat treatment at the two selected temperatures are about 80-85%. The porous carbon (C) heat-treated at T K is denoted as C@T in this study.Metal-loaded carbons were prepared via incipient wetness impregnation method [31,32] with Pd(O2CCH3)2 (toluene solution), Ni(NO3)2•6H2O (ethanol solution), and Fe(NO3)3•9H2O (ethanol solution). All reagents were supplied by Wako Pure Chemical Industries, Ltd (Japan). The metal loading amounts of samples were adjusted to 1.0wt% or 3.0wt%. The impregnated samples were dried at room temperature, then heated up to 623 K at a heating rate of 1.5 K min−1 and maintained at the target temperature for 10 min before cooling down. The entire process was conducted under a pure argon flow (200 mL min−1). The carbon sample (C) loaded with metal M is denoted as M-C in this study.The microscopic morphology of the carbon sample was observed using a field emission scanning electron microscope (SEM, JEOL JSM-6330F, Japan) and a transmission electron microscope (TEM, JEOL 2100F, Japan). The crystallinity change of carbon was examined by synchrotron X-ray diffraction (XRD, Aichi Synchrotron Radiation Center, Japan, λ = 0.07997 nm) and Raman spectroscopy (Renishaw, inVia reflex 785S, UK, λ = 532 nm). Porosity analysis of porous carbons was performed using a surface analyzer (Micromeritics, ASAP 2020, US) by N2 adsorption at 77 K. Before N2 adsorption, the samples were pre-evacuated at 523 K for 5 h. The surface area was determined by the subtracting pore effect (SPE) method [33,34] with carbon black (Mitsubishi 32#) as a reference. The Brunauer-Emmett-Teller (BET) method was also employed to analyze the N2 adsorption isotherms in P/P0 range of 0.05-0.30 for comparison. The micro- and mesopore volumes were determined using the quenched solid density functional theory (QSDFT) with a slit-pore model [35]. The total pore volume was determined using the Gurvitch rule with the N2 adsorption amount at P/P0 = 0.98 [36]. The thermal gravimetric (TG) analysis was performed using a thermogravimetric analyzer (Rigaku, Thermo Plus TG 8120, Japan) from 303 K to 1123 K under a dry air flow (200 mL min−1) at a heating rate of 2 K min−1.The high-resolution TEM images of the pristine and heat-treated carbons (Fig. 1 ) clearly indicate that the graphene units of different porous carbons have different stacking structures depending on the heat treatment. PG mainly consists of crumpled graphene sheets of considerably large size (Fig. 1(a1)), which is very different from conventional highly porous carbon (such as activated carbon or ACF) that consists of small graphene units [37,38]. The micropores and mesopores constructed by entangled and crumpled graphene sheets will be further discussed in Section 3.3. Heat treatment at 1873 K produces partially stacked graphene sheets of larger size (Fig. 1(a2)), and heat treatment at 2273 K leads to the formation of a well-ordered graphitic layer together with a few disordered parts (Fig. 1(a3)). The stacking behavior of graphene sheets of PG under heat treatment has been discussed in detail in our previous study [26]. ACF mainly consists of randomly oriented graphitic units of approximately 2 nm (Fig. 1(b1)), which is similar to the structure of the traditional microporous carbons [37,38]. Heat treatment at 1873 K does not cause significant structural changes of ACF (Fig. 1(b2)), but further heating at 2273 K produces entangled and few-layer-stacked graphene layers (Fig. 1(b3)). SEM images of pristine and heat-treated carbons indicate that the heat treatment at 1873 and 2273 K does not significantly change their exterior morphology (Fig. S2). The graphitization of each carbon under heat treatment will be further discussed with their Raman spectra and XRD patterns in the following section.The graphitic states of different carbons were evaluated by Raman spectroscopy. Pristine PG has two overlapping broad bands located at ~ 1350 and ~ 1590 cm−1 (Fig. 2 (a1)), which are associated with the defects or edges in the graphene unit (D band) and in-plane motion of carbon atoms in the aromatic planes (G band), respectively. Heat treatment at 1873 K leads to a decrease in the D band and an increase in the sharpening of the G band. The decrease in full width at half maximum (FWHM) can be attributed to the growth of the graphitic structure [39]. Further heating at 2273 K gives a very sharp G band and a negligible D band. The intensity ratios of the D band to the G band (ID/IG value) of PG, PG@1873, and PG@2273 are 1.61, 0.39, and 0.06, respectively, indicating the significant graphitization of PG through high-temperature treatment. It is worth mentioning that, PGs prepared from different graphite precursors show different graphitization behavior under heat treatment. PG heat-treated at 1873 K in our previous study [26] doesn't possess high graphitization degree as PG@1873 prepared in the present work does. The reason could be that PG prepared from Bay carbon graphite has smaller and more defective graphene sheets compare to those prepared from Madagascar graphite in current study.The shapes of the Raman spectra and ID/IG values of pristine ACF are similar to those of PG (Fig. 2(b1)). Heat treatment of ACF at 1873 K results in a sharpened D band, and the ID/IG values decrease from 1.63 to 1.42. Further heat treatment at 2273 K remarkably decreases the D band intensity, and the ID/IG value becomes 0.51, which is much larger than that of PG@2273. PG has a more graphitizable structure than ACF. The in-plane crystallite sizes La of graphitic structures obtained using the Tuinstra-Koenig equation [40,41] are shown in Table S1, together with the values of ID/IG and FWHM. The peaks corresponding to the overtone mode longitudinal optical phonons (~2450 cm−1) and the 2D band (~2700 cm−1) become distinct and sharp, while the peak of the (D + G) combination mode (~2930 cm−1) becomes weaker with the increase in heat treatment temperature, suggesting the growth of ordered graphitic structure during the thermal treatment [26].PG has broad (002), (10), and (11) XRD peaks around 13°, 22°, and 39° (Fig. 2(a2)) with three-dimensional X-ray reflection peaks indicating that the stacking structure of graphene units is turbostratic. After heat treatment at 1873 K, all the above-mentioned peaks become sharper, and the (004), (105) + (006), (201) peaks start to appear at 28°, 41°, and 45°, indicating the growth of three-dimensional ordering. Intensive heat treatment at 2273 K results in sharp and distinctive XRD peaks, suggesting the development of a three-dimensional crystalline structure in PG@2273 (Table S2). On the other hand, heat treatment does not significantly change the XRD patterns of ACF (Fig. 2(b2)), although the stacking of graphene layers can be observed in the TEM images (Fig. 1(b3)). The classical work on the graphitization of carbon by R.E. Franklin [42, 43] indicates that graphitizable carbons have weakly linked, compact, and nearly parallel-oriented carbon crystallites, while non-graphitizable carbons have rigidly linked carbon crystallites with random orientations. ACF is a typical non-graphitizable carbon and does not form a highly crystalline structure even upon heat treatment at 2273 K. On the contrary, PG has a representative graphitizable nature even if the as-prepared PG has a disordered graphene unit structure. The observed graphitizable behavior of PG is ascribed to the microscopic assembly structure and property of disoriented graphene units, which means a large planar size and high flexibility.The N2 adsorption isotherms of pristine and heat-treated carbons at 77 K can provide their surface and porosity information. The adsorption isotherm of PG is a combination of type I and IV (Fig. 3 a) according to the IUPAC classification [44]. A significant N2 uptake occurs at low P/P0 region and a distinct hysteresis above P/P0 = 0.4, indicating the presence of both micropores and mesopores. ACF has an N2 adsorption isotherm of type Ⅰ(b) (Fig. 3b), indicating the presence of wider micropores and/or narrow mesopores [44]. Heat treatment at 1873 K induces a slight decrease in N2 uptake for both PG and ACF, while heating at 2273 K results in a significant decrease in N2 uptake due to reduction in porosity. The porosity evolution of PG (Bay carbon graphite as precursor) with the increase of heat-treatment temperature has been systematically investigated in our previous study [26]. The porosity degradation mostly occurs within the temperature region of 1373 ~ 2273 K at an accelerating elevation and finishes at 2473 K. This is because the mutual-stacking of graphene layer in PG needs to overcome a certain energy barrier. The porosity of PG degrades very fast once this energy barrier is broken through. The PG and ACF heated at 1873 K must be applicable to metallic catalyst supports. The porosity parameters of the pristine and heat-treated carbons are listed in Table S3.Heat treatment can induce the graphitization of porous carbons by decreasing their defects and edges, thereby increasing their oxidation resistance. Herein, TG measurements of pristine and heat-treated carbons under dry air were performed. Pristine PG shows a slow weight loss up to 600 K (Fig. 4 a), which is related to the decomposition of oxygen functional groups and desorption of adsorbed gas. Then, it starts burning rapidly and finishes at approximately 800 K. The temperatures at 5% and 50% burn-off ratios are 668 and 802 K, respectively. These two temperatures are defined as “burning threshold temperature” (BTT) and “half-burned temperature” (HBT) in this study to evaluate the oxidation resistance. The pristine PG residual after TG measurement is less than 0.1wt%. PG@1873 has a BTT and HBT of 830 K and 906 K, respectively, and is more thermally stable than pristine PG. The PG@1873 residual after TG measurement is undetectable. PG@2273 shows almost no weight loss at temperatures below 900 K, and starts burning rapidly at higher temperatures. The BTT and HBT of PG@2273 are 945 K and 1016 K, respectively.Heat treatment induces a similar change to ACF compared with PG, i.e., the BTT and HBT of ACF increase with the heating temperature (Fig. 4b, Table S4). However, the increase in BTT and HBT of PG is more significant than those of ACF (Fig. 4c, 4d), which results into higher thermal stability of heat-treated PGs. Pristine PG, on the other hand, shows lower thermal stability than pristine ACF. The increase in the thermal stability of carbon through heat treatment can be ascribed to the following two factors. The heat treatment can remove trace amounts of metal impurities [45, 46], which act as catalysts during the gasification of carbon. Another reason is that graphitization reduces the edges and defects of carbon, on which the oxidation is 102 to 103 times faster than that on the basal planes [27]. Pristine ACF is metal-free carbon, which has a higher oxidation resistance than that of pristine PG, which contains trace amounts of metal impurities from the Hummer process and KOH activation even after soaking in 1 mol L−1 HCl solution. Heat treatment can effectively remove trace amounts of metal impurities in PG, thus diminishing the catalytic effect of oxidation. Moreover, PG has a more graphitizable structure than ACF; its defects and edge carbons decrease faster than those of ACF, ensuring a more pronounced increase in thermal stability under heat treatment.Porous carbons are often used as supports in many catalytic reactions. Consequently, the high oxidation resistance of porous carbon in the presence of metal is of significant importance. PG@1873 prepared in the present study has both a large surface area and high crystallinity, and is used as the model for further investigation. Herein, a TG study was performed on PG@1873 loaded with three kinds of metals (Pd, Ni, and Fe; 1.0wt%). Metal-loaded ACF@1873 was studied in the same way for comparison. The synchrotron XRD patterns of all metal-loaded carbons show no extra diffraction peaks except for the carbon supports (Fig. S3), suggesting a highly dispersed state of the loaded metals. The above three metals are usually dispersed in single-atom states if they are loaded at low weight ratios (around 1.0wt%) on carbons [47–49]. TG measurements indicate that the BTT and HBTs of all carbons decrease after the metal loading (Fig. 5 a, 5b and Table S5). By comparing the TG curves of PG@1873 and ACF@1873 loaded with the same metal, it was found that M-PG@1873 showed a smaller decrease in both BTT and HBT than M-ACF@1873 (Fig. 5c, 5d and Fig. S4), which means that PG@1873 is more oxidation-resistant than ACF@1873 in the presence of a metal catalyst. In particular, the oxidation resistance of PG@1873 for Fe catalysts is evident; the BTT of Fe-PG@1873 is 130 K higher than that of Fe-ACF@1873. Molecular dynamic studies indicate that Ni atoms at the edge of graphene can fracture the C-C bond and promote the diffusion of carbon atoms, while those at the graphene plane are much less active [50]. PG@1873 prepared in this work has a highly crystalline graphene plane with fewer edges and defects, on which the metal catalyst cannot break the C-C bond effectively, and therefore is more thermally stable than less ordered metal-loaded [email protected] is important to understand the oxidation resistance as the metal loading increases. Here we also studied the oxidation resistance of PG@1873 and ACF@1873 with Fe loading of 3.0wt%. The synchrotron XRD pattern of thus prepared Fe-PG@1873 show a few diffraction peaks from Fe-contained compound (Fig. S5), while that of Fe-ACF@1873 show no extra diffraction peaks. Nanoparticles with higher crystallinity may be preferentially produced on graphene walls with higher graphitization degree. Nevertheless, the BTT and HBT of both PG@1873 and ACF@1873 decrease with the heating temperature, and PG@1873 shows much smaller decrease in both BTT and HBT than ACF@1873 with Fe loading of 1.0wt% and 3.0wt% (Fig. S6). In particular, PG@1873 loaded with 3.0wt% of Fe still have considerably high oxidation resistance.The oxidation resistance of metal-loaded PG@1873 is also compared with those of other reported carbon supports (Fig. 6 a, 6b, 6c). Because of the difficulty in many studies to determine the BTT caused by the decomposition of metal compounds for catalysts at lower heating temperatures, we only compared the HBT of different carbons. The nomenclature, abbreviation, and detailed information of these metal-loaded carbons are given in Tables S6, S7, and S8. It is evident that the oxidation resistance of PG@1873 is remarkable compared with most of the carbon supports in the presence of metal, while that of ACF@1873 is average. The oxidation resistance of metal-containing carbon is highly related to its graphitization degree, as discussed in this section. The high graphitization degree of PG@1873 ensures superior oxidation resistance than in case of other carbons. It is worth mentioning that MWCNT-2, MWCNT-4 and MWCNT-5 show higher oxidation resistance than the PG@1873 reported in this study. It is explained by the almost perfect crystal structure of the well-prepared multiwall carbon nanotubes (MWCNT). The TEM image and XRD pattern of MWCNT-2 clearly indicate its highly ordered graphene layers and sharp (002) diffraction peak [51], suggesting that it has a higher ordering than that of PG@1873. Nevertheless, MWCNT has the limitation of small specific surface area and nonporous features as catalyst supports. A high surface area is essential for high-quality catalytic support. We plotted the HBT versus specific surface areas for the carbon supports whose surface areas are available in their references. The present PG@1873 is situated near the right-upper corner of Fig. 6d, which is far from other catalyst supports. The partially graphitized porous graphene PG@1873 prepared in the present work has both the merits of high oxidation resistance and large surface area, showing great potential as catalyst support for reactions under mild oxidative conditions.This study reveals that heat treatment can improve the oxidation resistance of porous graphene more efficiently than conventional porous carbon, such as activated carbon fiber. Porous graphene has a more graphitizable structure than ACF because of its unique constitutional graphene units with large planar size and high flexibility. The partially graphitized porous graphene PG@1873 prepared in this work has both the merits of large surface area and high oxidation resistance. After metal loading, the oxidation resistance of PG@1873 is well maintained, while that of ACF@1873 deceases. Such unique properties make PG@1873 a promising catalyst support for catalytic reactions conducted under oxidative conditions. Shuwen Wang: Conceptualization, Methodology, Formal analysis, Investigation, Data curation, Writing-original draft. Yasunori Yoshikawa: Investigation. Zhipeng Wang: Investigation. Hideki Tanaka: Formal analysis, Investigation. Katsumi Kaneko: Conceptualization, Methodology, Writing- review & editing, Supervision.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This work was supported by a Grant-in-Aid for Scientific Research (B) [grant number 17H03039] and the OPERA Japan Science Technology Agency project, Japan.Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.cartre.2021.100029. Image, application 1

Porous graphene (PG) prepared from reduction and KOH activation of graphene oxide was heat-treated under argon atmosphere to obtain a partially graphitized porous carbon with high oxidation resistance. Transmission electron microscopy, Raman spectroscopy, synchrotron X-ray diffraction, and N2 adsorption isotherms (77 K) clearly illustrate the structural ordering and porosity change of PG under heat treatment. Pitch-based activated carbon fiber (ACF) was studied for comparison. PG is more graphitizable than ACF under heat treatment because it consists of highly flexible graphene units of larger size than those in ACF. Thermogravimetric studies indicate that heat treatment enhances more the thermal stability of PG than ACF, and metal-loading has a less detrimental effect on the thermal stability of heat-treated PG than heat-treated ACF and other reported carbon supports. Heat-treated PG shows great superiority to other carbon supports due to its both splendid oxidation resistance and high surface area. This study provides a promising route for the preparation of carbon-based catalyst supports for mild oxidative environments.

The steam reforming of hydrocarbons is a widely used process for obtaining hydrogen. Nowadays, almost 50 % of its worldwide production is acquired via steam reforming reaction [1,2]. The process analyzed in this research occurs between methane and steam, specifically. It produces a mixture of H2 and CO, commonly known as syn-gas. Until the two last decades, H2 was mainly obtained using large scale reforming reactors, for the needs of the industrial ammonia synthesis on the way of the Haber-Bosch process [3]. Currently, the demand for scaled-down reactors is rising, due to the rapid development of fuel cells [4] and the necessity of producing hydrogen on-the-spot [5]. Small-scale reactors currently surpass the large-scale reactors, due to the lack of proper pipelines for the hydrogen transport. For now, gas distribution networks are not suitable for transporting hydrogen [6]. Therefore, it is more cost-effective to set up a small-scale reactor, directly in the location of hydrogen demand [6]. The small-scale reactors are also advantageous when it comes to the processing of distributed or stranded resources for hydrogen production [7]. These facts force a growing interest in the improvement of the small-scale reactor's performance. However, scaling down the steam reforming reaction is not a simple task. To keep the cost of a small-scale unit relatively low, they often have to come in a modular and standardized design. The operating conditions of the small-scale applications are also required to be less severe than in the industrial-scale reactors, to reduce the material cost of a single reactor [6]. Following, the temperature of the process has to be reduced. That results in a reduction of the methane conversion [8], being just one of the issues connected with scaling down the process [6].By far, researches were mostly focused on the parametric study and optimization of the reaction conditions [9,10]. However, this approach has limited effectiveness, as the reaction conditions can be improved just to a certain level. Therefore, further development of the process is pursued by the introduction of new materials and design concepts, like new catalyst structures [11,12], introducing new kinds of catalyst supports [13], or by rethinking the design of the reactor itself [14–16]. Other works focus on an optimization of the process' cost effectiveness by the introduction of renewable sources for hydrogen production [17–19]. The cited researchers have focused on different aspects of the steam reforming and the possibilities of its enhancement and cost reduction.Another captivating opportunity was pointed out by Palma et al. [20], who introduced a structured catalyst for the intensification of the reforming reaction. Their work concentrated on the importance of an effective heat transfer and its beneficial influence on the reaction rate. They confirmed that improving axial and radial temperature distribution results in better performance of the reactor, due to rising catalytic activity. Yun et al. [21] have also focused on the enhancement of heat transfer during the process. They proposed a reactor design with a maximized heat transfer area, which had also led to better performance of the reforming unit. Dubinin et al. have stated that proper handling of heat in the reforming process can also lead to an increase of the process' cost-effectiveness, as excess heat could be reused by applying a heat exchanger and incorporating the reactor into a mini-CHP unit [22].The strong endothermic character of the reaction causes the development of a non-uniform temperature field inside the reformer [23]. Rapid temperature decay at the upstream region has a consequence in the presence of thermal stresses inside the reactor, leading to its uneven degradation and reduction of the unit's lifetime [24]. Thus, the unification of the temperature distribution may not only improve the operational conditions but also achieve easier control of the process. Therefore, our team decided to develop an original strategy for the improvement of the temperature distribution across a steam reforming reactor. Motivated with promising results obtained by Settar et al. [25], we decided to improve their idea of the macro-patterned active surfaces [25] with the introduction of metallic foam matrices [26]. Their first research, which was reported in Ref. [25] introduced the division of the catalytic region of a wall-coated reformer and separating the created zones with inter-catalytic spacing, which led to an intensification of the hydrogen production. The second one [26] focuses on providing advantageous thermal conditions for the reaction, as metallic foams were previously reported to deliver a considerable heat transfer area [27], leading to the improvement of the temperature distribution inside the reactor. First of all, instead of plating a catalyst only on the wall surface, we decided to fill a whole reactor's volume with a catalytic composite of nickel and yttria-stabilized zirconia (Ni/YSZ), to maximize its active surface contained in the reactor. Then, the reforming unit was divided into separate segments, and part of them got substituted with non-catalytic metallic foam, to adjust the intensity of the reaction proceeding, leading to the unification of the thermal field inside the reactor. The acquired results showed that the macro-patterning concept could be successfully used to alter the temperature field in the reforming reactor and enhance the reaction rate [16]. This was also confirmed by further investigations by Settar et al. [28,29].Basing on the previously cited works, we decided that the proper optimization of the segments' alignment is necessary, to achieve a uniform temperature distribution, simultaneously maintaining as high a methane conversion rate as possible. The presented research regards the optimization of the particular segments distribution and adjustment of Ni/YSZ densities used in the catalytic ones, for the proper altering of the reforming reaction rate (Section Mathematical model). Optimization of the given issue is meant to allow the unification of the temperature distribution in the macro-patterned reactor [16]. However, it has to be conducted with respect to maintaining high hydrogen production effectiveness, determining that the problem we are dealing with is a multiobjective optimization. The objectives behave contrarily, as the temperature gradients can be only reduced by limiting the amount of the catalyst used in a particular reactor, which has a consequence in the decreasing of the methane conversion rate. The less a catalyst is used, the more a reaction rate is diminished. The described problem becomes even more complex if we consider the amount of possible segments combinations (Section Genetic algorithm). Many configurations can be falsely recognized as a global optimum when being only an optimal solution in the currently investigated part of the search space. Moreover, the steam reforming numerical simulation has a considerable computational time, as it combines heat and mass transfer phenomena, as well as chemical reactions occurring during the process [30]. Therefore, its implementation in a proper optimization algorithm is a demanding task.The genetic algorithm (GA) was chosen to be the mean of this optimization. Due to its non-deterministic nature and a vast search region, it is expected to handle with even adversely conditioned problems successfully. Moreover, it has a high plausibility of finding the global optimum in a limited time [31]. Previous researches have confirmed the applicability of GAs to the steam reforming optimization. Taji et al. have successfully conducted a multiobjective optimization of the methane conversion rate and hydrogen yield in an industrial hydrogen plant [32]. The optimization have taken into account the catalyst deactivation over time. On the other hand, Zheng et al. have reported the results of a single objective GA optimization of the methane conversion, by the adjustment of the reaction parameters and the catalyst load in a micro-reactor [33]. The presented works validate GA to be an effective technique for the steam reforming optimization, as the method allows for definition of many objectives, which may be pursued effectively, regardless of the number of considered parameters [33,34]. Although, implementations of GA in the reforming optimization have been reported previously, no papers regarding a multiobjective optimization of the reactor design has been found. The multiobjective optimization of the catalyst distribution in a small-scale steam reforming reactor constitutes the original contribution of this paper. The presented work introduces new aspects in the development of the steam reforming process, as follows: • incorporation of the physical properties of the metallic foams into the mathematical model, both for the catalytic and non-catalytic segments, • the multiobjective optimization of the segments alignment and their morphological features using a genetic algorithm, • the improvement of the temperature distribution, by altering the amount of catalyst loaded into a single reactor, maintaining a relatively significant methane conversion rate. incorporation of the physical properties of the metallic foams into the mathematical model, both for the catalytic and non-catalytic segments,the multiobjective optimization of the segments alignment and their morphological features using a genetic algorithm,the improvement of the temperature distribution, by altering the amount of catalyst loaded into a single reactor, maintaining a relatively significant methane conversion rate.The geometry of the reformer considered in this analysis is presented in Fig. 1 . It was assumed to be an axisymmetric tubular reactor. The analyzed model is steady. The reforming unit consist of a cylindrical pipe, divided into thirty separate segments, indicated with the red dashed line. The segments vary in values of porosity, pore diameter and density of the used catalytic material. The flow of gases is assumed to be laminar, steady and in one direction. The fluids taking part in the process are assumed to be Newtonian. The properties of the substances included in the reactions were taken from the literature [35].The reactor is supplied with a mixture of hydrogen (H2) and steam (H2O). The relation between their amounts is described with steam-to-carbon ratio (SC). This parameter is very vital for the steam reforming, as too low a SC may have a consequence in carbon deposition, limiting the active surface of the catalyst [36]. The two main reactions occurring during the considered process are as follows: - methane/steam reforming (MSR) reaction: (1) C H 4 + H 2 O → 3 H 2 + C O , Δ H M S R = 206 . 1 k J m o l , - water-gas shift reaction (WGS): (2) CO + H 2 O ⇌ H 2 + CO 2 , Δ H WGS = − 41.15 kJ mol . methane/steam reforming (MSR) reaction:water-gas shift reaction (WGS):Equations (1) and (2) were incorporated into the model by the formulation of a proper expression describing the rates of the reactions. The MSR reaction proceeds rather slowly and its rate can be described with: (3) R MSR = w ˙ cat A MSR exp − E a R ¯ T p CH 4 α p H 2 O β . The exact values of parameters contained in (3) were acquired during experimental research conducted earlier by our team [30]. The WGS reaction, expressed with (2) has a contrary nature, as it proceeds fast. It has been assumed to remain in equilibrium under the conditions present during the considered process, as explained in Ref. [37]. This was also confirmed by analyzes conducted previously [38–41]. According to the given literature review, this assumption can be validated. Therefore, CO, CO2, H2 and H2O have to satisfy the equilibrium equation, given below: (4) K WGS = k WGS + k WGS − = p CO 2 p H 2 p CO p H 2 O = exp − Δ G WGS 0 R ¯ T , allowing to formulate the WGS reaction rate equation, as follows: (5) R WGS = k WGS + p CO p H 2 O + k WGS − p H 2 p CO 2 . The value of Eq. (5) can be acquired through the analysis of the reforming process stoichiometry and balancing the chemical species. Following that, we are able to specify the methane conversion rate x c r and carbon monoxide conversion rate y c r : (6) x c r = 1 − n CH 4 inlet − R MSR V n CH 4 inlet , (7) y c r = K WGS + 3 x c r − χ − ω 2 K WGS − 1 , where: (8) χ = K WGS S C + 3 x c r 2 , (9) ω = 4 K WGS x c r K WGS − 1 S C − x c r , and consequently calculate the partial pressures included in Eq. (4), basing on the reactions' stoichiometry as well, leading to the following relation [30]: (10) R WGS = n CH 4 outlet V = n CH 4 inlet x c r V y c r . After connecting Eq. (6) with Eq. (10) and applying simple mathematical transformations, a final expression for the WGS reaction's rate is formed: (11) R WGS = R MSR y c r . The mass consumption and production rates of the mentioned reactions (Eqs. (1) and (2)) are summarized in Table 1 . These values are further applied into the heat transfer model (Section Heat and mass transfer model). Now, the chemical reactions model is almost complete, as only thermodynamic heat generation rates of the reactions are left for formulation. They can be acquired by multiplying the reaction rates (Eqs. (3) and (11)) by their enthalpies: (12) Q MSR = − Δ H MSR R MSR , (13) Q WGS = − Δ H WGS R WGS . The mathematical model used in this analysis is based on the fundamental transport equations (Eqs. (14)–(18)). The governing equations were derived using volume-averaging method, due to application of porous structures [42]: (14) ∂ ρ 0 U x ∂ x + 1 r ∂ r ρ 0 U r ∂ r = 0 , (15) ρ 0 ε 0 2 U x ∂ U x ∂ x + U r ∂ U x ∂ r = − ∂ P ∂ x + μ ε 0 [ ∂ 2 U x ∂ x 2 + 1 r ∂ ∂ r r ∂ U x ∂ r   − μ K p U x − ρ 0 c ine K p U x U x 2 + U r 2 , (16) ρ 0 ε 0 2 U x ∂ U r ∂ x + U r ∂ U r ∂ r = − ∂ P ∂ r + μ ε 0 ∂ 2 U r ∂ x 2 + 1 r ∂ ∂ r r ∂ U r ∂ r − U r r 2 − μ K p U r − ρ 0 c i n e K p U r U x 2 + U r 2 , (17) ρ 0 C p U x ∂ T loc ∂ x + U r ∂ T loc ∂ r = ∂ ∂ x λ eff ∂ T loc ∂ x + 1 r ∂ ∂ r r λ eff ∂ T loc ∂ r + Q s , (18) ρ 0 U x ∂ Y j ∂ x + U r ∂ Y j ∂ r = ∂ ∂ x ρ 0 D j , eff ∂ Y j ∂ x + 1 r ∂ ∂ r r ρ 0 D j , eff ∂ Y j ∂ r + S j . The effective mass diffusivity of species D j , eff is calculated using Eq. (19), which is explained below [43]: (19) D j , eff = 1 − 1 − ε 0 D j . The permeability K p of the specific segment is calculated using Eq. (20), basing on the information about its porosity ε 0 [44]: (20) K p = ε 0 1 − 1 − ε 0 1 / 3 36 1 − ε 0 1 / 3 − 1 − ε 0 d p 2 , where d p stands for an average pore diameter. The inertial coefficient c ine was calculated using [45]: (21) c ine = 0.0095 g s − 0.8 ε 0 3 τ − 1 1.18 1 − ε 0 3 π 1 g s − 1 , where tortousity τ and shape function g s are expressed with following equations [44,45]: (22) τ = ε 0 1 − 1 − ε 0 1 / 3 , (23) g s = 1 − exp ( − 1 − ε 0 0.04 ) . The properties of gases taking part in the process were taken from the literature [46]. The heat transfer model applied in this analysis, allows calculation of the effective thermal conductivity λ eff [47]. It describes heat propagation in a structure of a metallic foam, which can be calculated using: (24) λ eff = 2 l 2 R A + R B + R C + R D , where R A−R D stand for the thermal resistances of the porous media cell subsections and can be calculated as follows [47]: (25) R A = 4 d l 2 e 2 + d π 1 − e λ solid + 4 d l 4 − 2 e 2 + d π 1 − e λ mix , (26) R B = e − 2 d l e 2 λ solid + 2 − e 2 λ mix , (27) R C = 2 − 2 e l π d 2 λ solid 2 + 2 − π d 2 2 λ mix , (28) R D = 2 e l e 2 λ solid + 4 − e 2 λ mix . Formulas (25)–(27) require knowledge of the foam ligament radius d, acquired by solving [47]: (29) d = 2 2 − 2 ε 0 − 3 2 4 e 3 π 3 − e − 4 e 2 . For the needs of the numerical analysis, the dimensionless cubic node length e was set to be equal to 0.0339, as explained in Ref. [48]. The values of the thermal conductivity λ solid for the catalytic material and metallic foam were taken from the literature [49,50]. The heat conductivity of the gases mixture λ mix inside the reformer was calculated using the mixing laws [46].After having defined the mathematical model for the methane/steam reforming (Section Mathematical model) preparation of an adequate numerical model is needed. The Finite Volume Method was chosen for the discretization of the governing equations used in the mathematical model [51,52]. Each of the partial differential equations, described in Section Heat and mass transfer model, can be written in a generalized form, as follows Eq. (30): (30) Ψ x ∂ φ ∂ x + Ψ r ∂ φ ∂ r = ∂ ∂ x Γ ∂ φ ∂ x + 1 r ∂ ∂ r r Γ ∂ φ ∂ r + S . ¯ The coefficients given in Eq. (30) originate from the transport Eqs. (17 and 18), and their values are gathered in Tables 1 and 2 Although, if we consider a non-catalytic segment, the sources S ¯ for Eqs. (17) and (18) are equal to 0, as the chemical reactions are assumed to be suppressed on these segments [16].The discretized transport Eq. (31) was obtained after the integration of Eq. (30) over the created control volumes. Following simple mathematical transformations, it can be presented in a form, given by Ref. [51]: (31) Ψ x φ e − Ψ x φ w r m Δ r + r Ψ r φ n − r Ψ r φ s Δ x = Γ ∂ φ ∂ x e − Γ ∂ φ ∂ x w r m Δ r + r Γ ∂ φ ∂ r n − r Γ ∂ φ ∂ r s Δ x + S ¯ r m Δ r Δ x , (32) a P φ P = a E φ E + a W φ W + a N φ N + a S φ S + b , where: (33) a P = a E + a W + a N + a S , (34) b = S ¯ r m Δ r Δ x , The Power Law scheme was applied to calculate the fluxes crossing the control volume faces. Each of these fluxes is represented by a specific coefficient a j [51]. The subscripts E, W, N, S dicate the location of a specific face, relating to the analyzed node. They correspond to the geographic directions, therefore a E represents a face on the left side of the node, a W on the right, a N a face above and a S underneath the node. The momentum equations were solved using the SIMPLE algorithm, which allowed to calculate the velocity values in the specific grid points [51]. The created systems of partial differential equations were solved using a prepared Gauss-Seidel linear solver, due to its high robustness [51]. Afterwards, the numerical model was implemented in the C++ programming language. The complete numerical analysis (Section Numerical Results) was conducted using the designed in-house numerical code. The convergence criterion set for this analysis was defined as acquiring the difference between the values of the mass and heat sources, in two subsequent iterations, lower than 10−5. The grid's resolution was established at 150 elements in the longitudinal and 25 elements in the radial directions, resulting in a square-shaped element. These dimensions were assumed to be sufficient for the conducted analysis, as our previous research [16] already confirmed it.For this analysis, the S C was set to be equal to 2.0, as this ratio was previously reported to be high enough to effectively prevent carbon deposition [53]. Each segment of a specific reactor can be filled with non-catalytic steel foam or catalytic Ni/YSZ of various densities, calculated basing on the segment's porosity (Section Genetic algorithm). The segments' alignment is the most crucial aspect for altering the temperature distribution inside the reformer [16]. The endothermic character of the reaction results in a temperature drop within the catalytic segments and the higher density of the catalyt in a specific segment. The drop is expected to be greater, as the reaction becomes more intense [8]. The metallic foam segments were introduced for the limitation of the temperature decrease. They tend to have good thermal conducting properties, what combined with their considerable surface of heat exchange, is expected to allow the gases mixture to reheat effectively [27]. The value of λ solid was set at 22 W m−1 K−1 for the catalytic material [49] and at 30 W m−1 K−1 for the steel foam [50]. The temperature of the fuel flowing inside the reforming unit is considered to reach the temperature of the reformer instantly. The symmetry boundary conditions were applied at the symmetry axis, whereas the no-slip boundary conditions were set at the wall of the reactor. All of the boundary conditions were summarized in Fig. 2 . The thermal conditions in this analysis were set as follows: • inlet temperature T = T in = 900 K at x = 0 and 0 ≤ r < R , • outlet temperature ∂ T / ∂ x = 0 at x = L and 0 ≤ r < R , • symmetry boundary condition ∂ T / ∂ r = 0 at 0 ≤ x < L and r = 0 , • wall temperature T = T wall = 900 K at 0 ≤ x < L and r = R . inlet temperature T = T in = 900 K at x = 0 and 0 ≤ r < R ,outlet temperature ∂ T / ∂ x = 0 at x = L and 0 ≤ r < R ,symmetry boundary condition ∂ T / ∂ r = 0 at 0 ≤ x < L and r = 0 ,wall temperature T = T wall = 900 K at 0 ≤ x < L and r = R .The thermal boundary condition was chosen to be of the first-type, as setting a constant temperature value has the most negative influence on the unification of the temperature distribution inside the reformer. Due to this fact, the results of the conducted analysis are more reliable. The boundary conditions are also essential for calculating the mass transport equations: • inlet mole fractions Y j = Y j , in at x = 0 and 0 ≤ r < R , • outlet mole fractions ∂ Y j / ∂ x = 0 at x = L and 0 ≤ r < R , • symmetry boundary condition ∂ Y j / ∂ r = 0 at 0 ≤ x < L and r = 0 , • no-slip boundary condition Y j = 0 at 0 ≤ x < L and r = R . inlet mole fractions Y j = Y j , in at x = 0 and 0 ≤ r < R ,outlet mole fractions ∂ Y j / ∂ x = 0 at x = L and 0 ≤ r < R ,symmetry boundary condition ∂ Y j / ∂ r = 0 at 0 ≤ x < L and r = 0 ,no-slip boundary condition Y j = 0 at 0 ≤ x < L and r = R .The boundary conditions for the Navier-Stokes equations were provided as follows: • inlet velocity U=U in=0.15 m s−1 at x = 0 and 0 ≤ r < R , • outlet velocity ∂ U / ∂ x = 0 at x = L and 0 ≤ r < R , • symmetry boundary condition ∂ U / ∂ r = 0 at 0 ≤ x < L and r = 0 , • no-slip boundary condition U = 0 at 0 ≤ x < L and r = R . inlet velocity U=U in=0.15 m s−1 at x = 0 and 0 ≤ r < R ,outlet velocity ∂ U / ∂ x = 0 at x = L and 0 ≤ r < R ,symmetry boundary condition ∂ U / ∂ r = 0 at 0 ≤ x < L and r = 0 ,no-slip boundary condition U = 0 at 0 ≤ x < L and r = R .The genetic algorithm (GA) was used to find the most optimal catalyst distribution in the methane/steam reforming process. This kind of optimization technique was developed by John Holland and his associates [31]. It is an example of a stochastic method, having its origins in the evolution process. Thus, vocabulary originating from biology and genetics is used for their description [31]. Although GAs are qualified as search algorithms, they can be distinguished from the conventional ones.First of all, they operate not with the parameters themselves, but with their representations in the form of finite-length strings, binary in the case of this analysis. GAs start their operation with a whole population of solutions, instead of a single solution. This approach is intended to limit the risk of the algorithm engage in a local extremum when a multimodal search space is considered. After acquiring results for the given set of solutions, they are evaluated with fitness functions, assigning a specific fitness value for each of the specimens. The calculated fitness is further used for determining which specimens perform best and should be preferred for the composition of a consecutive population of solutions. A gene is the smallest element of the GA, and it represents a single character of a code string. A chromosome stands for a complete code string, and in this analysis, it represents the parameters of a specific segment of the macro-patterned reactor, which are the segment's porosity and pore size. A set of chromosomes composes a genotype, which should be understood as a single specimen in a specific population.Having introduced the essential vocabulary allows to describe the principles of the GA used in this analysis. Its operation starts with randomizing the parameters of an initial population, composed of thirty reactors, each divided into thirty independent segments. The algorithm starts with defining segments’ average pore size and porosity, which was constrained to be selected from values between 0.5 and 0.8. As for the lower porosities the pressure drop was reported to be overly significant [54]. The upper boundary was set at 0.8, due to the lower accessibility of the manufacturing methods for metallic foams of higher porosities [55]. Having the porosity for a specific segment, the algorithm proceeds to the selection of its average pore diameter. Its values were constrained as explained in Ref. [56]. What is more, digital material representations (DMR) of metallic foams were used to acquire exact ranges of pore size for foams of particular porosity (Section Digital material representation). After this part is done, it randomizes if a specific segment is meant to be a non-catalytic or catalytic one. If the later is chosen, the catalyst density w ˙ cat is calculated with: (35) w ˙ cat = ρ cat ⋅ ( 1 − ε 0 ) , where ρ cat was equal to 5.3448 106 g m−3, as the ratio of Ni to YSZ used for derivation of the reaction kinetics (Section Mathematical model) was 60:40 and the solid Ni density is equal to 8.908 106 g m−2. Afterwards, the GA calls the reforming simulation code over each of the specimens and when calculations are complete, it gets to the evaluation process, basing on two separate fitness functions, determining that we deal with a multiobjective optimization [57]. First of the fitness functions Eq. (36) calculates the methane conversion rate, determining the amount of methane which has undergone conversion during the reaction. The second one, expressed with Eq. (37), is based on the difference between the maximal and the minimal temperatures present inside reactors, T max and T min respectively. What is more, the described functions are in conflict, as maximization of Eq. (36) is connected with rising of the catalyst density used in the specific reactor and minimization of Eq. (37) requires an opposite approach, making the whole optimization process even more arduous. The equations for fitness calculation were formulated as follows: (36) f CH 4 = f r a c CH 4 in − f r a c CH 4 out f r a c CH 4 in , (37) f T = 1 − T max − T min Δ T max , where Δ T max stands for the temperature difference acquired for a reference case (Section Numerical results). After calculating values of Eqs. (36) and (37) a single fitness value f is needed. This can be acquired by application of the weighted-sum method [58]: (38) f = ∑ w j f j . The weights w j were chosen arbitrarily and their values are equal to 0.6 for f CH 4 and 0.4 for f T , as it was decided that methane conversion is a more important factor in this process. Afterwards, the algorithm chooses the two best performing reactors, which advance to the next population intact and the crossover procedure begins. The crossover starts with a partially-random selection of potential parent specimens. Basing on the acquired fitness values, the probability of selection for each specimen is calculated. The probability value is acquired by dividing the fitness of each specimen by the sum of the fitness values acquired for the whole population. Then, specimens are paired using a roulette wheel selection and the recombination of their chromosomes starts [31]. The point of crossover is randomized, both chromosomes split and their bit strings are interchanged. In the case of this analysis, the algorithm performs two crossovers, first explicit and second implicit. The explicit one occurs first and is used for a crossover of the pore size and the porosity of a considered segment. Following that, the implicit one determines if that segment is supposed to be a catalytic one or not. It happens via a crossover of the catalyst density value. If its results are contained in the permissible range, which boundaries are calculated using Eq. (35), the segment is set to be catalytic and the catalyst's density is calculated basing on the porosity value acquired during the explicit crossover procedure. Next, the mutation is performed. It is a random process, converting a single gene in the bit string representing the segment's parameters, to have an opposite value. This mechanism helps avoiding the local extremum trap, by the introduction of genes unable to be acquired during the crossover process, as they are limited by the parameters generated for the initial population [31]. The mutation is described with its rate, corresponding to the probability of its occurrence. Depending on the literature source, many different rates are advised. Typical GAs with numerous populations are suggested to have mutation rate equal to 2 % [31] or 1 / X %, where X stands for the bit string's length [59]. However, in the case of this analysis, the population is strictly limited to thirty individuals and according to Ref. [60], a mutation rate equal to 10 % was applied. The mutation is individually performed for each segment, meaning that it is randomized thirty times for a specific reactor. This part of the GA's operation is repeated until a new population is complete. Then, the whole process starts over again, until the convergence criteria are met, which were set to be the methane conversion rate over 60 % and a temperature difference lower than 25 degrees. The algorithm operation is summarized in Fig. 3 .For the improvement of pore size range accuracy, a set of over three hundred digital representations of metallic foam was generated. The process was conducted using an in-house code, developed basing on the random geometric graphs. The DMR generation approach is very similar to the algorithm presented in Neumann et al. [61].Exemplary foams are provided in Fig. 4 . The composed algorithm starts its operation with the generation of a random graph in a cube of specified dimensions. Every node and connection is a source of a field similar to the gravity field. Afterwards, it creates a grid of voxels, composing a cuboid. Every voxel has its phase assigned basing on the phase force coefficients defined by the user. The force coefficients decide how highly a node or a connection can influence the given voxel.The generated set of DMRs was further uploaded to the Avizo software, which allowed for a quantitative analysis of the structures’ morphology. Basing on the acquired results, the upper border of the average pore size for the considered range of porosity, is set at 0.002 m. The lower border of the average pore size range is described by the following equation: (39) d p = 3.2894 ε 0 − 0.6315 . Defining the borders of an admissible average pore size completes the necessary assumptions. Now, the GA is provided with every information essential for calculations commencement.Having developed the described mathematical model and optimization algorithm allows to perform a set of numerical calculations. The strongest emphasis is put on analyzing the thermal conditions inside the reactor and the reaction's products. The pursued reactor has to combine improvement in the thermal conditions, maintaining as high a methane conversion rate as possible. Otherwise, the application of the identified solution would be unprofitable.Preceding the optimization process, a definition of a reference case is needed. The reference reactor has a homogeneous catalyst distribution, and its density in the particular segments is set to have the maximal possible value. Following, a reactor with maximal CH4 conversion and thermal conditions typical for a conventional reactor is composed. The maximal Ni/YSZ density is acquired for segments of the lowest admissible porosity, equal to 0.5, resulting in a density value equal to 2.67 ⋅ 106 g m−3 (Eq. (35)).The fitness value calculated for the reference case is equal to 0.53 and its temperature distribution is presented in Fig. 5 . After analyzing the temperature field, the most significant decrease in its value can be noticed at the inlet of the reactor. The observed drop in the tempaerature value occurs due to the activation of the reforming reaction [8]. A conclusion can be drawn, that the MSR reaction dominates the temperature field formation. Closing to the reactor's outlet, less CH4 remains left for conversion. Thus, less heat is consumed by the MSR reaction and the temperature gradients start to diminish.For the reference case, the methane conversion rate reached 0.82. The distribution of the mole fractions is presented in Fig. 6 . The mole fractions change in the longitudinal direction, confirm the previous conclusion about change of temperature gradient, being correlated to the amount of CH4 left for conversion. Therefore, it can be concluded that the thermal conditions can be moderated by changing the rate of the reforming reaction.The developed genetic algorithm starts its operation with the generation of the initial population. It is composed of thirty specimens, with fully randomized parameters of the segments. Afterwards, the methane/steam reforming simulation is called over each of them. The initial set's distribution of thermal fitness f T and methane conversion rate f CH 4 , is presented in Fig. 7 . The calculated fitness value for the best solution included in the initial population is equal to 0.47. Its temperature distribution (Fig. 8 ) and mole fractions are presented in Figs. 8 and 9 .The results acquired after computing ten subsequent generations brought a substantial improvement. Both, in the unification of the temperature field and methane conversion rate, when compared to the results for the initial generation. The fitness of the best specimen improved to 0.68.The overall distribution of fitness values in the 10 th generation is presented in Fig. 10 . The specimens begin to be focused in the pursued region of the search space, what can be noticed in Fig. 10. The temperature distribution acquired for the most fit specimen in this generation is provided in Fig. 11 b). Its methane conversion rate reached 0.56 and the exact changes of mole fractions in the longitudinal direction are presented in Fig. 12 .The 30th generation brought only a slight improvement, considering the optimal solution. The fitness values acquired for the specimens became to be focused in a very narrow region of the search space (Fig. 13 ). Therefore, the algorithm has ended global exploration and started to search for the optimum locally.The fitness value acquired for the best solution contained in the 30th generation is equal to 0.74. However, no considerable improvement has been noted, considering the thermal conditions (Fig. 14 ). Therefore, the reason of improvement of the fitness value should be sought in the methane conversion rate. As expected, it rose to 0.61. The exact change of the mole fractions in the longitudinal direction of the fittest reactor is presented in Fig. 15 .The final set of solutions characterizes itself with an improvement of the methane conversion rate. The acquired thermal conditions are similar to the ones in the 30th generation. The overall distribution of fitness values for the 50th generation is presented in Fig. 16 . The best solution reached a fitness of 0.78. Its temperature distribution is provided in Fig. 17 b). Like for the 30th generation, only f CH 4 has improved and the value of f T was maintained (Fig. 16). The mole fractions distribution is presented in Fig. 18 . The CH4 conversion rate is equal to 0.64, setting around 80 % of the value acquired for the reference case.The observed decrease of f CH4 is directly connected with the reduction of the Ni/YSZ amount used. Apparently, the algorithm detected the region of the highest temperature decrease at the reactor's inlet. Thus, the amount of the catalytic material in that region was reduced (Fig. 19 b), to improve the thermal conditions. As it can be observed, the Ni/YSZ amount is increasing, closing to the reactor outlet. The algorithm did it on purpose, to maximize the CH4 conversion rate for the presented solution. The overall amount of catalytic material was strictly limited, resulting in ten out of thirty catalytic segments only (Fig. 19b).To summarize the algorithm's operation, the differences between subsequent specimens should be confronted. Radius averaged temperature distribution in the reactor is an adequate method of comparison between fittest solutions in particular generations. Following the results presented in Fig. 20 , a significant change in the temperature distribution can be noted. The temperature difference value was reduced to 23.7 K, from 44.8 K acquired for the reference case. However, the presence of peaks in the temperature values may induce catalyst degradation, just like for the reference case. The observed peaks may be a consequence of a single segment's dimensions. Possibly, narrower segments would result in smoothing the noticed peaks, although their future manufacturing would be more challenging.The presented analysis aimed to improve the methane/steam reforming process, through the optimization of the thermal conditions inside a reforming reactor. The macro-patterning concept was introduced, as it appears to be a valid strategy for the improvement of the reforming process. The concept's principle is to divide the reformer's tube into separate segments, which are further filled with non-catalytic metallic foam or Ni/YSZ. The optimization was conducted using a genetic algorithm. The GA altered the segments composition and their porosity, to maximize the CH4 conversion rate and minimize the temperature gradients.Implementation of the morphological properties of the metallic foams allowed to check the influence of their introduction on the heat and mass transfer in the reforming reactor. The generated DMRs brought information about flow characteristics and made it possible to define how much the process has improved precisely.After the calculation of fifty generations, a solution with improved thermal conditions has been acquired. The algorithm decided to use only 33 % of the reference amount of the catalytic material. Despite a significant reduction of the Ni/YSZ present, the CH4 conversion fell only by 22 %. Simultaneously confirming that the effectiveness of the reforming process can be elevated by improvement of the thermal conditions inside the reactor.Considering the catalytic segments' parameters, the acquired solution might be suspected to be a local optimum only. The Ni/YSZ segments have similar parameters, which propagated since the GA's early operation. Therefore, the algorithm requires further development, to ensure that the global optimum was found. The mechanism of computing multiobjective fitness has to be revised, as the weighted sum approach appears to be inferior. Other possible improvements have to be sought in the introduction of an adaptive mutation rate, the enhancement of a crossover procedure and a fitness calculation. Moreover, changing the thermal fitness function to optimize the temperature gradients might alleviate the elimination of peaks in the average temperature profile.The following is the Supplementary data to this article: Multimedia component Multimedia component Supplementary data to this article can be found online at https://doi.org/10.1016/j.ijhydene.2020.02.228.

The presented research focuses on an optimization design of a catalyst distribution inside a small-scale methane/steam reforming reactor. A genetic algorithm was used for the multiobjective optimization, which included the search for an optimum of methane conversion rate and a minimum of the difference between highest and lowest temperatures in the reactor. For the sake of computational time, the maximal number of the segment with different catalyst densities was set to be thirty in this study. During the entire optimization process, every part of the reactor could be filled, either with a catalyst material or non-catalytic metallic foam. In both cases, the porosity and pore size was also specified. The impact of the porosity and pore size on the active reaction surface and permeability was incorporated using graph theory and three-dimensional digital material representation. Calculations start with the generation of a random set of possible reactors, each with a different catalyst distribution. The algorithm calls reforming simulation over each of the reactors, and after obtaining concentration and temperature fields, the algorithms calculated fitness function. The properties of the best reactors are combined to generate a new population of solutions. The procedure is repeated, and after meeting the coverage criteria, the optimal catalyst distribution was proposed. The paper is summarized with the optimal catalyst distribution for the given size and working conditions of the system.

This study did not generate any datasets.Catalysis is a key process in science that allows to control all kinds of chemical transformations. In the presence of a suitable catalytic material the reaction rate can be dramatically increased, which enables the optimal use of resources, increasing the yield of desired products and at the same time avoiding waste formation as well as reduce specific energy requirements. 1–10 Nowadays, 90% of all modern processes in the chemical industry apply catalytic technologies. 7–10 In addition to this crucial role in chemical sciences, catalysts also provide the basis of innovation for many other industries based on life and material sciences as well as energy technologies. Thus, new catalytic materials including molecularly defined and nanostructured systems are continuously prepared by scientists all over the world and tested for all kinds of transformations. 1–10 Regarding the potential new catalysts, in particular, 3D-metal-based systems are gaining increasing importance and provide the basis for an advanced and sustainable chemical synthesis. 11–17 Due to the inherent beneficial aspects such as stability, recycling, and reusability, heterogeneous nanostructured materials, especially, are of prime importance. 17–20 Typically, for a specific benchmark reaction or more importantly for a given industrial process the “best” catalyst is desired. Especially, for the bulk chemical industry it is important to apply state-of-the-art catalysts with optimal activity (TOF, turnover frequency), productivity (TON, turnover number), and selectivity to be cost competitive on a global scale. 21 However, apart from such highly optimized systems, there is also significant interest in catalysts, which can be applied in a general way for various processes. This is especially true for applications in organic synthesis, for drug discovery, and for basic sciences. Here, it is a common practice to develop new catalysts only for one specific synthetic methodology and the generality of a given catalyst is measured by its robustness toward different reaction conditions, but especially by its functional group tolerance and a wide substrate scope. Considering that elementary steps of many chemical processes are similar, we believe that “general” catalysts can be developed more efficiently by not only focusing on one specific transformation. As an example, in the oxidation of alcohols diverse compounds B–F can be formed. In general, alcohol (A) is oxidized to the corresponding aldehyde (B) first, which then can react with different nucleophiles such as H2O, alcohol, and ammonia to generate either geminal diol (X), hemiacetal (X), hemiaminal (X), or primary imine (Y), respectively, as intermediates (X and Y). All these intermediates might be further oxidized to produce the corresponding acid (C), ester (D), primary amide (F), and/or nitrile (E), respectively (Figure 1 ). 22–28 Looking at the individual steps of Figure 1, clearly the conversion of A to B and X to C, D, and F are mechanistically related and indeed can be performed with similar type of catalysts. However, traditionally each of these methodologies is studied separately using different catalyst systems, which is time and resource consuming.Among the many kinds of chemicals, functionalized aromatic and heterocyclic compounds are most valuable, which provide the basis for countless products of our daily life. In fact, synthetic organic chemistry and drug discovery majorly rely on the valorization of such compounds. 29–31 Among these, (hetero)aromatic carbonyl compounds (B), carboxylic acids (C), esters (D), nitriles (E), and amides (F) represent valuable fine and bulk chemicals widely used in research laboratories and industries. 32–38 Notably, these compounds can be easily functionalized/upgraded. Hence, they serve as precursors and intermediates for the synthesis of advanced chemicals, pharmaceuticals, agrochemicals, biomolecules, and materials. Moreover, many life science molecules, natural products, fragrances, and cosmetics as well as other daily life products contain, –CHO, –C=O, –COOH, –COOR, –CN, and –CONH2 functionalities, which play vital roles in their physical properties and functions.In general, products B–F can be conveniently accessed by oxidation of benzylic alcohols and related heteroaromatic compounds, 39–102 which are broadly commercially available. As an example, more than >200 benzylic alcohols are available from Sigma-Aldrich. 103 Regarding potential oxidants, air is ideal because it is abundant, inexpensive, and green, and it produces only water as by-product. 104 Favorably, air is much safer and more convenient to use than dioxygen. To perform the oxidation of alcohols using molecular oxygen or air to produce B–F, both homogeneous and heterogeneous catalysts based on precious and non-precious metals were developed in the past (Figure 2 ). 39–102 Despite these achievements, until now, there is no single general catalyst developed or applied for the oxidative conversion of alcohols to synthesize carbonyl compounds (aldehydes and ketones; B), carboxylic acids (C), esters (D), nitriles (E), and amides (F).In this regard, here, we show that a general catalyst development can be achieved efficiently by directly including different related benchmark reactions and parallel testing of the catalyst materials under investigation. Following the presented strategy, we demonstrate that it is possible to develop graphitic-shell-encapsulated cobalt nanoparticles as a “most general” oxidation catalyst, which can not only be applied in one of the above-mentioned aerobic oxidation reactions but many related transformations. The highly stable and reusable catalyst allows for the synthesis of functionalized and structurally diverse aromatic and heterocyclic aldehydes, ketones, carboxylic acids, esters, nitriles, and primary amides in good to excellent yields.In the past decade, we prepared a variety of nanostructured 3D-metal (Fe, Co, Ni, and Cu)-based materials by immobilization of either organometallic complexes or metal organic frameworks on inorganic supports and subsequent pyrolysis under inert atmosphere. 105–108 Some of these materials proved to be highly active and selective for catalytic hydrogenations, oxidations, and reductive amination reactions.105–108 A typical feature of these active catalysts is the core-shell structure of the metal nanoparticles, which are embedded in graphene or graphitic layers. 105–108 To obtain this specific structure, ligated metal complexes have been used as precursors. 105–108 To be cost-efficient, the respective ligands should be as simple, abundant, and inexpensive as possible. In this respect, amines and carboxylic acids are interesting as a plethora of them is easily accessible.In continuation of our previous work, 105–108 we started to prepare a library of supported 3D-metal nanoparticles using Co, Mn, Fe, and Cu salts with piperazine (PZ) and DL-tartaric acid (TA) as ligands, which will form metal coordination polymers or metal organic frameworks. As an example, Co(NO3)2·6H2O was dissolved in DMF, and then this mixture was heated to 150°C. At this temperature, PZ and TA were added and stirring was continued for 30 min. After addition of the support (carbon; Vulcan XC72R) and 4 h of additional stirring, the solvent was removed, and the resulting dark solid material was grinded and pyrolyzed at different temperatures (400°C–1,000°C) under argon atmosphere for 2 h to provide the desired cobalt-based nanoparticles supported on carbon (Figure 3 ). Similarly, other 3D-metal nitrates (Fe(NO3)3·9H2O, Mn(NO3)2·6H2O, and Cu(NO3)2·3H2O) were applied following the same procedure. For comparison, metal salts without ligands were pyrolyzed on carbon, and Ru- as well as Pd-containing materials were made using PZ and TA ligands, too.Following our concept to develop a universal oxidation catalyst, we evaluated the generality and applicability of the prepared materials not only for one type of reaction, but five different aerobic oxidation reactions were chosen. More specifically, all potential catalysts as well as selected commercial ones were tested for their activities in the conversion of benzyl alcohol (A1) to benzaldehyde (B1), benzoic acid (C1), methyl benzoate (D1), benzonitrile (E1), and benzamide (F1) (Figure 4 ). In general, all these benchmark reactions were performed in the presence of air (1 bar or 10 bar) at 55°C–120°C using either alcohols, water, or heptane as solvent. Interestingly, aldehyde and ester formation are observed at ambient pressure and low temperature, while the formation of acid, amide, and nitrile proceeded at temperatures >100°C and 10 bar of air vide infra.First, we tested in a parallel manner, the materials prepared by the pyrolysis of Fe-, Mn-, Co-, and Cu-nitrates on carbon (Fe(NO3)3@C-800, Mn(NO3)2@C-800, Co(NO3)2@C-800, and Cu(NO3)2@C-800) (Figure 4). All these materials exhibited no or poor activities for all the benchmark reactions (<16% yields of the corresponding products B1–F1). Next, we tested catalysts prepared by the impregnation and pyrolysis of PZ- and TA-ligated metal complexes (Fe-PZ-TA@C-800, Mn-PZ-TA@C-800, Co-PZ-TA@C-800, and Cu-PZ-TA@C-800) (Figure 4). Among these materials Fe-PZ-TA@C-800 was completely inactive for the formation of benzoic acid and methyl benzoate, whereas it showed low to moderate activity for the synthesis of benzaldehyde, benzonitrile, and benzamide in 16%, 20%, and 60%, respectively. Mn-PZ-TA@C-800 was even more specific producing only 20% of B1, while no or very little activity is observed in the other model reactions. Interestingly, Co-PZ-TA@C-800 exhibited remarkable activity and selectivity in all the benchmark reactions and produced almost quantitative of yields (>98%) of benzaldehyde, benzoic acid, methyl benzoate, benzonitrile, and benzamide. Finally, Cu-PZ-TA@C-800 was tested and showed no activity for the formation of C1 and D1 as well as very low activity for B1 formation (30%). However, this material was found to be efficient for the preparation of benzonitrile (98%) and benzamide (97%). Because of the unique behavior of the cobalt-based material, variation of the pyrolysis temperature of the templated Co-PZ-TA@C was performed. However, materials prepared by pyrolysis at 400°C, 600°C, and 1,000°C showed lower activity. Similarly, pyrolysis of cobalt-complexes with single ligands either PZ or TA (Co-PZ@C-800 or Co-TA@C-800) gave less active materials and provided the desired products B1–F1 in 50%–68% yields. Using the Fe, Mn, or Co salts in the absence and presence of PZ and TA under homogeneous conditions exhibited no or minor activity in all five benchmark tests (<5%) (Table S2). Likewise, the non-pyrolyzed supported pre-catalysts (metal-PZ-TA@C) behave. However, in the presence of the homogeneous Cu-PZ-TA system and its supported derivative some activity for the formation of benzaldehyde (10%–15%) is observed (Table S2).To compare the activities and selectivities of the optimal system (Co-PZ-TA@C-800) with commercially available precious-metal-based catalysts, Ru/C and Pd/C were also applied in the benchmark reactions (Figure 4). Under similar conditions, Ru/C showed no activity for alcohol to ester oxidation, and in all other cases, product yields were lower compared with Co-PZ-TA@C-800, while Pd/C exhibited only high activity for the preparation of benzaldehyde. Likewise, Ru-PZ-TA@C-800 and Pd-PZ-TA@C-800, exhibited moderate to low activity for most reactions. Thus, among all the tested materials Co-PZ-TA@C-800 was found to be the most general oxidation catalyst, which allows for diverse aerobic oxidations of benzyl alcohols to produce a variety of product classes in a selective manner.To demonstrate the stability, recycling, and reusability of this Co-material (Co-PZ-TA@C-800), the synthesis of benzonitrile from benzyl alcohol in presence of aqueous ammonia and air was performed for seven times under standard conditions. Notably, in the presence of ammonia supported nanoparticles easily encounter stability and reusability problems. Nevertheless, as shown in Figure 5 , Co-PZ-TA@C-800 was stable and is conveniently recycled and reused up to 7th run.To know the structural features and to understand the catalytic activities, we carried out detailed characterizations of the most active (Co-PZ-TA@C-800), moderately active (Co-PZ@C-800), (Co-TA@C-800), and less active (Co(NO3)2@C-800) materials using X-ray powder diffraction (XRD), scanning transmission electron microscopy (STEM) with electron energy loss spectroscopy (EELS), and X-ray photoelectron spectroscopy (XPS). The XRD patterns of the most active catalyst, Co-PZ-TA@C-800, showed the presence of mainly metallic cobalt particles (Figure S1), while the moderately active catalysts Co-PZ@C-800 and Co-TA@C-800 contained a mixture of metallic cobalt and oxidic cobalt (Co3O4) particles (Figure S1). STEM analysis of Co-PZ-TA@C-800 proved the formation of metallic cobalt particles with different sizes ranging from 1 to 7 nm and from 25 to 40 nm (Figure 6 A). However, some bigger particles with sizes up to 80 nm were also observed. The smaller particles are usually found in groups, while other areas of the material contained no cobalt. Interestingly, most of the particles in this material are surrounded by few layers of graphitic carbon (Figure 6A, right image). In addition to metallic cobalt, the presence of a very small amount of cobalt oxide is observed (Figures 7 A, 7B, and S2). Co-PZ@C-800 contained also metallic and oxidic cobalt; however, the presence of the oxide seems to be more than in Co-PZ-TA@C-800, and it can be found either on the surface or as partially oxidized particles (see e.g., the biggest particle in the left image in Figures 6B and S3).The sizes of these particles are in the range between 25 and 60 nm with few particles being bigger up to 100 nm and fewer below 25 nm compared with Co-PZ-TA@C-800. In the case of metallic cobalt, these particles are covered by graphitic layers (Figure 6B, right image). Likewise, Co-TA@C-800 showed the presence of both metallic and oxidic cobalt particles with sizes of 15–50 nm and only very few below this size. Similar to Co-PZ-TA@C-800, the nanoparticles of metallic cobalt are surrounded by graphitic layers in both Co-PZ@C-800 and Co-TA@C-800 (Figure 6C). The least active material, cobalt nitrate@C-800, contained completely Co3O4 particles, which are not surrounded by graphitic layers (Figure S4). The material obtained after three reaction cycles using the active catalyst Co-PZ-TA@C-800 showed that there is not much difference in the structure compared with the fresh catalyst (Figure 6D). In this reused material, metallic nanoparticles with sizes of 3–10 and 25–40 nm are observed, which are in few cases partially oxidized at the surface. However, analysis of the material after 7 reaction cycles revealed that cobalt is oxidized in more proportion (Figure S5). This implies that the cobalt is successively oxidized during the reaction cycles. EELS was applied to analyze the elemental composition of a selected area in the most active material, Co-PZ-TA@C-800 (Figures 7A and 7B). Analysis of the edge features of the elements enables the visualization of the spatial distribution of the corresponding elements (C, N, O, and Co) in a single-color elemental map as shown in Figure 7 (right). As can be seen there the support mainly consists of carbon (Figure 7, red map, C-K edge) and some content of nitrogen, which is originated from the ligands (Figure 7, green map, N-K edge). Inspection of the distribution of the Co-L edge signal and the O-K edge signal (Figure 7, yellow and blue map, respectively) reveal that the two bigger particles consist of a metallic cobalt core and a shell of cobalt oxide. Two selected spectra that show different features of selected areas in the material are shown in Figure 7B.To obtain further insights into the surface chemistry of these materials, we performed XPS analysis. The sample surfaces of all the four catalysts (Co-PZ-TA@C-800, Co-PZ@C-800, Co-TA@C-800, and Co-PZ-TA@C-800 recycled) consists mainly of C with small concentrations of Co, O, N, S, and Si (Table S1) with the last two probably originating from Vulcan XC-72R and N from the starting chemicals such as ligands and cobalt nitrate. As found by STEM the Co particles are surrounded by carbon layers leading to the very low surface concentrations of Co between 0.2 and 0.4 atom% for the fresh catalysts and 0.9 atom% for the recycled catalysts. The high-resolution Co 2p spectra of all four samples (see Figure 8 A) confirmed the presence of metallic Co as sharp peaks at 778.7 (Co 2p3/2) and 793.8 eV (Co 2p1/2) as well as oxidic structures as broad peaks at higher binding energies. 109 Considering the satellite features at around 786 and 803 eV, a mixture of CoO and Co3O4 seems to be present. Looking at the recycled catalyst (three reaction cycles) Co-PZ-TA@C-800R (see Figure 8A) an oxidation of the surface can be observed so that only a minor part Co is still in the metallic state. Note that the Co concentration at the surface increases to 0.9 atom% (Table S1) in the recycled catalyst (three reaction cycles), which indicates a partial breakup of the protective carbon shell probably also leading to the observed oxidation during the use of the catalyst. In case of the recycled catalyst after 7th run, only oxidized Co is observed on the surface (Figure S6).The N 1s spectra in Figure 8B are fitted with four peaks which can be assigned to pyridinic-N at binding energies around 398.8 eV, pyrollic-N, and/or N bonded to a metal in a Me–Nx center (∼400.1 eV), graphitic N (∼401.3 eV) as well as oxidized pyridinic-N (∼404 eV). 110 Interestingly the concentration of N is higher in Co-PZ-TA@C-800 (1.7 atom %) compared with the other fresh catalysts (0.6 and 0.7 atom%; Table S1). After recycling the N concentration becomes even higher (4.7 atom%) and is dominated by pyridinic and pyrollic-N/Me–Nx. This is explained by ammonia side-reactions on the catalyst surface.All these characterization data revealed that the immobilization and pyrolysis of cobalt-complexes containing PZ and/or TA ligands produced dissimilar kinds of cobalt nanoparticles supported on carbon, which in turn revealed varying catalytic activities. The material (Co-PZ-TA@C-800) containing predominately metallic cobalt nanoparticles exhibited highest activity. Apparently, fully oxidized cobalt has a negative impact on the overall catalytic performance as such particles have not been observed in the most active catalyst, and even not in the recycled one. Catalytic performance likely depends on the particle nature, sizes, and their distribution. The combination of PZ and TA ligands seems to favor the formation of a higher share of smaller cobalt containing particles and thus induce an increased number of accessible active sites in the catalyst, Co-PZ-TA@C-800. Hereafter, we represent the most active catalyst Co-PZ-TA@C-800 as Co/GS@C, where GS denote graphitic shell.After having a general catalyst system Co/GS@C (Co-PZ-TA@C-800) in hand, we performed additional tests with >90 different alcohols. As shown in Figures 9, 10, 11, and 12 , simple substituted as well as functionalized and structurally diverse aromatic and heterocyclic aldehydes, ketones, acids, esters, nitriles, and amides can be prepared in good to excellent yields. For example, alkyl- and phenyl-substituted alcohols produced the corresponding products B–F in up to 98% yield (Figure 9; products B2, B6, B7, B11, C2, C3, C8–C10, D2–D5, E2–E4, F2–F4, and F8).Similarly, fluoro- and thio-trifluoromethyl-substituted products were obtained yields in up to 96% (Figure 9; products B3–B5, B12, B13, C4–C7, D6, D7, E5–E7, and F5). Such compounds are interesting building blocks for the discovery of new pharmaceuticals and agrochemicals. 111 In addition to benzyl alcohols, related condensed arenes gave corresponding products in up to 89% yields (Figure 9; products B8, B9, D8, D9, E8–E10, F6, and F7). Likewise, benzophenone, and 1-phenylbutan-1-one were obtained in 83%–85% yields (Figure 9; products B14–B15). Notably, in the oxidative cross-esterification reaction apart from methanol, other aliphatic alcohols can be used to provide ethyl, propyl, iso-propyl, butyl, and hexyl benzoates in up to 90% yields (Figure 9; products D10–D14). Interestingly, in case of ammoxidation to give nitriles, Co-PZ-TA@C-800 showed good activity for aliphatic alcohols at elevated temperature (140°C). As a result, 4-phenylbutanenitrile, and several alkyl nitriles were obtained in up to 78% yield (Figure 9; products E12–E15).Next, the ability and selectivity of Co-PZ-TA@C-800 for the refinement of more complex molecules as well as the tolerance of functional and sensitive groups was studied. Thus, functionalized as well as multi-substituted benzylic alcohols were subjected to aerobic oxidation under the optimized conditions (Figure 10). Chloro-, bromo-, and iodo-substituted benzylic alcohols smoothly reacted to the corresponding halogenated benzaldehydes, acetophenones, benzoic acids, methyl benzoates, benzonitriles, and primary benzamides in good to excellent yields (Figure 10; products B16–B18, B27, B36–B38, C11–C14, C19, D15–D17, D25, E16–E19, F9–F13, and F18). These further functionalized halogenated molecules are indispensable for many applications and serve as valuable starting materials and intermediates. 112 As an example, 2,6-dichloro-benzyl alcohol was reacted in presence of ammonia in water and produced the corresponding benzamide in 85% yield (F18). Substrates containing ether, hydroxyl, amine, nitro, ester, boronic ester, or nitrile substituents, were selectively converted to desired products B19, B20, B23, B26–B33, B39, B40, C15, C16, C20, D18, D19, D21–D25, E21, E23, E27, F14, F15, and F17. Interestingly, sulfur-containing alcohols were also selectively converted without oxidation of S-moiety (Figure 10; products B24, B25, C17, D20, E25, E26, and F19). In case of 1,3- and 1,4-benzenedimethanol, both CH2–OH groups were selectively oxidized and produced terephthalaldehyde B30 and terephthalonitrile E24 in 95%–96% yields. In addition, di- and multi-substituted substrates, which possess additional challenges, were efficiently oxidized, and gave products B21, B22, B26–B29, C18, C19, D23, D25, E20, E22, and F16–F18 in high yields (Figure 10). Even the dinitro-substituted benzyl alcohol produced the corresponding benzaldehyde B28 in 85% yield. Sterically hindered tri-methyl benzyl alcohol also reacted to provide the corresponding benzonitrile in 86% yield (Figure 10; product E22). In addition to benzylic alcohols, allylic alcohols such as cinnamyl and perillyl alcohols can be efficiently transformed to cinnamaldehyde, perillyl aldehyde, and cinnamyl nitrile (Figure 10; products B34, B35, E28). Furthermore, aliphatic cyclic secondary alcohols were oxidized to produce cyclic ketones (Figure 10; products B43 and B44).Subsequently, the synthesis of heterocyclic carbonyl compounds, carboxylic acids, esters, nitriles, and amides from corresponding alcohols was explored. In general, heterocyclic compounds find wide range of applications, especially in life sciences. Indeed, such scaffolds are ubiquitous in pharmaceuticals, natural products, agrochemical, and other biomolecules. Thus, they play a pivotal role in modern small molecule drug discovery processes. 113 , 114 As shown in Figure 11, different kinds of heterocyclic alcohols were oxidized to give the desired compounds. Interestingly, nicotinic derivatives such as nicotinaldehyde, methyl nicotinate, nicotinonitrile, nicotinic acid (niacin), and nicotinamide—the latter two are used as food supplement and nutrition medications—as well as 3-acetylpyridine are smoothly prepared from 3-pyridinemethanol in up to 97% yield (Figure 11; products B45, B61, C21, D26, E29, F20). Similarly, bromo-, di-methoxy-, and di-chloro-substituted 2- and 3-pyridinemethanol are selectively oxidized to produce B47–B49 and D28. Other N-heterocycles such 2-pyrazine and quinolinemethanol are well accepted and provided the respective products in 88%–94% (products E30, E31, and F21). Interestingly, 2-thiophenmethanol also allowed for selective oxidation in up to 95% yield (Figure 11; products B50, B62, D29, E41, and F22). At this point, it should be noted that sulfur-containing compounds constitute common poisons for most heterogeneous catalysts. However, Co-PZ-TA@C-800 tolerated the presence of many sulfur-containing molecules and a variety of sulfur-containing products, e.g., B24, B25, B50, B52, B53, B62, C17, C22, D20, D29, E25, E26, E33, E34, E41, F19, and F22 were obtained in good to excellent yields. Apart from the oxidation of hydroxymethyl-substituted N-, O-, and S-heterocycles a variety of benzylic alcohol containing heterocyclic motifs such as thiazole, morpholine, pyrazine, tetrahydropyran, diazepane, N-methyl diazepane, and triazole underwent aerobic oxidation under the previously optimized conditions and furnished the desired products (Figure 11; B51–B53, B55–B60, and E33–E40). As an example, 2,1,3-benzothiadiazol-5-yl-methanol gave corresponding aldehyde and nitrile (Figure 11; products B52, E34. Other notable examples include the oxidation of 3,4-(methylendioxy)-benzylalcohol, an important motif present in drugs and natural products (Figure 11, products B51, C23, D30, E32, F23) and 3,5-dimethyl-1-phenyl-1H-pyrazol-4-yl-methanol as well as 2-(2-morpholinoethoxy)phenyl-methanol (Figure 11, products B54, B60, E40, E42).In recent years, the valorization of hydroxymethylfurfural (HMF, A2) and furfuryl alcohol (A3) attracted significant interest for the preparation of sustainable polymers and fuels (Figure 12). 115 , 116 Among these, the synthesis of 2,5-furandicarboxylic acid (FDCA) and dimethyl furan-2,5-dicarboxylate (FDCM) from HMF is of actual interest to produce poly(ethylenefuranoate) (PEF) polymer. 115 Applying our Co/GS@C catalyst FDCM, (D31) is prepared in up to 85% yield. More sensitive furan-2,5-dicarbaldehyde (B66) can be also obtained in up to 87% yield. Furthermore, furan-2,5-dicarbonitrile (E43) is available from this latter intermediate. Similarly, furfuryl alcohol (A3) was selectively transformed to corresponding aldehyde (B67), carboxylic acid (C24), methyl ester (D32), nitrile (E44), and amide (F24) in good to excellent yields (Figure 12). Again, these products have various interesting applications, for example, 2-furoic acid is a known preservative, flavoring ingredient, food, and color additive in food, 117 while 2-furonitrile has been suggested as a potential sweetening agent, which has about thirty times the sweetening power of sucrose. 118 In general, catalytic oxidations were performed in 50–150 mg scale with respect to substrate. To demonstrate the utility of this catalyst system, reactions of five alcohols were also performed on 1–10 g (Figure 13 ). The yields of the desired products from these upscaling experiments were similar to those obtained from the smaller scale.Further, we calculated TONs and TOFs of our Co-catalyst for the oxidation of benzyl alcohol to benzaldehyde (Table S3). Under standard conditions (0.5 mmol alcohol, 35 mg catalyst, 80°C, 24 h) these values are found to be 15.6 and 0.65 h−1, while at 100°C and increased amount of substrate (2.5 mmol of benzyl alcohol, 35 mg catalyst, 24 h) both values increased (TOF and TON are 46.8 and 1.95 h−1). These numbers are at least comparable to reported non-noble metal-based catalysts for the individual transformations (Table S3).We performed kinetic investigations on the Co/GS@C-catalyzed oxidation of benzyl alcohol to benzaldehyde and examined the effect of (1) reaction time, (2) reaction temperature, (3) catalyst amount, and (4) substrate (benzyl alcohol) concentration (Figure S7). By increasing the time, temperature, or catalyst loading the yield of benzaldehyde increased, and quantitative yield was obtained for 24 h, at 80°C with 35 mg of catalyst. On the other hand, increasing the substrate (benzyl alcohol) concentration, the yield of benzaldehyde is decreased. Next, we calculated the reaction order with respect to substrate (benzyl alcohol), which is found to be −0.85 (Figure S7E). This also confirmed that the substrate has a negative effect on the rate of the reaction.Next, we conducted experiments to identify the formation of possible reactive oxygen species (ROS) during the Co/GS@C-catalyzed aerobic oxidation reactions. For this purpose, under standard conditions, the oxidation of benzyl alcohol to benzaldehyde was tested in the presence of different radical quenchers/trapping agents such as NaN3, i-PrOH, and p-benzoquinone (PBQ) (Table S4). All these reagents have been used to trap singlet oxygen (1O2), hydroxyl (⋅OH) or super oxide (O2 ⋅−) radicals, which are considered as the ROS in aerobic oxidations. These experiments showed that there is no effect after adding i-PrOH or NaN3 on the reactions. However, the reaction is inhibited after the addition of 80 mg PBQ. This makes the formation of super oxide (O2 ⋅−) species likely. In addition, we performed an experiment for trapping super oxide (O2 ⋅−) species using butylated hydroxytoluene (BHT) (Figure S8). Under similar experimental conditions, without the substrate (35 mg Co/GS@C, 0.5 mmol BHT, 1 bar air, 10 mol % K2CO3, 2 mL n-heptane, 80°C, 24 h), we performed the reaction with BHT and observed the formation of BHT-OOH, which is detected by GC-MS (Figure S8). These experiments indicate that super oxide (O2 ⋅−) is formed during the reaction.Further to prove the formation of a superoxide radical intermediate, EPR spin-trapping studies using 5,5-dimethyl-1-pyrroline N-oxide (DMPO) as spin-trap reagent were performed. The EPR spectrum of the reaction mixture containing a suspension of Co/GS@C catalyst, Cs2CO3 and benzyl alcohol in heptane after heating at 80°C for 3 min under bubbling of O2 followed by addition of DMPO exhibited a signal at g = 2.006 characteristic of the DMPO-OOH spin adduct indicating again the formation of a superoxide radical intermediate during the catalytic reaction (Figure 14 ). It should be noted that no EPR signal is detected in the absence of benzyl alcohol suggesting that its adsorption on the surface of the Co/GS@C-800 catalyst induce the activation of molecular oxygen and superoxide formation.Regarding the general mechanism, in all these oxidations, the first step is the Co/GS@C-catalyzed oxidative conversion of benzyl alcohol (A) to benzaldehyde (B). Thus, for the formation of carboxylic acid (C), ester (D), nitrile (E), and amide (F), (B) serves as the key intermediate (Figure S9). Indeed, for all transformations the formation of benzaldehyde was detected by GC-MS. In case of benzoic acid, the aldehyde reacts with water and generates benzaldehyde hydrate (X) as another intermediate, which is then oxidized in the presence of Co/GS@C and gives the corresponding acid. Similarly, in the formation of benzoic acid esters, aldehyde reacts with another alcohol and provides hemiacetal (X / ) as another intermediate, which finally converts to the corresponding ester in presence of Co/GS@C and air. For the formation of nitrile, benzaldehyde couples with ammonia and generates primary imine (Y), which finally yields the corresponding nitrile. In case of amide synthesis, two pathways are possible: (1) the aldehyde can react with ammonia to form hemiaminal as the intermediate, which could be then oxidized to give the corresponding primary amide or (2) formation of benzonitrile takes place, which can undergo hydrolysis to form the primary amide. To prove these two pathways, we performed the reaction of benzonitrile in water in presence of ammonia and air using Co/GS@C catalyst at 120°C for 24 h. From this experiment, we obtained 98% of benzamide (Figure S10). Thus, we conclude the formation of amide occurred mainly by the hydrolysis of benzonitrile. It should be noted that the intermediates, aldehyde hydrate (X), hemiacetal (X /), and primary imine (Y) are unstable, and we were not able to detect or isolate them.Based on the identified active oxygen species and the proposed reaction pathways and intermediates, we suggest the following general mechanism for the different aerobic oxidations of primary alcohol in the presence of Co/GS@C (Figure 15 ). In the first step, (1) adsorption and activation of alcohol and oxygen takes place on the catalyst surface. During this process, the generation of the observed superoxide species occurs. In the next step (2), oxidation of the activated alcohol takes place. In the last step (3), the desorption of the product, aldehyde takes place by the regeneration of catalyst. Similar catalytic cycles for the formation of esters, carboxylic acid, and nitrile are proposed. The hydrolysis of benzonitrile to benzamide occurs best in presence of catalyst, water, ammonia, and air.In conclusion, we demonstrate that a new catalyst can be efficiently developed not only for one specific synthetic transformation but also for related methodologies with similar elementary reaction steps. In particular, we show that the here presented cobalt catalyst is able to perform the selective aerobic oxidation of alcohols to a variety of functionalized aromatic products. This catalyst is based on carbon-supported graphitic-shell-encapsulated specific cobalt nanoparticles, which are prepared by immobilization of in-situ-generated cobalt-PZ-TA template on carbon and subsequent pyrolysis under argon at 800°C. Applying the optimal material, functionalized and structurally diverse (hetero)aromatic aldehydes, ketones, carboxylic acids, esters nitriles, and primary amides were prepared in selective manner from alcohols in the presence of air. The resulting compounds represent valuable fine and bulk chemicals, which serve as key starting materials and intermediates for the synthesis of advanced chemicals, pharmaceuticals, agrochemicals, and materials. We believe that the presented concept is not only valid for the here-described case of alcohol oxidations but offers manifold opportunities for other chemical transformations, too.Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Matthias Beller ([email protected]).All materials generated in this study are available from the lead contact without restriction.We gratefully acknowledge the European Research Council (EU project 670986-NoNaCat) and the State of Mecklenburg-Vorpommern for financial and general support. We thank the analytical team of the Leibniz-Institut für Katalyse e.V. for their excellent service.R.V.J. and M.B. supervised the project. T.S., R.V.J., and M.B. planned and developed the project. T.S. prepared catalysts and performed catalytic experiments. V.G.C. performed catalytic experiments and reproduced the results. N.R. performed TEM measurements and analysis. J.R. conducted EPR measurements. S.B. performed XPS measurements and analysis. R.V.J., T.S., M.B., and V.G.C. wrote the paper.The authors declare no competing interests.Supplemental information can be found online at https://doi.org/10.1016/j.chempr.2021.12.001. Document S1. Figures S1–S224, Tables S1–S4, supplemental experimental procedures, and supplemental references Document S2. Article plus supplemental information

Functionalized (hetero)aromatic compounds are indispensable chemicals widely used in basic and applied sciences. Among these, especially aromatic aldehydes, ketones, carboxylic acids, esters, nitriles, and amides represent valuable fine and bulk chemicals, which are used in chemical, pharmaceutical, agrochemical, and material industries. For their synthesis, catalytic aerobic oxidation of alcohols constitutes a green, sustainable, and cost-effective process, which should ideally make use of active and selective 3D metals. Here, we report the preparation of graphitic layers encapsulated in Co-nanoparticles by pyrolysis of cobalt-piperazine-tartaric acid complex on carbon as a most general oxidation catalyst. This unique material allows for the synthesis of simple, functionalized, and structurally diverse (hetero)aromatic aldehydes, ketones, carboxylic acids, esters, nitriles, and amides from alcohols in excellent yields in the presence of air.

Hydrogen is hailed as an environmentally benign alternative to traditional fossil fuels because it has a high energy density and produces zero pollution [1,2]. However, to date, hydrogen is primarily produced by steam reforming of fossil resources. From an environmental and sustainability perspective, hydrogen production from electrochemical water splitting, which produces no carbon emissions, is a desirable method. In general, water splitting is divided into two half reactions: the hydrogen evolution reaction (HER) and the oxygen evolution reaction (OER) [3–6]. To date, Pt-based and Ir-based catalysts are still the most popular systems for the HER and OER, due to their low overpotentials and small Tafel slopes. Nevertheless, the limited supplies and high cost of precious metals make them impractical for large-scale application in electrocatalytic water splitting reactions [7].Recently, extensive studies have focused on exploiting cost-effective catalysts based on abundant transition metals (TMs), such as chalcogenides [8,9], phosphides [10,11], nitrides [12,13], and carbides [14,15]. However, these require a high operating voltage for water splitting due to their inferior anodic OER kinetics. To address this problem, highly efficient, energy-saving hydrogen production has been achieved by combining the HER with oxidation reactions that have low theoretical voltages and use small organic molecules, including methanol [16,17], hydrazine [18,19], urea [20,21], and 5-hydroxymethylfurfural [22,23]. The urea oxidation reaction (UOR) has an extremely low theoretical voltage of 0.37 V, would generate a 70% energy saving, and offers the potential to purify industrial and sanitary wastewater [24]. The UOR is a six-electron transfer reaction and is also limited by intrinsically sluggish kinetics (CO(NH2)2 + 6OH– → CO2 + N2 + 5H2O + 6e–). Therefore, exploring highly efficient bifunctional catalysts with high HER and UOR activity to achieve urea-assisted energy-saving hydrogen production remains a formidable task.Transition metal phosphides (TMPs) have been explored as intriguing electrocatalysts for water electrolysis by virtue of their low electrical resistance and similarity to hydrogenase [25]. TMPs (Ni2P, CoP, MoP, etc.) with various morphologies and structures have been fabricated and have yielded good performance. However, most of them exhibit only monofunctional catalytic activity. Among the various strategies for preparing excellent bifunctional catalysts for both oxidation and reduction reactions, constructing a synergistic interface with two different electrocatalysts is an effective method. Interfacial engineering can tailor the electronic environment, expose abundant active sites, promote electron transfer, and optimize the adsorption of reaction intermediates [26–29]. Experiments as well as theoretical calculations have shown that bimetal catalysts with abundant heterogeneous interfaces usually display better electrocatalytic performance than the corresponding single compound [30]. Wang et al. [31] reported a heterogeneous bimetallic Mo-NiPx/NiSx catalyst for robust overall water splitting that exhibited a low voltage of 1.42 V at 10 mA cm–2. Yu et al. [32] designed a heterogeneous bimetallic material, Ni2P-FeP, that proved to be a startling effective bifunctional catalyst for overall water splitting, revealing rich active sites and an improved transfer coefficient. Hence, the rational design of a hierarchical structure to engineer coupling interfaces is important for optimizing catalytic activity and creating high-performance bifunctional catalysts for the HER and UOR.Herein, we design and prepare a Ni2P/NiMoP nanosheet catalyst with a hierarchical architecture on a nickel foam (NF) substrate, using simple hydrothermal and phosphorization methods. The Ni2P/NiMoP electrode demonstrates excellent performance for both the HER and the UOR, requiring an overpotential of only 22 mV for the HER and a small working potential of 1.33 V for the UOR at a current density of 10 mA cm–2. Moreover, a two-electrode system using Ni2P/NiMoP as a bifunctional catalyst shows an ultralow cell working voltage of 1.35 V at 10 mA cm–2 and long-term durability for 80 h. Its intriguing activity can be ascribed to the engineered heterostructure interface, which leads to charge redistribution, regulates the electronic structure, and promotes electron transfer during the reactions.The procedure for synthesizing Ni2P/NiMoP nanosheets grown on NF is schematically illustrated in Fig. 1 a. First, the NiMoO4 precursor nanosheets were grown directly on NF through a simple hydrothermal method previously described [33]; these nanosheets were characterized using SEM and XRD (Figures S1 and S2). Next, uniform Ni2P/NiMoP nanosheets were prepared via a phosphorization process that left the original morphology undamaged (Fig. 1b–d). The resulting uniformly distributed nanoflake arrays on the NF substrate were beneficial for exposing active sites and for allowing electrolyte permeation and gas release [34]. The TEM images in Fig. 1e show the typical nanosheet structure. HRTEM images show 2.21 Å and 2.29 Å spacing between the lattice fringes, which correspond to the (111) planes of Ni2P and NiMoP (Fig. 1f and g), confirming the existence of a heterointerface structure between Ni2P and NiMoP. The heterogeneous interfaces may have modulated electron distribution and enhanced electron transfer, thereby exposing abundant active sites and optimizing the material's chemical adsorption capacity. For comparison, Ni2P was also prepared without the Mo precursor (Figures S3 and S4). The HRTEM image and energy-dispersive X-ray (EDX) elemental maps in Fig. 1h show that elemental Ni, Mo, and P were evenly distributed in the as-prepared samples, implying the successful fabrication of metal phosphides. Fig. 2 a presents the XRD pattern of the as-prepared Ni2P/NiMoP scraped from the NF substrate. The diffraction peaks are easily indexed to the Ni2P phase (PDF#03-0953) and the NiMoP phase (PDF#31-0873), indicating the successful synthesis of a Ni2P/NiMoP heterostructure. XPS was conducted to investigate the chemical composition and electronic states of the Ni2P/NiMoP and Ni2P samples. Comparison of the XPS survey spectra (Figure S5) confirmed the presence of elemental C, P, O, Mo, and Ni in the Ni2P/NiMoP. In the high-resolution Ni 2p spectrum (Fig. 2b), the two peaks centered at 852.9 and 870.0 eV are consistent with Ni-P bonds [31]. The peaks located at 856.3 and 874.2 eV are assigned to Ni-O bonds, arising from surface oxidation, while the remaining two signals are satellite peaks. In the P 2p spectrum (Fig. 2c), the two peaks at binding energies of 129.2 and 130.3 eV are compatible with P 2p3/2 and P 2p1/2, respectively, and the P-O species (133.8 eV) is due to surface oxidation [35]. In the Mo 3d spectrum (Fig. 2d), the peaks at 230.3 and 233.3 eV are ascribed to Mo 3d5/2 and Mo 3d3/2, respectively. It should be noted that the peaks of P 2p and Ni 2p for Ni2P/NiMoP are positively shifted compared with those for Ni2P, indicating charge redistribution due to the strong coupling interfaces between Ni2P and NiMoP, which could have promoted electrocatalytic activity. In addition, the higher valence state of Ni in Ni2P/NiMoP would have been conducive to optimizing the binding interaction between the catalyst and reaction intermediates [27,36,37].The HER performance of each as-prepared sample was recorded using a typical three-electrode setup in 1 M KOH electrolyte. We first investigated the effect of phosphorization temperature on the HER activity (Figure S6), finding the Ni2P/NiMoP catalyst prepared at 350 °C has the best catalytic activity. The linear sweep voltammograms (LSV) in Fig. 3 a indicate the Ni2P/NiMoP has notable electrocatalytic activity for the HER compared with bare NF, Ni2P, NiMoO4, and commercial Pt/C catalysts. As shown in Figs. 3b and S7, the Ni2P/NiMoP displayed an ultralow overpotential of 22 and 91 mV to yield current densities of 10 and 100 mA cm–2, which were much lower than those for NF (169 and 374 mV), Ni2P (92 and 189 mV), and NiMoO4 (144 and 318 mV). It also outperformed most TM-based HER materials (Table S1) and was comparable to or even higher than the benchmark catalyst, Pt/C (16 and 121 mV). Tafel slopes were calculated to investigate the reaction kinetics (Fig. 3c). Ni2P/NiMoP exhibited a low value of 34.5 mV dec–1, which was close to that of Pt/C and smaller than those of NF, Ni2P, and NiMoO4, indicating the fast HER kinetics of Ni2P/NiMoP.To determine the origins of this incredibly high HER catalytic activity, we conducted electrochemical impedance spectroscopy (EIS) analyses. Ni2P/NiMoP had a lower charge transfer resistance (R ct) than the other materials, suggesting fast kinetics in the electrocatalytic HER process (Fig. 3d and Table S2). The electrochemically active surface area (ECSA) was estimated by measuring the double-layer capacitance (Cdl), which is correlated with ECSA. Figure S8 shows that Ni2P/NiMoP exhibited a Cdl value of 135.5 mF cm–2, about twice that of Ni2P (61.5 mF cm–2), demonstrating that the engineered heterostructure interface of Ni2P/NiMoP made more active sites accessible. When the LSV curves were normalized by the ECSA, the Ni2P/NiMoP still showed excellent catalytic performance (Figure S9).To further assess the intrinsic specific activity of Ni2P/NiMoP, we also investigated the turnover frequency (TOF) (Fig. 3e). At an overpotential of 200 mV, the Ni2P/NiMoP exhibited the largest TOF value, at 0.92 s–1, about twice that of Ni2P and NiMoO4, confirming the Ni2P/NiMoP heterostructure strongly enhanced the intrinsic catalytic activity. Stability is also a critical parameter in assessing catalyst activity. Notably, the LSV curves of the Ni2P/NiMoP electrode remained almost identical after 5,000 and 10,000 cycles (Fig. 3f), implying it had excellent cycling stability. We also assessed long-term electrochemical durability using chronoamperometry at 100 mA cm–2, and the overpotentials remained reasonably stable for 60 h of continuous operation (Figure S10). These results demonstrate the robustness of the Ni2P/NiMoP electrode for the HER.We also investigated the morphology and structure of Ni2P/NiMoP after the HER, finding the nanosheet morphology to be intact after long-term durability testing (Figure S11). The XRD pattern also showed no significant differences from the original (Figure S12), and the heterointerface structure was preserved (Figure S13). These results indicate that the engineered heterostructure interface of Ni2P/NiMoP enhanced the charge transfer rate and the number of accessible active sites, resulting in admirable electrocatalytic activity.Given the sluggish kinetics of the OER, we replaced it with the UOR, in light of the latter's distinctly low theoretical voltage of 0.37 V. Testing was performed in 1 M KOH with 0.33 M urea. First, the polarization curves of Ni2P/NiMoP catalyst for the OER and UOR were compared (Fig. 4 a). The UOR current density was 10 mA cm–2 at a voltage of 1.33 V, much lower than for the OER (10 mA cm–2 at 1.49 V), confirming the crucial role of urea in lowering the anodic potential. We also assessed the effect of urea on the HER performance of Ni2P/NiMoP catalyst (Figure S14) and found negligible difference in the HER polarization curves in the presence of urea, indicating Ni2P/NiMoP resisted urea interference during the HER. The UOR performance of NF, Ni2P, NiMoO4, and commercial Pt/C were probed for comparison (Figs. 4b and S15), with Ni2P/NiMoP emerging as better than all of these (Fig. 4c and Table S3), requiring an ultralow potential of 1.37 V to reach 400 mA cm–2. The Tafel slope of Ni2P/NiMoP was only 23.3 mV dec–1, far lower than that of NF (159.1 mV dec–1), Ni2P (31.8 mV dec–1), NiMoO4 (57.2 mV dec–1), and Pt/C (102.0 mV dec–1), implying faster UOR kinetics (Fig. 4d). The low R ct meant Ni2P/NiMoP had a faster charge-transfer process for the UOR (Fig. 4e).Next, to assess the prospects of Ni2P/NiMoP for industrial applications, we tested its electrochemical stability at a density of 500 mA cm–2 for the HER and UOR. As shown in Figs. 4f and S16, the voltage underwent negligible change after 10 h of testing, indicating Ni2P/NiMoP is suitable for industrial application. We also investigated the morphology, phases, and chemical states of Ni2P/NiMoP after UOR stability testing. An SEM image (Figure S17) shows the nanosheet morphology was well preserved, indicating its excellent structural stability. The XRD pattern showed that in the post-UOR sample, a new phase of Ni(OH)2 had been formed and the phosphide phase had deteriorated (Figure S18), which is attributable to surface oxidation on the electrode. After UOR stability testing (Figure S19), no lattice fringe was evident, indicating the crystalline Ni2P/NiMoP had been transformed into amorphous (oxy)hydroxide species. XPS was conducted to investigate variations in the chemical valence states of the post-UOR sample (Figure S20). The disappearance of the Ni-P peak and the notably decreased peak intensities of P and Mo after UOR experimentation were ascribed to surface corrosion and the formation of hydroxide species [38]. Based on the above experimental evidence, the (oxy)hydroxide species produced on the surface may have served as the actual catalytic active sites for the UOR, and the remaining phosphides as the core to support the efficient transport of electrons.In view of the Ni2P/NiMoP catalyst's extraordinary performance for the HER and UOR, a two-electrode setup employing a Ni2P/NiMoP electrode as a bifunctional catalyst in 1 M KOH with 0.33 M urea was used instead of the traditional water splitting method (Fig. 5 a). The ΔE values of urea-assisted water electrolysis were 1.34 and 1.53 V at 10 and 100 mA cm–2, much lower than for traditional overall water splitting (Fig. 5b), suggesting the prospects for application are good. As depicted in Fig. 5c, the LSV curve of the Ni2P/NiMoP electrode exhibited a voltage of 1.50 V to reach 10 mA cm–2 for traditional water splitting. In contrast, the full cell voltage dropped noticeably to 1.35 V at 10 mA cm–2 after the introduction of 0.33 M urea (Fig. 5d), showing that energy-saving hydrogen generation can be achieved by substituting the anodic OER with UOR. The urea electrocatalysis performance of Ni2P/NiMoP also exceeded those of the most-reported catalysts (Table S4). This two-electrode system was able to steadily generate hydrogen for 80 h of operation at 10 mA cm–2, and the operating voltage showed no significant decay, demonstrating the system's excellent durability (Fig. 5e).Based on the above experimental results and previous reports, the compelling activity of Ni2P/NiMoP may stem from the following features. First, the engineered interface heterostructure could have led to charge redistribution, thereby regulating the electronic structure and optimizing the adsorption capacity of active species during the catalytic process. Second, the hierarchical architecture of the Ni2P/NiMoP nanosheets on the NF substrate not only might have ensured an efficient charge transfer rate but also could have improved the number of accessible active sites and the release of gas. Third, the hydroxide species produced on the phosphide surface may have served as the actual catalytic active sites, favoring the electrocatalytic UOR reaction.In conclusion, we developed a Ni2P/NiMoP nanosheet catalyst with a hierarchical architecture on a NF substrate. The structure and the strong coupling interfaces between Ni2P and NiMoP were proved by XRD and XPS. The interface engineering may have led to charge redistribution, regulating the electronic structure and optimizing the adsorption capacity of active species during the catalytic process. Moreover, the uniform nanosheets on the NF substrate could have promoted charge transfer and improved the number of accessible active sites. The Ni2P/NiMoP catalyst exhibited notable HER and UOR properties. Importantly, a two-electrode electrolyzer assembled with Ni2P/NiMoP as a bifunctional catalyst for both the anode and the cathode required an ultralow cell voltage of 1.35 V to achieve a current density of 10 mA cm–2 and exhibited excellent long-term durability during 80 h of operation. This work offers great potential for developing TMP catalysts to use in energy-saving high-purity hydrogen production via the engineering of heterojunctions.L.F. Jiao proposed the concept. T.Z. Wang and X.J. Cao performed the experiments. T.Z. Wang wrote the manuscript. All authors participated in data analysis and manuscript discussion.The authors declare no competing financial interests.This work was financially supported by the National Natural Science Foundation of China (52025013, 51622102), Ministry of Science and Technology of China MOST (2018YFB1502101), the 111 Project (B12015), and the Fundamental Research Funds for the Central Universities (63191523, 63191746).Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.esci.2021.09.002. Image, application 1

Electrochemical water splitting is a sustainable and feasible strategy for hydrogen production but is hampered by the sluggish anodic oxygen evolution reaction (OER). Herein, an effective approach is introduced to significantly decrease the cell voltage by replacing the anodic OER with a urea oxidation reaction (UOR). A Ni2P/NiMoP nanosheet catalyst with a hierarchical architecture is uniformly grown on a nickel foam (NF) substrate through a simple hydrothermal and phosphorization method. The Ni2P/NiMoP achieves impressive HER activity, with a low overpotential of only 22 mV at 10 mA cm–2 and a low Tafel slope of 34.5 mV dec–1. In addition, the oxidation voltage is significantly reduced from 1.49 V to 1.33 V after the introduction of 0.33 M urea. Notably, a two-electrode electrolyzer employing Ni2P/NiMoP as a bifunctional catalyst exhibits a current density of 10 mA cm–2 at a cell voltage of 1.35 V and excellent long-term durability after 80 h.

The accumulation of waste plastics in landfills and oceans has caused a global environmental crisis. 1–3 In particular, microplastics have been entering the food chain and become a potential threat to human health (B. Liebmann et al., 2018, Microplastics 2018, conference). Although there are thousands of plastic materials in use, only six of them—polyethylene (PE, high and low density), polypropylene, poly(vinyl chloride), polystyrene (including expanded polystyrene), polyurethane, and poly(ethylene terephthalate)—are widely used. Collectively, ~6.3 billion metric tons of plastic waste were produced by 2015, of which 79% was landfilled, 12% was incinerated, and only 9% was recycled. 4 PE is the polymer with the most massive volume produced globally, and the production could reach over 100 million metric tons per year. 4 , 5 Therefore, the efficient upcycling of waste plastics, especially PE, is critical to mitigating the severe environmental problem.Technologies for recycling waste plastics mainly include three types: mechanical recycling, incineration, and chemical recycling. Mechanical recycling is the only technology used commercially for the large-scale plastic recycling process, but it still suffers from decreasing product quality after the consecutive melting and remolding cycles. 6 Although incineration converts mixed waste plastics to heat and electricity, the energy recovery efficiency cannot be as much as that from chemical recycling because of the massive loss of energy. 7 Therefore, chemical recycling is considered a promising process for valorizing waste plastics, whereby plastics are the low-cost feedstock for producing value-added chemicals or fuels.Recently, pyrolysis has been extensively investigated as a chemical recycling technology. The world's largest resin producers, including Chevron Phillips Chemical (CPC), Saudi Basic Industries Corporation, and BASF, have been using this technology to produce circular polymers from plastic waste. 8–10 Indeed, CPC has already accomplished the first commercial-scale production of circular PE in the United States. In addition to the commercial application, catalytic pyrolysis has also drawn much interest from research communities. The production of syngas or liquid hydrocarbon fuels from PE waste is technically feasible. 11 However, elevated temperatures (>300°C) are needed in catalytic pyrolysis processes, 12–14 which might not be economically sound given the high energy consumption. Moreover, it is challenging to control product distribution at high temperatures. In addition to linear alkanes, branched, cyclic, and aromatic hydrocarbons are produced during pyrolysis. 15–17 Aromatics are of value, but they can readily be transformed into coke that might cause catalyst deactivation. 18–21 Even though the catalyst could be regenerated after the coke is burned, the operation cost would increase substantially.Therefore, developing effective catalytic processes that could selectively convert PE to high-value chemicals under mild reaction conditions is of utmost importance for chemical upcycling of PE waste plastics. 22 For instance, Sadow and coworkers 23 designed a mesoporous catalyst with a Pt core@SiO2 shell structure to selectively convert high-density PE (HDPE) into a narrow distribution of diesel- and lubricant-range alkanes in a solvent-free system (300°C, 24 h, 1.38 MPa H2). The polymer molecules thread and bind into the silica pores, and the small-molecule products desorb and exit the pores after the cleavage from the polymer end at the active sites on the Pt metal catalyst surface. Likewise, Scott and coworkers 24 developed a tandem solvent-free hydrogenolysis-aromatization process to produce valuable alkyl aromatics from PE with a Pt/Al2O3 catalyst at 280°C. Although these solvent-free methods provided a strategy for manufacturing higher-value products from PE waste, the kinetic performance is still an issue because it requires an extended processing time (24 h).In general, compared with solvent-free pyrolysis, PE depolymerization can be promoted dramatically with the use of solvents, where mass transfer and heat transfer rates can be improved. 25–27 Adams et al. 28 used ionic liquids to convert PE at 120°C, and the yield of low-molecular-weight hydrocarbons reached 95% in 72 h. Although the reaction temperature was much lower, the reaction time had to be prolonged to achieve satisfactory outcomes. Meanwhile, the separation might be an issue given that another solvent was needed for extracting the products from the ionic liquid solvent. Jia et al. 29 reported that PE was degraded into transportation fuels and waxes through cross-alkane metathesis with hexane, 98% of which were converted into liquid hydrocarbon oils at 150°C in 3 days. Ideally, a well-designed solvent system with appropriate heterogeneous catalysts could promote highly selective PE depolymerization under mild conditions. However, for the current solvolysis process, catalytic deconstruction rates still need to be enhanced. Practically, the recovery, reuse, and lifetime of solvents and catalysts could also be limiting factors for large-scale applications.In our previous study, we found that ruthenium on a carbon (Ru/C) catalyst was able to convert n-heptadecane into short-chain hydrocarbons under mild conditions. The Ru catalyst is known to be capable of cleaving the C–C bond. 30 , 31 The dehydrogenative chemisorption of the hydrocarbons is considered the first step in the mechanism of hydrogenolysis on active metal, and then the formed hydrogen-deficient surface species go through C–C bond scission. 32 After the cleavage of C–C, the reaction is finally completed by hydrogenation and desorption. PE, consisting of long hydrocarbon chains, has the simplest structure of any of the polymers. While our manuscript was under review, the remarkably high activity of the Ru catalyst in the hydrogenolysis of PE was also reported by Rorrer et al. in the absence of solvent. 33 We hypothesize that Ru catalysts can break the C–C bonds in PE polymers by using a suitable solvent. Hence, in the current study, we investigated the conversion of PE to liquid fuels with a Ru/C catalyst in the liquid-phase reaction, which has not been reported previously to the best of our knowledge. Table 1 shows the structural parameters of fresh and spent Ru/C catalysts. The specific surface area, the metallic surface area, and the active-metal dispersion decreased after the first run but remained the same after the second run. The result showed that the catalyst structure became stable after the first cycle. The decrease in Ru dispersion could be partly due to metal leaching during the reaction. The Ru particle size increased from 2.9 to 4.1 nm, indicating that sintering occurred after the first run. These structural changes could explain the decrease in the catalytic activity after the first run.Transmission electron microscopy (TEM) images of the fresh and spent Ru/C catalysts are displayed in Figure 1 , showing that the Ru nanoparticles were well dispersed on the C support. The mean particle size on the fresh catalyst was approximately 3.1 nm. A slight shift in the particle-size distribution was observed on the used catalysts, although the particle size was in the range of 2–5 nm. According to the TEM images, the mean particle size of the spent Ru/C catalysts after the first and second cycles was 4.2 and 4.0 nm, respectively, which is consistent with the CO pulse chemisorption result. Both characterization results demonstrated that the aggregation occurred on the Ru/C catalyst after the first cycle, whereas the Ru particle size was nearly unchanged in the subsequent cycles.We employed X-ray photoelectron spectroscopy (XPS) to investigate the change in valence state in the Ru particles before and after the reaction. Because the Ru 3d doublet overlaps C 1s, Ru 3p is commonly used for characterizing the change in the Ru element valence state. Figure 2 shows that the Ru 3p1/2 and 3p3/2 binding energies of the fresh Ru/C catalyst were 462.9 and 485.0 eV, respectively, whereas those of the spent catalyst shifted to low values, 462.4 and 484.8 eV, respectively, after the reaction, indicating that Ru oxide on the catalyst was reduced by H2 during the reaction. Meanwhile, the Ru atomic percentage decreased from 1.6% to 1.05% after the first cycle, whereas it remained the same after the second cycle, which is consistent with the trend of the decrease in the metallic surface area in Table 1.The crystalline structures of the fresh and used catalysts before and after the HDPE depolymerization, respectively, were characterized through X-ray diffraction (XRD) (Figure 3 ). Two XRD peaks at about 2θ = 25° and 43° are associated with the (002) and (100) phases of the C support, respectively. No Ru or Ru oxide peaks were observed, indicating that the Ru particles were very small and dispersed on the C support very well. 34 No significant change in the XRD patterns was observed before or after the reaction, implying that the catalyst's crystal structure might be unchanged.The HDPE depolymerization reaction was investigated with a variety of C-supported metal catalysts under the same reaction conditions. The experimental results in Table 2 show that the copper, iron, palladium, platinum, and nickel catalysts displayed no effect on the HDPE depolymerization at 220°C. Although other groups have reported that iron, palladium, and nickel can promote PE deconstruction, high temperatures (e.g., 430°C) are still necessary for such processes. 35 , 36 Recently, Pt@SiO2 catalysts were reported to carry out the hydrogenolysis of HDPE in a solvent-free system for an extended reaction time, 24 h, at a relatively low temperature (250°C). 23 In contrast, in our study, only <0.5 wt % of the HDPE depolymerization products (C8–C38) was detected on gas chromatography-mass spectrometry (GC-MS) with the Pt/C catalyst in n-hexane even when it was reacted for 6 h at 250°C. The solvent system's poor performance could be ascribed to HDPE's low solubility in supercritical n-hexane (critical temperature: 234.5°C). Rhodium (Rh) was reported to have catalytic ability in C–C cracking, which is similar to Ru. 37 However, with the Rh/C catalyst, no detectable liquid hydrocarbon products by GC-MS were observed at 220°C, although there was no residue after the reaction. Long-chain hydrocarbons (>C45) with high molecular weights, which are beyond the detection limit of our mass spectrometer, could be the main products. As the temperature increased to 280°C, an ~75.3 wt % yield of alkanes in the range of C8–C38 was obtained (Figure S1A), demonstrating that Rh is also active for C–C hydrogenolysis at elevated temperatures. In contrast, the full conversion of HDPE to hydrocarbon fuels by pyrolysis with the Ru/Y-zeolite catalyst was accomplished at 600°C. However, the severe coke deposition on the catalyst in pyrolysis raised concerns about the catalyst’s stability. 38 Here, we found that the Ru/C catalyst was superior among all the screened catalysts in this study. The HDPE strips were converted to 60.8 wt % jet-fuel-range and 14.1 wt % diesel-range alkanes at 220°C in just 1 h with the Ru/C catalyst in n-hexane, and no long-chain products could be detected (Figure S1B). Compared with other metals, Ru metal was reported to have the lowest activation energy in ethane hydrogenolysis, favoring the C–C bond cleavage. 32 In the comparison of ethane hydrogenolysis on transition-metal catalysts, ∗CHCH∗ was found to be the primary intermediate in the C–C bond scission for Ru, Rh, and Pt because it has the lowest free-energy barrier in C–C bond cleavage. 39 Meanwhile, both ∗CHCH∗ and ∗CH3CH∗ were considered dominant intermediates for Pd. Among these transition metals, the turnover rate in ∗CHCH∗ cleavage decreases in the order Ru > Rh > Pt > Pd, which is consistent with our result that Ru could cleave the C–C efficiently and that Pd has the lowest cleavage turnover rate.The temperature effect on the HPDE depolymerization is shown in Figure 4A. We detected no cracking product at 150°C. When the depolymerization was carried out at 200°C, a complete HDPE conversion to liquid-phase alkanes was obtained. With increasing temperature, the yield of high-molecular-weight alkane products decreased. The yield of the jet-fuel-range alkanes (C8–C16) reached a maximum of ~60 wt %, whereas that of the diesel fuels (C17–C22) was ~15 wt % at 220°C, and almost all long-chain hydrocarbons (C number > 23) were converted to short-chain alkanes in 1 h. As the temperature increased to 230°C, the yields of jet- and diesel-fuel-range alkanes decreased to ~55 and ~5 wt %, respectively, as a result of excess cracking. The HDPE polymer is not easily solvated in a supercritical solvent. At 240°C, which is higher than n-hexane's critical temperature (234.5°C), we observed an abrupt change in the product distribution compared with that at 230°C. The yield of the long-chain hydrocarbon products (C17–C38) increased dramatically from <5 to ~50 wt % as the temperature increased just 10°C (from 230°C to 240°C), implying that the low solubility of HDPE in the supercritical n-hexane solvent could lead to much slower C–C bond cracking rates.The reaction time is another crucial parameter for determining the product distribution. Here, the effect of reaction time on the HDPE depolymerization was also investigated, and the results are shown in Figure 4B. Surprisingly, HDPE was rapidly degraded to liquid hydrocarbons (C number < 38) in only 0.5 h at 220°C. With increasing reaction time, the yield of jet-fuel-range alkanes increased first and then decreased as a result of excess cracking. The maximum yield (~60 wt %) of jet-fuel-range alkanes was achieved in 1 h. Almost no high-molecular-weight products were observed after 1 h.Further, we also investigated the catalyst loading effect on the depolymerization by varying the amount of catalyst. As shown in Figure 4C, the depolymerization reaction did not occur in the absence of a catalyst. The depolymerization reaction rate increased with increasing catalyst loading. With a low loading of the catalyst ([Ru]/[HDPE] ratio was 2.1%), the yield of lubricant-range hydrocarbons (C24–C35) reached 31.6%. While the [Ru]/[HDPE] ratio increased to 8.3%, the yield of jet-fuel-range alkanes achieved the maximum value (~60 wt %). As the catalyst amount continued to increase, the corresponding jet-fuel yield decreased. Meanwhile, more short-chain hydrocarbons (C number < 8) were observed after the [Ru]/[HDPE] ratio surpassed 1.2%, indicating that an increasing amount of catalyst would promote the cracking reaction. Figure 5 shows that hydrogen pressure played a significant role in the HDPE depolymerization. In the absence of H2, no product was detected. With increasing H2 pressure from 0 to 60 bar, the depolymerization reaction rate increased first and then decreased after the H2 pressure passed 30 bar, indicating that higher hydrogen pressure could inhibit the depolymerization reaction. Iglesia and coworkers also observed that hydrogenolysis of the linear and branched alkanes (C2–C8) was reduced as the H2 pressure increased. 40 They found that H2 pressure could also influence the C–C bond cleavage position in long-chain alkanes, probably as a result of the dehydrogenated intermediates formed by quasi-equilibrated adsorption and dehydrogenation. 41 , 42 At low hydrogen pressures, the hydrogenolysis rates were proportional to the concentration of the reactive unsaturated intermediate [∗CnH2n+2−y∗], and the rates increased with hydrogen pressure. 43 At high hydrogen pressures, the surface was mainly occupied by chemisorbed hydrogen atoms (H∗), hindering the adsorption of intermediates and decreasing the hydrogenolysis rates. Note that Iglesia and coworkers studied the alkane hydrogenolysis in the gas phase, which could significantly differ from PE's hydrogenolysis in solvents. HDPE's structure resembles those of long-C-chain linear alkanes (varying in C chain length), consisting of only Csecondary–Cprimary and Csecondary–Csecondary bonds. Hence, the Ru-catalyzed HDPE hydrogenolysis includes primarily two independent reactions: regioselective hydrogenolysis of the easily accessible C–C bonds (e.g., Csecondary–Csecondary) and hydrogenolysis of Csecondary–Cprimary bonds (i.e., chain-end scission). 44 Thus, the scission of Csecondary–Csecondary is preferred for acquiring more valuable long-chain hydrocarbons.Also, the hydrogenolysis mechanism of linear liquid-phase alkanes would be analogous to the dissociation mechanism for the C–C bonds in HDPE and its degradation intermediates. Herein, the hydrogen pressure effect was further explored with eicosane, a C20 linear alkane, as the probe reactant (Figure 6 ). We found that at low H2 pressure (10 bar), the C19 alkane, n-nonadecane, was the dominant product, indicating that terminal dissociation was the main pathway. As the H2 pressure increased to 60 bar, the main products were octadecane and heptadecane (C18H38 and C17H36), demonstrating that the primary pathway was changed to internal dissociation. Nakagawa et al. reported that with a Ru/CeO2 catalyst and the absence of solvents, the reaction order to the H2 partial pressure for cracking n-hexadecane (C16H34) was 0.4. Non-stoichiometric methane formation from n-hexadecane ([methane] − [C15] = −0.8) was observed, indicating that high hydrogen pressure suppressed excess methane formation, i.e., the cleavage of Csecondary–Cprimary. 44 The same group also observed that under higher hydrogen pressures, the yield of C15 from terminal dissociation was lower than the average of the internal dissociation product yields, which is similar to our result that only a low yield of C19 was obtained at 60 bar of H2. Notably, Nakagawa et al. found no significant difference between the yields of C2–C14 hydrocarbons, whereas we observed that the main products, C18 and C17, were acquired with the presence of a solvent.Likewise, HDPE is a linear alkane polymer containing predominantly secondary C atoms and a few primary C atoms; the influence of hydrogen pressure on the hydrogenolysis of HDPE seems similar to that of eicosane. At low H2 pressures, the liquid alkane products might mainly be generated from the terminal dissociation, which was suppressed with increasing H2 pressure. After the H2 pressure passed a threshold value, the internal dissociation became dominant. At 60 bar of H2, ~90% of HDPE was converted to C8+ liquid hydrocarbon products, implying that internal dissociation is the primary depolymerization pathway at high H2 pressures. However, both terminal and internal dissociation can coexist in a wide range of H2 pressures during HDPE depolymerization.Solute solubility and thermodynamic equilibrium coefficients are critical parameters that affect the reaction kinetics in solutions. 45 Here, the role of different organic solvents in HDPE depolymerization was investigated. In a polar solvent, e.g., water, the HDPE degradation rate was found to be very slow at 220°C, as shown in Figure 7 . Typically, PE can be degraded in supercritical water whose dielectric constant is comparable to those of the polar organic solvents. 46 , 47 Although the supercritical hydrolysis process requires a very high energy input, the low polarity of supercritical water facilitates PE's dissolubility and thus promotes the reaction rate. However, at 220°C, subcritical water is much denser and more polar than supercritical water, leading to a low PE solubility and thus a slow depolymerization reaction rate. Meanwhile, we observed that the HDPE strips were transformed into spherical solid particles after the reaction, which was different from that in the organic solvents (Figure S2). These plastic strips usually melted at over 150°C. 48 The formation of spherical solids indicated that the plastic strips were melted but were not solvated in the water at 220°C as a result of the low-solubility HDPE in subcritical water. Therefore, non-polar solvents were preferred for PE dissolution and depolymerization. Figure 7 shows that n-hexane was the optimal organic solvent for HDPE degradation with the Ru/C catalyst, whereas other non-polar solvents exhibited much different performance in the depolymerization reaction. Notably, no cracking products were detected in n-pentane solvent, although the polarity of n-pentane is very similar to that of n-hexane. Here, the reaction temperature (220°C) was higher than n-pentane's critical temperature (196.45°C) but lower than n-hexane's critical temperature (234.5°C). Therefore, the supercritical pentane solvent behaved very differently from those at lower temperatures. HDPE polymers might not be solvated in the supercritical n-pentane, causing high resistance to mass and heat transfer. We also observed that the HDPE strips were transformed into spherical particles in the supercritical n-pentane after the reaction, implying that HDPE was melted rather than dissolved.We evaluated the solvation effect by using the Hansen solubility parameters, which are based on the theory of “like dissolves like.” 49 As shown in Tables S1 and S2, the relative energy difference (RED) of water and PE is much larger than 1, indicating that water is not a suitable solvent for PE. The RED values are less than 1 for other organic solvents that show a high affinity, consistent with the experimental results that HDPE polymer could be dissolved in these solvents. It is reasonable that PE solvation in the solvents is the first step in the degradation reaction (Scheme 1 ). We observed that the solvent molecular structure profoundly affects the depolymerization, as shown in Figure 7. For instance, methylcyclohexane was not as efficient as n-hexane for depolymerization because of its obstructive cyclic molecular structure. Under identical reaction conditions, the dominant products with the n-hexane solvent are the medium-chain n-alkanes (C8–C16), whereas the longer-chain n-alkanes (C17–C38) are the main products in methylcyclohexane. Nevertheless, the appropriate inhibition effect on the PE depolymerization in methylcyclohexane was desired for controlling the product distribution given that the long-chain hydrocarbons (C17–C38) are the target products, such as lubricants, with a higher profit margin than the medium-chain n-alkanes (C8–C16), which are jet-fuel components. A similar steric hindrance effect was also observed with decalin as the solvent, whereby no cracking liquid hydrocarbon products were detected after the reaction. The solvated polymer molecules in decalin might be obstructed from being in contact with the heterogeneous Ru/C catalyst surface. Note that the molecular size of n-hexane is 1.03 nm (length) × 0.49 nm (width) × 0.4 nm (height), which is much larger than methylcyclohexane (0.79 × 0.73 × 0.5 nm) and slightly longer than decalin (0.91 × 0.72 × 0.5 nm). 50–52 Nevertheless, the linear molecules, e.g., n-hexane, were more flexible, compensating for their bulky molecular size. 53 , 54 The similarity in shape between n-hexane and HDPE could facilitate the diffusion of large PE oligomer molecules in the solvent, which allows the access of bulky reactant substrates to the Ru/C catalyst surface. In addition, methylcyclohexane and decalin are known as the hydrogen-donor solvents, 55 which can transfer hydrogen even in the H2 atmosphere. The solvent-donated H∗ could quickly react with the polymer radicals, terminating the consecutive cracking reactions. 56–58 According to the results of the molecular dynamics (MD) simulations, PE adopts a compact conformation in pentane and hexane, with the lowest radius of gyration value (Rg), followed by water and methylcyclohexane, and finally it adopts an extended conformation in trans-decalin (Table S3). The extended conformation of PE in decalin can be attributed to the high degree of hydrophobicity of decalin solvent. A PE molecule is also hydrophobic in nature and thus prefers to be in hydrophobic solvents, resulting in the fully extended conformation of the PE molecule in hydrophobic solvents such as decalin. To understand the influence that the structure might have on dynamic properties, we computed the end-to-end polymer chain autocorrelation function in different solvents, as shown in Figure 8 , which estimates how readily the polymer relaxes in a particular solvent. Figure 8 shows that PE polymer decorrelates fastest in n-pentane and n-hexane, followed by methylcyclohexane, water, and decalin. The decorrelation order of the PE polymer is in accordance with the amounts of short-chain hydrocarbon molecules produced in the experiment, except for n-pentane. In our simulations, PE decorrelated fastest in n-pentane; however, PE did not depolymerize in n-pentane to produce short-chain hydrocarbon molecules in the experiment. In this case, the difference observed between experiment and simulation is due to the limitation of simulations to capture supercritical behaviors of n-pentane. In our simulations at 493 K, n-pentane still behaved like a normal fluid rather than a supercritical one. As a result, the behavior of PE in n-pentane is similar to that in n-hexane.The PE end-to-end length decorrelation rate correlates to the affinity of PE polymer toward the solvent it is immersed in, as described by the radius of gyration results. As the simulation progresses, the interaction between PE polymer and solvent molecules causes the conformation of the PE polymer to change. The cases in which the PE end-to-end length decorrelates fast (for example, in n-hexane) indicate that the PE polymer does not have a high affinity toward solvent molecules, thus causing the PE polymer to coil. We propose that the coiled polymer adsorbs in this state on the catalyst surface and undergoes cracking reactions. The coiled structure has a high tendency to pass through the solvent molecules to reach the catalyst surface. In contrast, the slow decorrelation rate of PE in decalin shows PE affinity toward decalin, where PE polymer can sustain extended conformations for a longer time. The comparison between PE conformation in decalin and in hexane after 500 ns of NVT (substance, volume, and temperature) simulations is shown in Figure 9 . The straight PE chain in decalin has a high affinity toward solvent molecules, preventing the straight PE chain from reaching the catalyst surface for depolymerization reactions and leading to poor kinetic performance. The extended configurations of PE in decalin suggest higher relative thermodynamic stability in the bulk solvent as a result of increased entropy arising from the chain flexibility in the solvent. This result suggests that considering both the solvent quality and the adsorption affinity of collapsed and extended PE chains could determine an additional important screening characteristic for solvents used in depolymerization processes.The catalyst stability is a big hurdle in plastic depolymerization via catalytic pyrolysis. 59 , 60 In our study, the catalyst did not show severe deactivation in the n-hexane solvent after being used for five cycles (Figure 10 ). The yield of jet-fuel-range alkanes (C8–C16) decreased only slightly after first use and then became stable in the subsequent runs, indicating that the catalyst stability would be reliable for depolymerization. We observed that more short-chain hydrocarbons were generated after the first cycle, which could be ascribed to the increase in Ru particle size. Nakagawa et al. found that the terminal dissociation was more prevalent if the Ru particle size increased from <1.5 to >2 nm. 44 Therefore, smaller particle size might favor the yield of jet-fuel-range products. Furthermore, the thermal gravimetric analysis (TGA) curves showed that the Ru loading decreased by 0.62% after the first cycle and remained almost the same after the second cycle (Figure S3), which is consistent with the trend of decrease in the metallic surface area in Table 1. Both results demonstrated that Ru would not continuously leach after the first use.Because of the high catalytic activity of the Ru catalyst in cleavage of the C–C bond, the solvent stability is important for the PE hydrogenolysis process. A blank experiment was conducted without the addition of HDPE (0.05 g Ru/C, 25 mL n-hexane, 220°C, p(H2) 20 bar, 1 h, 700 rpm). Approximately 5.6 wt % of the solvent (including 5.1 wt % loss by evaporation) was lost after the reaction, which was much lower than in the cross-alkane metathesis process for PE depolymerization (15.1 wt % loss) with light alkanes as both the solvent and the feedstock and (t-Bu2PO-t-BuPOCOP)Ir(C2H4)/γ-Al2O3 and Re2O7/γ-Al2O3 as catalysts at 175°C for 4 days. 29 Moreover, for process optimization, the short-chain hydrocarbon products from HDPE depolymerization could be reused as the makeup solvent in the process.In summary, we have demonstrated an efficient liquid-phase hydrogenolysis process with the heterogeneous Ru/C catalyst for selective depolymerization of waste HDPE plastic under mild conditions. Approximately 90 wt % HDPE was converted to C8+ liquid hydrocarbon products in an n-hexane solvent within 1 h under 30 bar H2 at 220°C. We were able to tune the product distribution by adjusting the process conditions, including catalyst loading, reaction temperature, hydrogen pressure, and reaction time. With high catalyst loading, high reaction temperature, or prolonged reaction time, excess cracking occurred during the reaction and led to the production of less valuable short-chain hydrocarbons. Hydrogen pressure played a significant role in the polymer dissociation pathway. Under low H2 pressures, terminal dissociation was dominant, whereas internal dissociation was prevalent when the H2 pressure increased.Furthermore, solvents also profoundly affected the depolymerization reaction kinetics and product selectivity. The solvation ability of PE in solvents was a key factor for depolymerization. The degradation of HDPE in subcritical water was slow because of its low solubility in polar solvents. Among the non-polar hydrocarbon solvents, n-hexane (a linear alkane) was better for HDPE depolymerization than the cyclic alkanes (methylcyclohexane and decalin). The highest yield of jet-fuel-range hydrocarbons (C8–C16) reached 60.8 wt % in the n-hexane solvent at 220°C. The MD simulations suggest that the interaction between PE polymers and solvent molecules causes the conformation of the PE polymer to change. The PE polymer with a low affinity toward solvent molecules tends to coil and then sieve through solvent molecules and get to the catalyst surface, where it will get cracked. PE adopts a compact coil conformation in pentane and hexane, followed by water, methylcyclohexane, and decalin. Although the steric hindrance from the solvents' cyclic molecular structure inhibited PE depolymerization, it promoted the production of long-chain hydrocarbons, such as lubricants.Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Hongfei Lin ([email protected]).This study did not generate any new unique reagent or material.This study did not generate codes, software, or algorithms.The feedstocks, HDPE plastic water jugs, were collected from the local recycling center in Pullman, Washington. Before the experiment, the jugs were cleaned with deionized water, dried at 100°C, and then cut into strips (5 × 5 mm). All chemicals were used as received without further treatment. The catalysts (Ru/C [5% Ru basis], Pd/C [5% Pd basis], Pt/C [5% Pt basis], and Rh/C [5% Rh basis]), the catalyst precursors (copper(II) nitrate trihydrate [99%] and iron(III) nitrate nonahydrate [98%]), and the self-synthesized catalyst support (activated charcoal Norit) were supplied from Sigma-Aldrich. Nickel(II) nitrate hexahydrate (99%) was purchased from Millipore Sigma. p-xylene (99%) was purchased from Alfa Aesar. Ultrapure water (specific resistance of 18.2 MΩ cm−1), n-pentane (Alfa Aesar, 98%), n-hexane (J.T. Baker, 95%), methylcyclohexane (Alfa Aesar, 99%), and decalin (Tokyo Chemical Industry, 99%) were used as the solvents.5% Cu/C, 5% Fe/C, and 5% Ni/C were synthesized through impregnation with copper nitrate trihydrate, iron nitrate nonahydrate, and nickel nitrate hexahydrate, respectively, as the metal precursors and activated charcoal Norit as the support. After being dried, the as-prepared 5% Ni/C, 5% Fe/C, and 5% Ni/C samples were calcined at 350°C (Ni/C) or 500°C (Fe/C and Ni/C) for 3 h in an atmosphere of nitrogen. Finally, the catalysts were reduced in H2 flow at 400°C (Ni/C) or 500°C (Fe/C and Ni/C) for 5 h prior to use.The specific surface area of the catalysts was determined through single-point adsorption of N2 at 77 K with a Micromeritics Autochem II 2920. The samples were prepared in helium at 200°C for 1 h before nitrogen adsorption (30% N2/He).The CO pulse chemisorption was used for determining the metal dispersion, active-metal particle size, and metallic surface area. The test was carried out on a Micromeritics Autochem II 2920. The sample was reduced for 2 h at 300°C with 10% H2/Ar at a 50 mL/min flow rate and then purged with helium for 1 h at a flow rate of 50 mL/min. After the sample was cooled to ambient temperature, 10% CO/He was added at each pulse, and the CO uptake profile was measured with a thermal conductivity detector (TCD) until no CO was adsorbed. The Ru dispersion was calculated under the assumption of a CO/Ru stoichiometry of 1:1. 61 The fresh and spent Ru/C catalysts were characterized by TEM on a JEOL 2010 J microscope at an accelerating voltage of 200 kV. The Gatan Digital Micrograph software was used for conducting data processing and analysis. The catalyst powder samples were dispersed on Formvar film nickel grids (200 mesh).The XPS analyses were carried out on a Kratos AXIS-165 with a monochromatized Al-Kα X-ray anode (1,486.6 eV) with the C 1s peak at 284.6 eV as the internal reference. The deconvolutions of Ru 3p were analyzed with the software XPSPEAK version 4.1.The crystalline catalyst structure was evaluated by X-ray powder diffraction (Rigaku Miniflex 600), with a Co-Kα radiation source (λ = Å) at a 2θ step of 10°–90° with a step size of 0.02°.TGA was performed with a TA Instruments Q50. The samples were loaded in aluminum crucibles and heated in airflow (60 mL/min) from 25°C to 600°C at a heating rate of 10°C/min.The depolymerization experiments were carried out in a 45 mL elevated pressure and temperature Parr Series 5000 multiple reactor system with a 4871 temperature controller. In a typical experiment, a certain amount of HDPE strips and catalyst were loaded in 25 mL solvent. The vessels were sealed and purged five times with 400 psi N2 and three times with 400 psi H2 and then pressurized with H2 to the set pressure at ambient temperature. Then the reactor was heated up to the set reaction temperature with magnetic stirring at 700 rpm. After the reaction, the vessel was quenched in a cold bath for fast cooling.After the reaction, the reactor was connected to a gas chromatograph Shimadzu GC-2014 with a TCD for analysis of the gas-phase product samples. The columns included a right 12.5 m (l) × 0.32 mm (i.d.) packed column, which comprised 3 m Hayesep D, 4 m HS, and 2.5 m HN, and a left 2 m (l) × 0.32 mm (i.d.) 10% Carbowax 20 m Ch packed column. After the reactor was disassembled, the solid catalyst and non-dissolvable residues were filtered out of the liquid phase. Then the liquid product samples were collected, and the internal standard, p-xylene, was added. The liquid samples were analyzed by a QP-2020 (Shimadzu) gas chromatograph-mass spectrometer for identifying and quantifying the unknown products. The QP-2020 was equipped with a Shimadzu SH-Rxi-5SIL MS column (30 m × 0.25 mm i.d., 0.25 μm film thickness), a flame ionization detector, and a high-performance ion source. The following definitions were used for quantitating the weight yield (y): y = ∑ m x m 0 × 100 % , where m0 is the weight of the HDPE feedstock before reaction and mx is the weight of the alkane hydrocarbons after the reaction, where x means the C number.A PE molecule C100H202 in length was packed into five different simulation boxes of 10 × 10 × 10 nm3. Each box was filled with one of the five different solvents: methylcyclohexane, n-pentane, n-hexane, water, or decalin. Water was modeled with the SPC/E water force field, 62 while the force fields for the organic solvents were obtained from the Automated Topology Builder repository. 63 For decalin, the isomer used was trans-decalin because trans-decalin is more stable than its cis counterpart as a result of its diequatorial chair conformation. Each system was simulated with the GROMACS 2018.3 simulation package. 64 The steepest descent algorithms were used for removing unfavorable contacts in the initial configuration. Electrostatic interactions were calculated with the particle mesh Ewald summation method 65 with an electrostatic cutoff value of 1.0 nm and van der Waals cutoff value of 1.0 nm. The system was evolved in the NPT ensemble (temperature 493 K, pressure 1 atm) for 2 ns with the Donadio-Bussi-Parrinello thermostat 66 (time constant τ = 0.1 ps) and the Berendsen barostat 67 (time constant τ = 1 ps). A temperature of 493 K was chosen to be consistent with the experiment. All the dimensions of the box were allowed to change during the NPT simulation. The production runs were carried out in the NVT ensemble (temperature 493 K), where the temperature was maintained by the Donadio-Bussi-Parrinello thermostat (time constant τ = 0.1 ps) for 500 ns.The polymer structure in the solvent was captured through the average radius of gyration calculated over the entire simulation time of 500 ns. To assess the dynamic behavior of the polymer in different solvents, we calculated the end-to-end autocorrelation function according to the following equation: e 2 e ( t ) ≡ ⟨ A ( t ) ⋅ A ( 0 ) ⟩ ⟨ A ( 0 ) ⋅ A ( 0 ) ⟩ , where A is the vector from the first C atom to the last C atom along the polymer chain.This project was partially funded by the Washington State University internal fund and Washington Research Foundation. C.J. is thankful for the Chambroad Fellowship from the Gene and Linda Voiland School of Chemical Engineering and Bioengineering at Washington State University. Computational simulation in this research was supported by NSF-CBET award 1703638 and was facilitated through the use of advanced computational, storage, and networking infrastructure provided by the Hyak supercomputer system at the University of Washington. The authors would like to acknowledge Zengran Sun and Prof. Steven R. Saunders for their support in the thermal gravimetric analysis. The authors also thank Dr. Baoming Zhao and Prof. Jinwen Zhang for valuable discussions.H.L. proposed, designed, and guided the project and revised the manuscript. C.J. performed most of the experiments and drafted the manuscript. S.X. and W.Z. also took part in the experiments and revised the manuscript. N.I., J.S., and J.P. performed the molecular dynamics simulations. All authors checked the manuscript.The authors declare no competing interests.Supplemental information can be found online at https://doi.org/10.1016/j.checat.2021.04.002. Document S1. Supplemental experimental procedures, Figures S1–S3, and Tables S1–S3 Document S2. Article plus supplemental information

Polyethylene (PE) is the most popular plastic globally, and the widespread use of plastics has created severe environmental issues. High energy consumption in the current process makes its recycling a challenging problem. In our report, the depolymerization of high-density PE was conducted in various liquid-phase solvents with the Ru/C catalyst under relatively mild conditions. The maximum yields of the jet-fuel- and lubricant-range hydrocarbons were 60.8 and 31.6 wt %, respectively. After optimization of the reaction conditions (220°C and 60 bar of H2), the total yield of liquid hydrocarbon products reached approximately 90 wt % within only 1 h. The product distribution could be tuned by the H2 partial pressure, the active-metal particle size, and the solvents. The solvation of PE in the different solvents determined the depolymerization reaction kinetics, which was confirmed by the molecular dynamics simulation results.

With the rapid industrial development, an increasing number of organic pollutants are discharged into the aquatic environment, producing negative impacts on humans, plants, and animals, as well as the entire ecosystem (Trejo-Castillo et al., 2021). Benzotriazole (BTA), a widely used chemical, has become a common additive in mineral flotation agents (Yao et al., 2021), circulating cooling water treatment agents (Yang et al., 2021), anti-icing fluids, dishwasher detergents, and metal corrosion inhibitors (Castaldo et al., 2020). This causes significant amounts of BTA to be discharged into the aquatic environment through industrial excess discharges (Cheng et al., 2021) or mixed surface runoff collected in sewage systems (Yang et al., 2021). Because of its persistence, bioaccumulation, and toxicity, BTA inhibits the growth and reproduction of aquatic organisms, and its estrogenic potential may have deleterious effects on the sex differentiation system of many organisms (Feng et al., 2020). In addition, BTA interferes with aquatic species and soil microbial communities, and it is carcinogenic and mutagenic in mammals (Li et al., 2020a). As a result, BTA is classified as an emerging polar pollutant. Given that almost no enzymes in organisms can degrade BTA, conventional wastewater treatment technologies can only remove about 30% of BTA in effluent (Yin et al., 2021). Thus, the existing BTA treatment technology needs to be improved effectively.Advanced oxidation technology (Liu et al., 2009) is a widely used and effective method to degrade organic pollutants. Photoelectrocatalysis (PEC) (Brillas, 2020) has gained attention because it is an efficient and simple method that does not produce secondary contamination and improves the cavity time, and it has shown excellent performance in the degradation of BTA (Hu et al., 2016). PEC has effectively degraded BTA using TiO2-coated electrodes, with a removal rate of 82.1% after 180 min (Ding et al., 2009). Wu et al. (2013) used ZnFe2O4 as a catalyst and performed a photoelectric-Fenton like reaction for 180 min to effectively remove 91.2% of BTA.The key to practical application of PEC is to develop a cheap and efficient catalyst. The doping of transition metal ions has become popular in environmental studies. Metals, such as Fe, Ni, Cu, and Zn, and their oxides in various valence states are dominant (Zhang et al., 2020a). Their self-doping can lead to the generation of oxygen vacancies that act as trap sites for holes, thereby facilitating the separation of photogenerated carriers (Zhang et al., 2021) and improving the PEC activity. Fe2O3 is one outstanding semiconductor with an ideal band gap (1.9–2.2 eV) structure (Leao-Neto et al., 2020). However, pure Fe2O3 has not been commonly used as an electrical conductor due to its excessively short vacancy diffusion length (2–4 nm) (Asif et al., 2021) and excited state lifetime (shorter than 10 ps) (Hannan et al., 2021). Cu2O is a p-type metal oxide semiconductor material, and it is stable and nontoxic (Wang et al., 2020a). Its 3d and 4s orbitals do not overlap, resulting in a semiconductor energy band structure with an empty conduction band, a full valence band (Li et al., 2020c), and a stereocrystalline configuration (Tan et al., 2019). Thus, Cu2O can precisely compensate for the easy compounding and low efficiency of Fe2O3 carriers in degrading pollutants (Polat, 2020). Cheng et al. (2021) prepared a CuO–Cu2O/WO3 film for the anode using a two-step deposition method and degraded oxygenated phenol in photocatalytic degradation with an efficiency of 87.6% after 180 min. Machreki et al. (2021) successfully developed porous Fe2O3 films for photocatalytic degradation of B41 dye wastewater, achieving a degradation efficiency of about 68% after 70 min. Given that experiments combining Cu2O and Fe2O3 are complex and expensive, studies have rarely been conducted on composites of Cu2O and Fe2O3. As a result, it is critical to develop materials that are simple to prepare with a high degradation efficiency.In this study, Fe2O3/Cu2O (FC) composites were produced by a simple one-pot hydrothermal process, and they were used as catalysts for efficient PEC degradation of BTA. Afterwards, the PEC degradation efficiency of BTA was determined, and the optimal operating conditions for the system and the degradation mechanism of BTA were investigated.Analytical reagents of CuCl2·2H2O, FeCl3·3H2O, polyethylene glycol-4000 (PEG-4000), CH3COONa, Na2SO4, H2SO4, NaOH, CH3CH2OH, and BTA were obtained from Aladdin Biochemical Technology Ltd (Shanghai, China). Deionized water was used for all experiments.In this study, FC composites were prepared using the one-pot hydrothermal method (Wang et al., 2021). Furthermore, appropriate amounts (molar ratios of 1:2, 1:1, 2:1, 3:1, and 4:1) of FeCl3·3H2O and CuCl2·2H2O were weighed, mixed with 40 mL of anhydrous CH3CH2OH, and stirred for 30 min. Next, 1 g of PEG-4000 and 3.6 g of CH3COONa were added and stirred vigorously for 4 h before being transferred to an autoclave at 200°C for 12 h of heating. The samples were filtered and cooled before being rinsed three times with distilled water and anhydrous ethanol to remove residual ions. They were then dried for 12 h at 60°C. The materials made were recorded as FC1/2, FC1, FC2, FC3, and FC4. Cu2O (or Fe2O3) was prepared without adding FeCl3·3H2O (or CuCl2·2H2O).Scanning electron microscopy (SEM) observation and energy spectrum analysis were performed using a field emission scanning electron microscope. X-ray diffraction (XRD) maps were obtained using a Bruker D8 advance X-ray diffractometer. X-ray fluorescence (XRF) analysis was performed using a Panalytical Axios FAST simultaneous wavelength dispersive XRF spectrometer (the Netherlands). X-ray photoelectron spectroscopy (XPS) analysis was conducted using a Thermo Fisher EscaLab 250Xi XPS analyzer. Ultraviolet–visible (UV–Vis) spectra were measured with a UV-2550 spectrophotometer (Shimadzu, Japan). Electrochemical measurements were performed with a CHI 660A electrochemical workstation (Shanghai Brilliance Instruments Co. Ltd., China) in a conventional three-electrode system. The modified electrode, platinum wire, and saturated glycerol electrodes were used as the working electrode, counter electrode, and reference electrode, respectively. The intermediates were analyzed using high performance liquid chromatography (HPLC) (Dinonex Ultimate 3000 UHPLC Column, Hypersil GOLD, 100 mm × 2.1 mm, 1.9 μm)–mass spectrometry (MS) (Thermo Scientific Q Exactive). HPLC l (C18 column, 5 μm, 4.6 mm × 150 mm, Agela ODS) was conducted for identification of the intermediates. Fig. A.1 shows the experimental setup for PEC degradation of BTA. The device used a quartz cylinder as the photoelectric reaction tank and used a titanium plate and a ruthenium-iridium loaded titanium plate (5 cm × 5 cm) as the cathode and anode, respectively (2 cm apart from each other). A 100-W high-pressure mercury lamp was used to simulate UV lamp irradiation with a primary emission wavelength of 365 nm. Circulating condensate was used to reduce the slight thermal effect of the UV lamp, and the reaction temperature was always maintained at room temperature. The simulated wastewater was mixed with 125 mL of BTA solution (20 mg/L) and 25 mL of Na2SO4 solution (0.2 mol/L). During the degradation process, BTA water samples were collected at different time intervals, and their concentrations were measured after filtration with 0.45-μm filters.The concentration of BTA was measured and analyzed using liquid chromatography (Wukong Instruments K2025) with a C18 column (with a wavelength of 268 nm). The removal efficiency (R) of BTA was calculated as follows: (1) R = C 0 − C t C 0 × 100 % where C 0 is the initial concentration of BTA; and C t is the concentration of BTA at the treatment time t. The chemical oxygen demand (COD) of BTA was determined according to the standard dichromate potassium method (HJ828-2017) proposed by the State Environmental Protection Administration of China.The optical properties of FC were studied using UV–Vis spectroscopy, and the band gap energy was calculated using the following equation: (2) α h v = A ( h v − E g ) η where α is the absorption coefficient; h is Planck constant; v is the frequency of light; A is the absorbance; E g is the band gap energy; and η is a variable depending on the nature of the optical leap, with η = 0.5 for the direct energy band leap and η = 2 for the indirect energy band leap (Wang et al., 2020b).The main degradation products were analyzed and identified using HPLC–MS. For HPLC, the mobile phase was 50% acetonitrile and 0.1% formic acid aqueous solution (0–10 min) at a flow rate of 0.2 mL/min, and its detection wavelength was 280 nm. For MS, the ion source electron spray ionization was used, the trans-gas rate was 40 mL/min, the auxiliary gas rate was 10 mL/min, the spray voltage was 3.0 kV, the capillary temperature was set to 300°C, the scan mode of S-lens50 was positive ion fullms-ddms2 top5, the resolution reached 70 000, and the scan range was 50–750 (mass-to-charge ratio).The SEM image of FC (Fig. 1 (a)) showed that many small disks of different sizes were attached to the cubic crystals. The rounded Fe2O3 increased the contact area for pollutants and easily trapped the electrons transferred to Cu2O in the cubic crystals in the reaction, thereby improving the degradation efficiency. The corresponding energy dispersive spectrometer (EDS) spectrum (Fig. 1(c)) confirmed the presence of Cu, Fe, and O. Additionally, the characterization analysis using XRF (Table A.1) showed that the contents of Fe2O3 and Cu2O agreed with those demonstrated by EDS, indicating the successful preparation of the composites. Fig. 2 shows the XRD characteristics of different materials. The diffraction peaks (Srivastava and Ingole, 2020) of α-Fe2O3 appeared at diffraction angles (2θ) of 24.20°, 33.14°, 35.67°, 40.82° 49.43°, 54.11°, 62.45°, and 64.02°, corresponding to Fe2O3 crystals of (012), (104), (110), (113), (024), (116), (214), and (300), and the main diffraction peak of Fe2O3 crystals appeared at 2θ = 35.67° (Sheikholeslami et al., 2020). Cu2O diffraction peaks (Sarto et al., 2019) appeared at 2θ = 29.54°, 36.50°, 42.24°, 51.98°, 61.34°, 73.61°, and 77.29°, corresponding to Cu2O cubic crystals of (110), (111), (200), (211), (220), (311), and (222) (Shi et al., 2019), and the main diffraction peak of Cu2O cubic crystals was at 2θ = 36.50° (Gaim et al., 2019). Clearly, the characteristic diffraction peaks of the composites covered all characteristic peaks of α-Fe2O3 and Cu2O, further confirming that Fe2O3 and Cu2O coexisted.The chemical states of the elements in the FC composites were analyzed using XPS. The XPS spectrum (Fig. 3 (a)) demonstrated the presence of O, Fe, and Cu elements in FC. These results were consistent with the EDS image (Fig. 1(c)). In addition, the fitted O 1s spectrum (Fig. 3(b)) contained four peaks at 531.9 eV, 530.6 eV, 530.1 eV, and 529.7 eV. 530.6 eV and 530.1 eV were mainly produced by the oxides of Fe (Bullen et al., 2020) and Cu (Liu et al., 2020), while the peaks at 529.7 eV and 531.9 eV were primarily attributed to lattice oxygen and surface hydroxyl groups. As shown in Fig. 3(c), the peaks of Fe 2p3/2 and Fe 2p1/2 were at 710.6 eV and 724.2 eV, respectively (Bagus et al., 2020), and the appearance of their satellite peaks at 718.4 eV indicated the presence of Fe3+ in the prepared samples (Imrich et al., 2021). Additionally, the detected peaks of Cu 2p3/2 and Cu 2p1/2 of Cu(I) were located at 932.3 eV and 952.3 eV, respectively (Fig. 3(d)) (Joseph and Sugunan, 2021), which were consistent with the peaks of Cu2O in Zhang et al. (2020b).The optoelectronic properties of the catalysts were further investigated in order to judge the suitability of FC for photocatalysis. The optical properties of FC were investigated using UV–Vis spectroscopy (Fig. 4 (a)). Pure Fe2O3 and Cu2O presented absorption fringes at 587 nm and 603 nm, corresponding to the band gap energies of 2.09 eV and 1.86 eV, respectively (Fig. A2). Fe2O3 had the same bad gap energy (2.09 eV) as that reported by Khasawneh et al. (2019), while Cu2O had a lower band gap energy than that (1.90 eV) reported by Li et al. (2020b). The decrease in Cu2O band gap energy was due to the appearance of d-orbit bands in the band gap. FC with a band gap of 1.96 eV not only improved PEC but also optimized the disadvantage of a narrow band gap prone to electron-hole complexation. Fig. 4(b) shows the photoluminescence (PL) spectral analysis of Fe2O3, Cu2O, and FC, which was used to determine formation, transfer, and complexation of photoexcited electrons and holes. At 472 nm, a strong emission peak was observed when excited by a 550-nm laser. The peak intensity of pure Fe2O3 was nearly twice that of FC, showing that electron-hole complexation occurred much more slowly in pure Fe2O3 than in pure Cu2O. The reduction in PL intensity of FC was due to the successful loading of Fe2O3 onto Cu2O where the built-in internal electric field at both loads provided an efficient path for charge carriers.The electrical properties were characterized by cyclic voltammetry (CV) curves. Prior to the test, the CV curves of the materials were determined using a Na2SO4 (0.2 mol/L) solution as the electrolyte. The CV curves of different materials (Fig. 4(c)) showed that the oxidation and reduction peaks of the pure materials were significantly lower than those of FC at −0.187 V and 0.253 V. This result shows that FC had a stronger redox activity due to the superposition of the valence changes of Fe3+/Fe2+ as well as Cu2+/Cu+ (He et al., 2017), and the two electron pairs interacted with each other to make the degradation of BTA more efficient.As shown in Fig. 4(d), there was no significant change in the photocurrent generated by all composites in the absence of light, and all photocurrents converged to zero. This indicated that photogenerated holes and electrons were not generated by the composites in the absence of light. The photocurrent generated by FC was much higher than that generated by the pure material. This confirmed that the separation of photogenerated electrons and holes was promoted by light exposure, which agreed with the PL analysis.The degradation of BTA in different processes was investigated. As shown in Fig. 5 (a), only 2% of FC was adsorbed after 90 min, meaning that FC had no adsorption effect on BTA. Under electrocatalysis with a current density of 20 mA/cm2, the removal efficiency of BTA was low, and 18.08% of BTA was removed in 90 min. This indicated that the degradation of BTA with electrocatalysis was not effective, a result that agreed with the findings of Li et al. (2021). The UV-catalyzed removal rate could only reach 26.25% for the same time period. This was mainly attributed to the fact that longer light exposure darkened the solution and reduced its light transmission, thereby preventing the degradation of BTA. The removal rate of BTA with PEC was significantly higher than under photocatalytic and electrocatalytic reactions. This was mainly because PEC generated superoxide radicals ( · O 2 − ) or oxidized radicals (·OH) and reduced the compound rate of photogenerated electron-hole pairs. This allowed these electron-hole pairs to release heat or migrate to the electrode plate, thereby reacting with nearby BTA. Therefore, PEC was suitable for the degradation reaction of BTA. Fig. 5(b) shows the effect of different materials on the degradation of BTA. The removal rate of BTA using FC under PEC treatment was 90.78%, significantly higher than with Fe2O3 and Cu2O (77.17% and 55.87%, respectively). This was due to the successful loading of Fe2O3 on Cu2O to reduce the electron-hole complexation rate. Fig. 5(c) shows the effect of the FC compounding ratio on the degradation of BTA. The degradation rate of BTA decreased significantly as the percentage of Cu2O increased. This was because the cubic crystal structure of Cu2O had a small contact area with BTA. The small circular shape of Fe2O3 compensated for this shortcoming of Cu2O with a large contact area with BTA, thereby significantly improving its catalytic efficiency through its internal electron transfer path. With an Fe:Cu molar ratio of 2:1, the highest BTA degradation rate reached 90.87%. However, when the percentage of Fe2O3 continued to increase, FC entered the solution and reduced the transparency of the solution. This hindered the photocatalysis and thus reduced the removal rate. Therefore, an Fe:Cu ratio of 2:1 (FC2) was considered the most suitable for subsequent experiments. Fig. 5(d) displays the effect of FC2 dosage on the degradation of BTA. When the dosage of FC2 was increased from 0 to 0.05 g/L, the 90-min removal rate of BTA increased from 54.94% to 94.87%. This was because the increase in the dosage enhanced the active site of FC, thereby augmenting the degradation rate of BTA. However, the removal efficiency decreased significantly with the further increase in FC2 dosage owing to the decreased solution transparency under high FC2 dosages. Therefore, 0.05 g/L was selected as the optimal dosage for subsequent experiments. The kinetic fitting curves (Fig. 6 (a)) showed that the reaction process was in accordance with the primary reaction kinetics, and the slope of the kinetic curve with a dosage of 0.05 g/L was higher than those with other dosages, demonstrating that FC2 showed an efficient PEC activity with a dosage of 0.05 g/L. Fig. 5(e) shows the effect of solution pH on the PEC degradation of BTA. The solution pH was adjusted using 0.1 mol/L of H2SO4 and NaOH solution. At a pH value of 3.06, BTA was completely degraded in 60 min. With the increase in pH, the removal efficiency decreased. This was due to the fact that BTA under acidic conditions is in a free state and can be easily degraded. In contrast, BTA under alkaline conditions has a molecular structure. As a result, the initial solution pH was adjusted to be 3.06 for further experiments. The kinetic fitting curves (Fig. 6(b)) showed that the reaction process accorded with the primary reaction kinetics, and the slope of the curve with a pH value of 3.06 was higher than those with other pH values. Therefore, FC2 demonstrated an efficient PEC activity at a pH value of 3.06. Fig. 5(f) shows the effect of current density on the degradation of BTA. When the current density was increased from 5 mA/cm2 to 25 mA/cm2, the removal efficiency of BTA increased because more photoelectrons were generated for the oxidative degradation of BTA, and strong oxidative radicals were generated to catalyze the degradation of BTA. When the current density was further increased, the degradation rate of BTA tended to be stable and did not increase. The kinetic fitting curves (Fig. 6(c)) showed that the reaction process fitted the primary reaction kinetics, and that the slopes of the curves with current densities greater than 20 mA/cm2 were higher than those with other current densities. This demonstrated that an efficient PEC activity was achieved using FC2 with a current density greater than 20 mA/cm2. Based on green energy-saving principles, 20 mA/cm2 was chosen as the optimal current density for subsequent experiments.In practical applications, the most important properties of PEC materials are their activity levels and their stability for long-term use. Stability experiments were performed by extracting, washing, and drying the recovered material. As shown in Fig. 7 , the removal rate of BTA using FC was maintained above 80.0% even after five repeated uses. This indicated that the FC catalyst was stable and retained a high level of PEC activity after repeated use. FC has significant potential for application in removal of organic wastewater with PEC.The COD removal rate can indirectly show the mineralization of organic pollutants, some of which are degraded to produce organic intermediates (Akintayo et al., 2021). Therefore, analysis of COD aids in understanding the degradation of organic pollutants (Can-Güven, 2021). Fig. 8 shows the removal efficiency of COD. The removal efficiency of COD was lower than that of BTA. This implied that intermediates formed during the BTA degradation. After 90 min of treatment, the removal rate of COD reached 96.61%. In contrast, the removal rate of BTA was already 100% at 60 min. This indicated that almost all degraded BTA and its intermediates were mineralized. Compared with photocatalysis, electrocatalysis, and PEC alone, PEC with FC as a catalyst was more efficient in degrading BTA and removing COD. · O 2 − , h+, and ·OH are the major active substances in the degradation of organic pollutants using PEC (Jiang et al., 2020). To understand the PEC mechanism of FC composite, the · O 2 − trapping agent p-benzoquinone, the photogenerated h+ trapping agent ammonium oxalate, and the ·OH trapping agent isopropanol were used to investigate the major active substances in the degradation of BTA using the PEC system (He et al., 2020). As shown in Fig. 9 (a), each trapping agent reduced the PEC removal efficiency of BTA to some degree. The addition of ammonium oxalate had a significant inhibitory effect on the removal of BTA, and the addition of isopropanol also had a certain inhibitory effect on the performance of PEC. Their 60-min removal rates were 47.2% and 55.1%, respectively. The experiment with p-benzoquinone had the least effect on PEC, with a removal rate of 71.8% after 60 min. As a result, h+, · O 2 − , and ·OH were all involved in the process of BTA removal using PEC, and the degree of involvement of the three free radicals in the oxidation of BTA was as follows: h+ > ·OH > · O 2 − . As a result, h+ in the composite PEC was mostly concentrated on the VB end of Fe2O3 rather than on the VB end of Cu2O (Liang et al., 2020) because the VB potential of Fe2O3 was higher than the VB potential of Cu2O (Fig. 10 ) (He et al., 2014).N element in BTA is presumably oxidized to NH3 and released into the atmosphere. Fig. 9(b) shows the generation of NH3-N during the degradation process. Clearly, the concentration of NH3-N increased as the concentration of BTA decreased. At 60 min, the concentration of NH3-N reached a maximum of 3.05 mg/L. At the same time, BTA was completely degraded. During the degradation of BTA, N element was finally released into the air in the form of NH3.To understand the photocatalytic degradation mechanism, UV full spectrum scanning and HPLC analysis were performed on the BTA solution. As shown in Fig. 9(c), the peak of the intermediate products appeared around 400 nm, which was determined to be the peak of N=N. This indicated that the intermediate product contained diazotrophic substances, which also contributed to the darkening with the degradation solution.As shown in Fig. 9(d), in addition to the major BTA peak, other small peaks appeared at 2.7 min and 3.8 min on the curves for 10 min, 20 min, 30 min, and 60 min but not on the curves for 0 min and 90 min. The peaks at 2.7 min were much larger than those at 3.8 min. This indicated that both peaks represented the intermediate products of BTA degradation. These intermediate products might be the substances that were not dehydrogenated after the opening of the ring. In addition, after 60 min of BTA degradation, the intermediate products were completely degraded during the remaining 30 min, with support from the change in the COD degradation rate.In order to investigate the possible degradation pathway of BTA, liquid mass spectra were measured in PEC for 40 min to observe the intermediates. As shown in Table A.2, diazo intermediates were produced during the degradation process, which was consistent with the UV full spectrum scan results showing intermediates after de-aminoamination. From this observation, the possible degradation pathway of BTA can be conceptualized as shown in Fig. 10. BTA exists in ionic form under acidic conditions, providing suitable conditions for PEC. PEC generates strong oxidation radicals for the oxidative decomposition of BTA. The radicals break the N–N bond, and an open loop forms for easy degradation. The oxidation of BTA by anode and materials occurs simultaneously during the degradation process. Thus, the reaction catalytic time is greatly reduced, resulting in cost savings and improving the degradation efficiency. Finally, BTA is oxidized to nontoxic H2O, NH3, and CO2. All these substances are discharged into the environment without producing secondary pollutants. Therefore, this FC-based PEC system provides a harmless and effective method for the degradation of BTA.In this study, composite FC was successfully prepared and used as a photocatalyst for the photocatalytic degradation of BTA. The results showed that, under optimal conditions, this photocatalytic treatment of 20-mg/L BTA for 60 min increased the BTA removal efficiency to 100%, and reduced the COD concentration of BTA by 96.61% after 90 min of treatment. After FC was added to the PEC system, it suppressed the compounding rate of photogenerated electrons and holes and improved the PEC degradation rate and mineralization rate of BTA, reflecting the strong photoelectric properties of FC. As a result, FC can be used as an effective catalyst for PEC. FC-based PEC provides an effective and promising method for degrading new organic pollutants.The authors declare no conflicts of interest.The following is the Supplementary data to this article: Multimedia component 1 Multimedia component 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.wse.2022.06.003.

Given the difficulties of degrading benzotriazole (BTA), this study used a one-pot hydrothermal method to prepare α-Fe2O3/Cu2O (FC) composites for photoelectrocatalytic (PEC) degradation of BTA. The characterization of FC structure showed that Cu2O in cubic crystals was loaded with circular sheets of Fe2O3. Owing to this structure, FC showed efficient PEC degradation of BTA when exposed to ultraviolet light. The experimental results demonstrated that FC efficiently degraded BTA. When the PEC degradation continued for 60 min, 100% degradation of BTA was achieved because FC enhanced the photoelectron-hole separation and the separation and transfer of articulated carriers. High performance liquid chromatography–mass spectrometry showed that intermediates formed during the PEC degradation of BTA. Finally, various pathways for degradation of BTA were postulated. This FC-based PEC system provides a harmless and effective method for degradation of BTA.

Data will be made available on request.Solid carbon is imperative for numerous applications, and its diverse properties stem from the many allotropes, and various compounds and composites that may be formed from carbon precursors, e.g. diamond, graphite, carbon fibers, cokes, carbon black, activated carbon, and carbon aerogels [1]. Some of these may in turn be surface-modified with functional groups to provide new properties [2,3]. The natural abundance of carbon presents a vast array of precursors for production of both mundane and advanced materials, in which heat treatment is often a key step. Pyrolysis, a term typically used when organic materials are involved, is a thermal treatment in an inert atmosphere, to prevent combustion and favor decomposition of the carbon precursor [4]. For advanced materials, the interest lies in the formation of solid carbon products, termed char. The pyrolysis conditions, such as temperature, heating rate, and dwell duration, are key variables for controlling the amount of char relative to gases and liquids, i.e., degree of carbonization. Char formation and degree of graphitization are also highly dependent on the carbon precursor and how it is treated, e.g., molecular structure, porosity, and the presence of impurities like metals [5].Porous carbons are utilized in several material applications, such as catalysis and energy storage, where transition metal nanoparticles are often confined within the carbon structure or pores [6]. Such materials have commonly been produced by impregnation or deposition-precipitation of pre-synthesized porous carbon structures, e.g. activated carbons, carbon nanofibers, and carbon nanotubes. Both methods are simple to execute, and impregnation grants control over metal loading but may result in poor metal dispersion at high metal loadings and pore blocking, which are often undesirable features [7]. A different approach is pyrolysis of carbon precursors with metal ions bound to their structure. A straightforward tactic is to use a polymer with appropriate functional groups that can bind metal ions, or advanced precursors such as metal-organic frameworks (MOFs), followed by pyrolysis [8,9].A polymer with great potential for this application is alginate – a biopolymer extracted from brown seaweed. Alginate is comprised of two monomers: (1–4)-linked β-D-mannuronate (M), and its C-5 epimer, α-L-guluronate (G) ( Fig. 1a). The spatial orientation and relative content of the two monomers will affect the interbonding and intrabonding properties in the presence of multivalent cations. The GG-diads have historically been of importance for cross-linking alginate chains, proposed to form so-called “egg-box structures” (Fig. 1b) [10]. This enables alginate to serve as a scaffold to obtain atomic dispersion of metal cations between the macromolecular chains.Studies of heating of metal alginates have focused on their flame-retardant properties [11–15] and the decomposition of alginate [16–18], typically only investigating the alginate or carbon by in situ characterization techniques such as thermogravimetric analysis (TGA) and Fourier-Transformed infrared spectroscopy (FTIR). The resulting materials after pyrolysis of metal alginates have been utilized for sorption of heavy metal ions and dyes from aqueous solutions [19–21], as anode materials [22,23], supercapacitors [24], electrocatalysts for oxygen reduction reaction (ORR) [25], and as catalysts for CO2 hydrogenation [26] and NOx abatement [27]. Specifications related to the M/G-content and monomer distribution in the alginate are rarely reported, and only alginate solutions with low concentrations have been utilized, less than 5 wt%. The pyrolysis temperature was typically in the range of 800–900 °C, resulting in well-carbonized carbon materials, but with extensive sintering of the metal particles, as there were no reports of particles below 10 nm. The metal species were usually only characterized after pyrolysis with ex situ techniques such as X-ray diffraction (XRD), Raman spectroscopy and transmission electron microscopy (TEM). The transition from metal-alginate to the final product remains unresolved but understanding the evolution of both carbon structure and metal species during pyrolysis is vital for tailoring material properties.In our previous work, the green and Na-alginate was ion-exchanged with different transition metal ions, followed by pyrolysis at 500 °C to produce [28]. In particular, pyrolysis of Fe-alginate yielded desirable material properties for heterogeneous catalysis. Due to the promising nature of this material, herein, the evolution of Fe and C species during pyrolysis has been corroborated by both ex situ characterization and advanced in situ techniques such as XAS and Mössbauer spectroscopy. This was mainly performed to understand how the pyrolysis conditions affect the material properties, but catalysts were also prepared by performing pyrolysis in a range of suitable temperatures to see how these material properties affect their performance in high-temperature Fischer-Tropsch synthesis (FTS). In FTS, synthesis gas (syngas, CO and H2) from biomass may react through a polymerization reaction to form green fuels and chemicals (hydrocarbons of varying lengths), and H2O as a by-product. Additionally, Fe catalysts exhibit water-gas shift activity, a reaction where H2O and CO form H2 and CO2. This enables Fe catalysts to use a syngas feed with low H2/CO ratios, which is typically the result when syngas is produced from carbon-rich feedstocks such as biomass.Sodium alginate (Protanal LFR 5/60: G monad frequency (FG) = 0.65–0.70, G diad frequency (FGG) = 0.5–0.6, average length of G-blocks (NG>1) = 11–20) was supplied by Dupont Nutrition Norge AS). Iron(III) nitrate nonahydrate (> 98 %) was supplied by Sigma-Aldrich, with impurities such as Cl- (< 5 ppm), SO4 2- (< 0.01 %), Ca2+ (< 0.01 %), Mg2+ (< 0.005 %), K+(< 0.005 %), Na+ (< 0.05 %). Ethanol (96 %) was supplied by VWR. Deionized water was produced by using a Milli-Q water purification system.Na-alginate was dissolved in deionized water by a magnetic stirrer to form a 20 w/w% alginate/water solution. A solution of 0.1 M Fe(NO3)3-solution was prepared, with five times greater volume than that of the alginate. The Na-alginate solution was dripped into the Fe(NO3)3-solution, which formed alginate beads on contact, and was kept in solution for 24 h. The alginate beads were then washed by placing them in 200 mL deionized water for 5 min, discarding and replacing the water, repeated three times before they were immersed in ethanol-water solutions of increasing concentration over time to gradually transform the beads from hydrogels to alcogels. The initial ethanol concentration was 10 %, which was discarded and increased by 20 % every 10th min up to 90 %, before finally leaving the beads in a 96 % ethanol solution for 24 h. The beads were collected from the ethanol solution and dried at 80 °C overnight, followed by mortaring. Approximately 1 g of the dried powder was placed in a calcination reactor that allows gas to pass through the sample. Pyrolysis was performed by using a heating rate of 2° min−1 in 100 mL min−1 N2 to the desired temperature, and dwelled at this temperature for a given time, as listed in Table 1. The samples were passivated in 1 % O2 in Ar for 2 h at room temperature after pyrolysis, which is important due to the samples’ pyrophoric nature.Powder XRD was recorded at ambient temperature with a Bruker D8 A25 DaVinci X-ray Diffractometer using a Cu Kα-radiation (λ = 0.15432 nm) X-ray tube and LynxEye™ SuperSpeed detector. The samples were scanned in the range 2θ = 10–80° for 60 min, using a 0.2° divergence slit. The powder diffraction files (PDF) used as standards were α-Fe (7-9753), FeO (PDF 6-615), γ-Fe3O4 (9-2285), γ-Fe2O3 (21-3968), χ-Fe5C2 (36-1248), θ-Fe3C (35-0772).N2 adsorption-desorption experiments were performed with a Micromeritics Tristar II 3000. Approximately 100 mg sample was degassed and evacuated for 24 h, at 353 K for the dried samples, and 473 K for the pyrolyzed samples. To determine the surface area, pore-volume, and pore diameters of the samples, the Brunauer-Emmett-Teller (BET) isotherm and Barrett-Joyner-Halenda (BJH) method (desorption) were used.Thermogravimetric analysis (TGA) with differential scanning calorimetry (DSC) was performed with a Netzsch Jupiter 449 unit. The analysis was performed in pure argon (100 mL min−1) or air (75 mL min−1), with a ramp rate of 10 °C min−1 , heating from RT to 900 °C. A QMS 403 C Aëolos quadrupole mass spectrometer (MS) was used to analyze the effluent gases.The elemental compositions of the iron alginate samples were measured by inductively coupled plasma mass spectrometry (ICP-MS). Between 10 and 30 mg of dried samples were mixed with 2 mL concentrated nitric acid (HNO3) in perfluoroalkoxyalkane (PFA) vials. Further, the samples were digested in an UltraClave, heated to 245 °C, and pressurized to 50 bar. The resulting solutions were diluted to a total volume of 216.6 mL, then 16 mL of this solution was sent for analysis along. Three blank samples were used to correct the results. The elemental analysis was performed with a High Resolution Inductively Coupled Plasma ELEMENT 2 connected to a mass spectrometer.The pyrolyzed samples were imaged by high-resolution (HR)-TEM. Experiments were performed with a JEOL JEM-2100 (LaB6-filament, side-mounted Gatan 2k Orius CCD) and a JEOL JEM-2100F (200k Schottky field emission gun (0.7 eV energy spread) and bottom-mounted Gatan 2k Ultrascan CCD) both with Oxford X-Max 80 SDD energy-dispersive X-ray (EDX) (solid angle 0.24 sr) and scanning option with bright-field (BF) and high-angle annular dark-field (HAADF) detector. For sample preparation, the samples were suspended in isopropanol, then deposited on a Cu-grid with lacey carbon.The D- and G-bands of the carbon in the pyrolyzed samples were analyzed with Raman spectroscopy. Experiments were performed with a Horiba ASD with a laser wavelength of 633 nm, employing a 600 g mm−1 grating, × 50 LWD objective, 15 acquisition, 3 accumulations, 25 % filter, and hole of 200. The powdered samples were placed on glass slides for analysis.Transmission 57Fe Mössbauer spectra were collected at different temperatures with conventional constant-acceleration or sinusoidal velocity spectrometers using a 57Co(Rh) source. Velocity calibration was carried out using an α-Fe foil at room temperature. The source and the absorbing samples were kept at the same temperature during the measurements. The Mössbauer spectra were fitted using the Mosswinn 4.0 program [29]. The experiments were performed in a state-of-the-art high-pressure Mössbauer in-situ cell – recently developed at Reactor Institute Delft [30]. The high-pressure beryllium windows used in this cell contain 0.08 % Fe impurity whose spectral contribution was fitted and removed from the final spectra. The experiment at 700 °C was performed in a standard tubular reactor and the pyrolyzed sample was measured quasi in-situ (via glovebox transfer).In situ XRD and X-ray absorption spectroscopy (XAS) was performed at the Swiss-Norwegian beamlines (BM31, European Synchrotron Radiation Facility (ESRF), France). The Fe-alginate sample was placed between two quartz wool plugs in a quartz capillary reactor (o.d. 1 mm), resulting in a bed length of 10 mm. The capillary was placed in an in situ cell, described elsewhere [31]. During the pyrolysis experiments, the capillary was purged with 10 N mL min−1 He, heating from RT (20 °C) to 700 °C at a rate of 2 °C min−1 at atmospheric pressure. The effluent gases that evolved were analyzed with an online MS. X-ray diffraction data were collected with a 2D plate detector (Mar-345) using monochromatic X-rays with a wavelength of 0.4975 Å. A lanthanum hexaboride (LaB6) standard was used as a calibration reference. X-ray absorption near-edge spectroscopy (XANES) and extended X-ray absorption fine structure (EXAFS) were measured at the Fe K-edge, ranging from 7.05 to 8.2 keV, in transmission mode. The XAS data were analyzed using DLV Excurve and Feffit. A linear pre-edge was subtracted, and the data were normalized by the edge-jump. The background was subtracted to yield the data in χ(k), which was k2-weighted before applying the Fourier-transform. EXAFS data were generally fitted in a k-range of 3.5–9.5 Å−1, due to the deterioration of the signal at high temperatures, and an R-range of 1.0–3.8 Å−1. For the amplitude reduction factor (S0), metallic Fe was extracted from the metallic Fe foil measurement, and the Fe-alginate (RT) was extracted from Fe(NO3)3. Coordination number (N), energy shift (ΔE0), scattering distance (R) and Debye-Waller factor (2σ2) were refined for each scattering path.The absorbances from the XANES measurements were placed in a matrix A with m rows (energy) and n columns (sample). The variance in the dataset was ranked by using singular-value decomposition (SVD), by the following equation: A = U × S + V where U (m x m) is the left singular matrix, S is the singular-value matrix and V T is the right singular matrix. The S matrix gives us the variance of each point in the dataset, for the initial evaluation of the number of pure components. Simple-to-use interactive self-modeling analysis (SIMPLISMA) was used to provide initial estimates for the pure components in the data set [32]. These estimates were then analyzed by multivariate curve resolution alternating least-square regression (MCR-ALS), to approximate the composition of components in a dataset with mixtures of components, using the equation [33]: A = C × S T + ε where C is the concentration of the relative concentration of the initial components contained in S, and ε is the error. The analysis assumes that the pure components have the lowest noise relative to the other spectra in the data. Constraints of non-negativity and normalization were applied, to obtain concentrations summing up to 100 %.Diffuse reflectance infrared Fourier Transform spectroscopy (DRIFTS) measurements of all samples were conducted in a Thermo Scientific Nicolet iS50 FT-IR Spectrometer with a Harrick Praying Mantis™ high temperature in situ cell, flushed with 30 mL min−1 Ar. A spectral range of 4000–600 cm−1 was used, with a resolution of 4 cm−1 and 32 scans being averaged for each spectrum. The samples were heated from RT to 500 °C at a rate of 5 °C min−1, halting the heating at every 50 °C to perform measurements.The FTS experiments were performed in a 10 mm i.d. tubular stainless-steel fixed bed reactor at 340 °C, 20 bar, and H2/CO = 1.0. The catalysts (0.10 g, 90–250 µm sieve fraction) were diluted and mixed with inert SiC (10 g, 90–250 µm sieve fraction) to minimize temperature gradients. To keep the catalyst bed fixed, the mixture was loaded into the reactor between two plugs of quartz wool. The reactor was mounted between two aluminum blocks in an electrical furnace to further improve heat distribution. The catalysts were reduced in H2 (100 N mL min−1) at 3 bars with a heating rate of 2° min−1 to 400 °C, with a dwell time of 3 h. Then the reactor was cooled to 330 °C and pressurized to 20 bar with 56,100 N mL g−1 h−1 H2. Syngas (48.5 % CO, 48.5 H2, 3 % N2) were then introduced in steps, replacing 25 % of the H2 flow every 5th minute, keeping the gas-velocity constant. The product stream was passed through a hot trap (90 °C) and a cold trap (25 °C) to collect condensable FT products, i.e., wax, light hydrocarbons, water, and oxygenates. The gas-phase products were analyzed with an Agilent Technologies 6890N gas chromatograph (GC) equipped with a stainless steel Carbosieve S-II and an HP-plot Al2O3 column with a thermal conductivity detector (TCD) and flame-ionization detector (FID). The 3 vol% N2 in the syngas mixture was used as an internal standard for the GC.The molar flows of H2O and H2 were estimated by using oxygen and hydrogen mass balances, where oxygenate formation was disregarded. The WGS equilibrium constant (Keq) was calculated by Keq = 10((−2.4198)+(0.0003855*(T)+(2180.6/T))), and was compared to the WGS quotient (QWGS) based on the fraction of products and reactants.The results are divided into two sections. First, ex situ characterization after pyrolysis of Fe-alginate at 400, 500, 600, and 700 °C will be regarded. These materials were also tested for FTS, to understand how the resulting material characteristics affect the catalytic performance. Second, in situ pyrolysis experiments were investigated to elucidate the evolution of both iron and carbon species.One batch of Fe-alginate was synthesized and split into four portions that were subjected to pyrolysis at different temperatures. The samples pyrolyzed at 400, 500, and 600 °C used a dwell duration of 8 h, whereas 1 h was selected for the 700 °C treatment to limit the extent of sintering.The porosity inherent in Fe-alginate can be attributed to the interconnection of alginate macromolecules, which is facilitated by ion-exchanged Fe3+. The measured BET (Brauner-Emmett-Teller) specific surface area of Fe-alginate was 178 m2 g−1, with pores primarily in the mesopore size range. After pyrolysis, a significant increase in BET surface area was observed for every step increase in temperature (Table 1, Fig. S1). In all cases, the pyrolysis treatment yielded pore sizes larger than for the Fe-alginate, and the total pore volume was also observed to increase with increasing temperature (Fig. S2). The observed changes in porosity can be attributed to both advancement of polymer decomposition and the development of by-products, which can lead to both expansion and formation of new pores.The dried Fe-alginate sample contained 10.3 wt% Fe, dictated by the amount of Fe3+ that can be crosslinked within the alginate gel. The decomposition of alginate constituents innately increased the Fe loading with increasing pyrolysis temperature. Relatively to Fe-alginate, the Fe loading doubled for P400, while for the other samples, roughly a threefold increase was achieved (Table 1). This indicates that a significant decomposition of alginate occurred above 400 °C.The powder X-ray diffractograms of the pyrolyzed samples, cooled to ambient temperature and exposed to air, are presented in Fig. 2. Broad and indistinct diffraction peaks were discerned for P400, which may originate from magnetite (γ-Fe3O4) or maghemite (γ-Fe2O3). These iron oxides are isostructural, the only difference is that magnetite has 2/3 Fe3+ and 1/3 Fe2+, whereas maghemite contains Fe3+ ions exclusively. This grants two highly similar diffraction patterns that are difficult to differentiate with the low-range ordering present in the present samples. Similar diffraction patterns were also present in P500 and P600, more defined yet still broad, indicative of increasing ordering due to particle growth. With increasing pyrolysis temperature, diffractions corresponding to ferrite (α-Fe) emerged, first for P500, then with an increase in intensity for P600, while the oxide diffraction diminished. If present, oxide phases were not discernible for P700, but the Fe species in the sample had reduced drastically and partially carburized into cementite. The most intense α-Fe diffraction peak, Fe(110), at 2θ = 45° overlaps with θ-Fe3C(211) and (103), but the Fe(200) diffraction peak at 2θ = 65° confirmed the presence of α-Fe.Images captured with TEM ( Fig. 3) showed amorphous carbon structures containing densely packed spherical Fe particles. As the average Fe particle size for P400 and P500 was relatively similar (Table 1), the increase in pyrolysis temperature from 400 to 500 °C did not lead to significant particle growth. The analysis of crystallinity and lattice fringes was not straightforward due to the size of the Fe particles. Larger particles were observed for P600, where some iron particles were completely reduced (Fe/FexCy), and some had a reduced core and a surrounding oxide shell, supported by an EDS line scan (Fig. S3). The EDS scan revealed a higher S density on the Fe particles, while Na was distributed more evenly in the sample, which is sensible, as S originated from the Fe precursor and the Na from the alginate. As the density difference of Fe atoms in iron oxides and metallic structure change the extent of electron transmission, darker areas indicate reduced Fe. Even larger particles and a broader particle size distribution were observed for P700 (Fig. S4). Similar Fe particle structures to P600 were seen for P700, but darker particles were also observed, indicative of more reduced Fe.The D-band (1350 cm−1) and G-band (1580 cm−1) region were measured with Raman spectroscopy, to investigate the bonding modes present in the carbon support. The relative intensity of the bands (ID/IG) is commonly used to quantify defects in graphitic and diamond-like structures, but the pyrolysis temperatures employed here are not likely sufficient to form graphite. Thus, the assignment of D- and G-band could be misleading. A study on pyrolysis of saccharose showed that the ID/IG ratio increased up to 2000 °C [34]. The ID/IG ratio of the pyrolyzed samples ( Fig. 4) increased with increasing temperature, similar to the aforementioned study, and also a blue-shift of the D-band and a red-shift of the G-band. An increasing ID/IG ratio in this temperature regime has been associated with increasing size of the structural carbon units [35]. The carbon material appears to be amorphous, but the increased ordering of the carbon structure was observed between 400–500 °C, and 600–700 °C.The samples were reduced in H2 at 400 °C for 3 h before being tested as catalysts at high-temperature FTS conditions, where a high initial CO conversion level was observed for all samples ( Fig. 5a). Because identical sample amounts were used, the samples with higher Fe content achieved a higher conversion level. However, the activities in terms of iron-time yield (molCO gFe −1 h−1) showed an inversely proportional relation to the pyrolysis temperature (Fig. 5b), where P400, P500, and P600 obtained similar activity profiles and the deactivation mainly occurred during the first 40 h on stream. Before reaction, P700 was the only sample containing θ-Fe3C, and the induction period experienced by this catalyst may be related to a slower interconversion of θ-Fe3C to χ-Fe5C2. As both FTS and WGS are highly exothermic, the high activity of P400 resulted in higher temperature (Fig. S12), which in turn increases the activity further. The oven temperature was unchanged during the first 100 h of stream due to the high heat of inertia in our system and to observe the deactivation profile. To obtain comparable data, the temperature was adjusted to 340 °C after 100 h, as reported in Table 2.Shorter hydrocarbons were produced by P400 and P500, compared to P600 and P700, but the C2-C4 olefins and paraffins selectivities were comparable for all samples. The changes in selectivity during reaction (Fig. S8), show similar results as discussed above, but P700 differs from the others by not stabilizing after 40 h on stream. All samples obtained the same CO2 selectivity, which also implies that the WGS activity is comparable, as the amount of CO2 is directly related to the former.At no point did the catalysts reach WGS equilibrium (Fig. S10).For comparison, Table 2 contains FTS performance data for relevant carbon-supported catalysts from literature. The samples in the current work have fair activity compared to the referenced works, but they do not have particularly specific selectivity. Compared to the previous work on this type of catalyst (Fe/15C [28]), a higher activity and olefin selectivity was reported due to higher Na and S content from the catalyst synthesis. Sodium and sulfur were not added to any of these catalysts by intention, but a much higher olefin selectivity could be achieved by optimizing the promoter amount, akin to the iron catalyst support on graphene oxide and promoted with potassium (Fe/K1-rGO [36]). Optimization of the catalysts for FTS is beyond the scope of this work.The diffractograms of the spent P400, P500 and P600 were similar, containing broad diffractions of χ-Fe5C2 and γ-Fe2O3, while P700 obtained more defined χ-Fe5C2 diffractions and a less prominent oxide phase (Fig. S7). At high-temperature FT conditions, χ-Fe5C2 is considered the active phase and is also the thermodynamically favored iron carbide.The measured particle size from TEM images of the spent samples (Fig. S6) revealed a relatively similar number-average particle size (± 95 % confidence interval), with sizes of 12.6 ± 0.3 nm (P400), 15.2 ± 0.5 nm (P500), 15.1 ± 0.3 nm (P600) and 15.6 ± 0.4 nm (P700). The exception was P400, which experienced less sintering than the other sample, likely due to lower metal loading. As the core is much darker than the surrounding shell, there is an indication that the spent particles had an α-Fe or χ-Fe5C2 core and an outer layer comprised of iron oxide.Mössbauer spectroscopy on spent samples that had been exposed to air revealed that the amount of χ-Fe5C2 increased from 12 % to 15 % for P400, P500, and P600, up to 38 % in P700, while all samples also contained 6–8 % ϵ’-Fe2.2C (Table S3). Deducing the extent of carburization during reaction from the contents of samples exposed to air is questionable, however, it seems reasonable that P700 achieved a greater extent of carburization. Before reduction in H2 and introduction of syngas in FTS, P700contained partially reduced and carburized particles, as seen from XRD (Fig. 2) and Mössbauer (Table S2), which when activated yielded more χ-Fe5C2 than the other treatments after over 100 h on stream. No apparent catalytic activity benefit was observed from the increased content of χ-Fe5C2. The oxide contributions originated from γ-Fe2O3, but with various particle sizes. For P700, only very small and superparamagnetic (SPM) γ-Fe2O3 was observed, while for the other samples, intermediately sized γ-Fe2O3 was observed in addition.An in situ Mössbauer experiment of the P500 sample after reduction and 20 h of FTS, showed that the catalyst consisted of 18 % χ-Fe5C2 and 82 % Fe1−xO. The fitting parameters show that a relatively crystalline χ-Fe5C2 phase was formed, while the oxide phase was relatively disordered. After the experiments, the sample was exposed to air and measured again, where the sample contained only 8% χ-Fe5C2 and the remainder Fe3+, due to oxidation (Table S2). The linewidth of χ-Fe5C2 increased significantly upon oxidation after reaction, indicating that the crystalline ordering decreased.The mass-loss and the accompanying effluent gases from the pyrolysis of the Fe-alginate were investigated using TGA-DSC coupled with MS. The sample was heated from RT to 900 °C in Ar at a rate of 10 °C min−1 ( Fig. 6). The decomposition was divided into four different segments, where the first segment (I) involved dehydration (H2O; m/z = 18) only, with an endothermic DSC signal. This process peaked between 100 and 120 °C, and the total dehydration yielded a mass-loss of 10 %. The second segment (II) took place between 160 and 350 °C, with the largest mass loss (50 %) throughout the entire experiment, the main effluent gases being H2O and CO2 (m/z = 40). This was followed by the third segment (III) between 350 and 550 °C, where a minor decomposition of 11 % mass loss occurred, with more CO2 than H2O evolving, compared to the previous segment. In the final segment, another mass loss of 11 % was observed, which was assigned to the reduction of the iron oxide, an endothermic process. The mass loss can be attributed to the carbon acting as a reductant, resulting in the release of both CO (m/z = 28) and CO2. A study of the pyrolysis of Fe2+-biopolymers (gelatin, chitosan, and alginate) demonstrated that a reduction of FeO to α-Fe or θ-Fe3C was observed around 650 °C [39].Mössbauer measurements were conducted after Fe-alginate was subjected to different heat treatments in Ar ( Fig. 7, Table S1, Fig. S13) and subsequently cooled to 120 K (300 K for samples subjected to air after treatment). The untreated Fe-Alginate contained 90 % Fe3+ and 10 % Fe2+, which was almost unchanged after heating the sample to 100 °C. At 200 °C, Fe2+ was almost exclusively observed, which was also the case at 300 °C, but small amounts of superparamagnetic (SPM) α-Fe could be fitted. For very small particle sizes α-Fe becomes SPM, when about half of their atoms are located at the surface, resulting in a nanoparticle that acts as a single magnetic domain. The quadrupole splitting (QS) of Fe2+ reduced as the temperature increased from 100 °C to 200 °C, implying higher charge symmetry at a higher temperature due to the removal of H2O ligands during drying.After treatment at 400 °C and a dwell time of 8 h (P400), the fraction of α-Fe (SPM) increased to 25 %, while the remainder consisted of Fe2+. Two different Fe2+ species were observed, where one was fitted as Fe1−xO (6 %) as it has a hyperfine field and a QS close to zero. The remainder (69 %) has a very different QS (1.42 mm s−1) than the measurement at 300 °C (2.54 mm s−1), and due to having no hyperfine field, it was assigned to very small particles of Fe1−xO that are superparamagnetic (Table S1), adding up to a total of 75 % Fe1−xO. Concerning this assignment of Fe1−xO, this entails that the Fe2+ species with high QS observed at 200 °C (2.72 mm s−1) and 300 °C are not Fe1−xO, and the low charge symmetry implies that these are Fe2+ ions still bound to alginate.The heat treatment at 500 °C with no dwell time yielded the same composition as for 400 °C for 8 h, but with species having smaller line widths, which could be due to higher crystallinity as an effect of higher temperature. However, dwelling at 500 °C for 8 h (P500) halved the number of Fe2+ (35%), and in addition to the SPM α-Fe (16 %), there was also SPM FexC (25 %) and α-Fe (24 %). The formation of the latter indicates that an increase in temperature has led to some particle growth, as some α-Fe has lost its SPM properties. When this sample was cooled down and exposed to air, it was almost completely oxidized (96 % Fe3+), containing small amounts of α-Fe (4 %) (Table S2). A sample was also measured after treatment at 700 °C for 1 h (P700), which yielded large quantities of θ-Fe3C (68 %), some α-Fe (24 %), and small amounts of SPM FexC (4 %) – the sample is completely reduced. Upon exposure to air, the sample was partially oxidized to Fe3+ (43 %), but the θ-Fe3C (51 %) appeared to be difficult to re-oxidize.The synchrotron experiment was performed by heating Fe-alginate from RT to 700 °C, while alternately recording XRD and XAS ( Fig. 8, Fig. S16), with an experimental objective to elucidate the fate of the Fe species during pyrolysis. Initially, the dried Fe-alginate sample was XRD-amorphous – the broad diffraction peak around 21° originating from the quartz capillary, but also overlaps with alginate (002) [40]. At 380 °C, very low-intensity diffraction peaks corresponding to FeO appeared, which increased in intensity with increasing temperature, indicating growth of the FeO particles. At 634 °C, FeO diffractions were promptly transformed into α-Fe and were accompanied by the release of CO – a stable signal until this point – and some CO2 (Fig S14). At the next measurement (660 °C), sharp α-Fe diffractions were observed, indicating a rapid reduction and particle growth.The initial XANES measurements exhibited a high absorption edge threshold energy (E0, maximum of the derivative signal), indicative of Fe3+ – the same species introduced into the alginate matrix in the material synthesis. At 150 °C, E0 was lowered, implying a reduction of Fe3+ to Fe2+, accompanied by a shift of the pre-edge to lower energy. Approaching 350 °C, E0 did not change significantly, but the spectra developed more characteristic features. The pre-edge also gained a wide feature towards the main edge, which is characteristic of FeO but may also have a contribution from broad features of the α-Fe edge. The next significant change was observed at 630 °C, where the spectra rapidly transformed to features that are distinct for α-Fe.A total of 189 XANES spectra were measured over the temperature range, and MCR in conjunction with SIMPLISMA and SVD was utilized to estimate compounds that are not among the measured standards, and their concentration. Many pure components may be obtained, but only those that can account for a significant degree of variance, have meaningful spectra, and are significantly different from the other components, were used. This evaluation resulted in four calculated components, which are shown in Fig. 8(b) along with the concentration of these components over the temperature range (Fig. 8(d)) The pure components were matched with measured standards, where component 1 had an edge position similar to α-Fe2O3 and Fe(NO3)3, however, the spectrum was relatively featureless compared to both of these, and was assigned as Fe3+ bound to alginate. Components 2 and 4 are somewhat similar in terms of edge position, but component 4 has features that are very close to FeO and were assigned accordingly. Component 2 contained less prominent features, similar to component 1, and had a more defined pre-edge that was shifted towards lower energies, and is therefore likely a Fe2+ species, as it does not resemble any spectrum among the measured standards. Component 3 has features matching with the Fe-foil and was assigned to α-Fe. Including more components than the four chosen here, yielded several variations of component 3 (α-Fe), where some may be linked to iron carbides. The concentration plot serves to give an estimate of the contributions of the selected components during the pyrolysis experiments. Some contributions, like α-Fe at the beginning and Fe3+ species between 400 °C and 600 °C, are not realistic but they provide the best fit given the limited set of components.An in situ DRIFTS experiment was performed to investigate the pyrolysis temperature’s effect on the alginate structure. The measurement obtained at RT ( Fig. 9(a)), shows the bands present in Fe-alginate, which were; the broad band ranging from ~3000–3600 cm−1 (O-H stretching mode, ν(O-H)s); a weak signal located at 2938 cm−1 (aliphatic C-H stretching mode, ν(C-H)s); the intense peaks at 1629 cm−1 (antisymmetric COO stretching mode, ν(COO)asym) and 1429 cm−1 (symmetric COO stretching mode ν(COO)sym); the weak band at 1236 cm−1 (C-C-H and O-C-H deformation, δ(CCH) and δ(OCH)); the intense bands between 1133 and 1109 (C-O stretching vibrations of the pyranose rings ν(C-O)s); and the band at 1054 (C-O or C-C stretching, ν(C-O)s and ν(C-C)s). The band at 1749 cm−1 was not observed for Na-alginate nor when gelated with divalent cations and does not appear to match with a carboxylic acid, and we have previously proposed that it might be related to an ester [41].The Fe-alginate was heated from RT to 500 °C in He, measuring DRIFTS spectra every 50 °C. The absorbance was calculated by using the single-beam data of the sample measured at RT °C as I0, to visualize the change towards 500 °C (Fig. 9(b)), while the raw single-beam data can be found in Supporting information (Fig. S18). The reduction of the wide ν(O-H)s band shows that water was gradually removed with increasing temperature, but the shoulder at 3480 cm−1 persisted until 200 °C. The position of this shoulder indicates intermolecularly bonded hydroxyl groups, perhaps linked to Fe-species, but it ultimately diminished when transitioning to 300 °C. The absorbance in the O-H region dropped severely from 250 to 350 °C. The absorbance of ν(C-H)s increased initially, but was diminished at 350 °C and appeared to be removed as the temperature reached 450 °C, at which point the signal deteriorated rapidly.In the region of 1800–1000 cm−1, the overall absorbance increased initially but reduced with increasing temperature after 150 °C. The most dramatic loss of overall absorbance was observed between 300 °C and 350 °C, affecting the entire region. A band appeared at 1767 cm−1 at 50 °C, accompanied by a gradually increasing band at 1801 cm−1. We assign the bands at 1767 cm−1 and 1801 cm−1 to the formation of an acid anhydride, where the peaks correspond to the symmetric and asymmetric stretching vibrations of the carbonyl groups, respectively. The higher intensity observed for the asymmetric peak indicates that the acid anhydride forms intermolecular bonds between the alginate chains, via the association of two carboxylate groups [42].The ν(CO)s at 1749 cm−1 and ν(COO)sym were the first bands in this region to decrease in absorbance, between 200 and 250 °C. Simultaneously, new bands appeared at 1706 cm−1 and 1596 cm−1, and 1510 cm−1. The band at 1596 cm−1, which we assign to ν(CC)s, was removed by 350 °C and appears to be an intermediate structure. The change for ν(COO)asym is difficult to assess precisely due to overlapping bands. The bands at 1510 cm−1 and 1706 cm−1 have frequencies that indicate ν(CC)s,skeletal and ν(CO)s, respectively [43, 44], and they persisted throughout the investigated temperature range. There were also bands emerging at 1390 and 1200 cm−1, in the region where O-H bending and C-O stretching is typically located.The Fe3+ introduced during the synthesis step was also observed when characterizing the dried Fe-alginate. Distinguishing the interaction between alginate and Fe3+ requires a powerful tool that may investigate the local alginate structure, such as nuclear-magnetic resonance (NMR). Unfortunately, the paramagnetism of the iron species makes high-resolution NMR unfeasible. The interaction of divalent transition metals with alginate is well-studied, but these results do not translate for trivalent cations. Not only is the binding strength between alginate and trivalent cations much higher than for the divalent cations, but the ionic radii are also of importance - a smaller ionic radius allows for tighter intermolecular interactions between the alginate macromolecules. Investigations of Al3+-alginate by NMR showed two different octahedral six-fold coordination sites [45]. The Fe3+ ion is slightly larger than Al3+, but similarities in binding mode can be expected. In our previous work, Fe3+-alginate was compared to alginates cross-linked with divalent cations, where the latter had more distinct ν(COO)asym and ν(COO)sym than Fe3+-alginate [41]. An EXAFS measurement was performed of Fe-alginate at RT (Table S4, Fig. S15), which indicated that the Fe atoms on average were coordinated with 6 oxygens. However, the oxygen can stem from both carboxyl and hydroxyl groups, and it is also difficult to differentiate light scatterers such as oxygen and carbon.During pyrolysis, the fate of the iron and alginate was linked, due to their initial interaction in Fe-alginate. At first, water was removed by drying as seen from the first mass-loss segment in TGA, but the iron species appeared unchanged. The reduction of Fe3+ to Fe2+ took place between 150 and 200 °C without involving the common reduction route through γ-Fe3O4. Instead, a direct reduction to Fe2+ was observed, seemingly while still bound to alginate. The reduction of Fe3+ liberated some carboxyl groups and might have resulted in the formation of intermolecular bonded acid anhydride. This was the beginning of the largest mass-loss (50 %) segment in TGA that started at 160 °C, and lasted up to 350 °C, encompassing H2O formation due to loss of hydroxyl groups and possibly also due to dehydration during the development of acid anhydrides and freeing up of carboxylates. The formation of FeO occurred in the range of 200–400 °C, which further destabilized the carboxyl groups, and possibly also hydroxyl groups, but crystallites with sufficient ordering were observed at 380 °C. This mass-loss was also observed for divalent metals in the same temperature range [41]. New CC and CO double bonds were also formed, while the bands related to the pyranose rings were severely reduced by 400 °C, implying that the ring structures of the monomers are fractured or transformed, which is consistent with the third mass-loss (11 %) segment.The carbon structure was further investigated by observing the D-band and G-band, which indicated that the carbon support carbonized by removing heteroatoms and formation of small carbon subunits up to 700 °C. Above 630 °C, the extent of carbonization appears to be sufficient to let carbon act as a reductant, as FeO was rapidly reduced to Fe, releasing CO and CO2 that resulted in a mass-loss of 11 %. Our previous investigations of the pyrolysis of metal-alginates with Fe, Co, Ni, and Cu, showed that Fe-Alginate was the only sample to have a mass loss in TGA after 550 °C, accompanied by CO and CO2 loss. The in situ XRD showed only clear diffractions of α-Fe, while both Mössbauer and XRD of the passivated P700 sample indicated that θ-Fe3C was also formed. Differentiating α-Fe to θ-Fe3C with XANES is not straightforward, but its absence during in situ XRD could be due to the low ordering of the FexC/θ-Fe3C and might require dwelling at 700 °C to develop.Analysis of the spent samples indicated that the Fe-alginate sample pyrolyzed at 700 °C was carburized to a greater extent and was able to activate more iron particles than those treated at lower temperatures. Yet, the samples pyrolyzed at low temperatures yielded the highest catalytic activity. A correlation was observed between catalytic activity at the end of FTS experiment and the average particle size in the spent sample (Fig. S10) – the sample pyrolyzed at 400 °C achieved noticeably lower Fe loading (20 %) than the others (27–33 %), due to less decomposition of the support. This was beneficial, as low metal loading reduced the proximity of the metal particles, effectively lowering the extent of sintering.All catalysts exhibited the same iron phases during reaction, although with different compositions. The active phase during reaction is χ-Fe5C2, but having a higher extent of carburization did not seem to be beneficial – the greater extent of carburization for P700 did not yield higher activity. The catalytic stability of P700 seemed to be the best due to the sample having relatively large particles before reduction and reaction, and therefore the loss of catalytic activity due to sintering was limited. The in situ FTS measurement of the sample pyrolyzed at 500 °C indicated that the sample consisted of 18 % χ-Fe5C2 after 20 h of FTS. Thus, the relatively small amount of the total iron that is carburized and responsible for the catalytic activity of P500 must be highly dispersed. Sintering should be limited to maintain the activity, as it appears to be the main cause of deactivation.While there were no apparent differences in olefin selectivity, more short-chain hydrocarbons were produced by the catalysts pyrolyzed at lower temperatures. All samples had the same amount of Na and S relative to Fe, but higher pyrolysis temperatures could potentially increase the segregation of sulfur to the surface of the Fe particles, but due to the low amount of S, it could not be quantified by XPS. Relative to the previous study (0.13 wt% Na and 0.19 wt% S) on this type of catalyst, a more efficient washing procedure was employed, which lowered the Na content to 0.05 wt% and reduced the C2-C4 olefin/paraffin (O/P) ratio from 2.0 to 1.0 [28]. The selectivity of C2-C4 olefins could therefore be improved by the addition of Na by incipient wetness impregnation. For this type of study, it is important to take into consideration that the Na and S originating from the precursor in the synthesis step are distributed differently in the material than if it is added to the pores of the support by impregnation.All the treatments performed on Fe-alginate resulted in materials with desirable catalytic properties, as the events that led to the decomposition of the alginate structure and initiated the carbonization took place around 400 °C. P400 obtained the most desirable catalytic properties for FTS, with high activity and without significant changes in hydrocarbon selectivity.The pyrolysis process of Fe-Alginate to form carbon-supported iron catalysts was investigated. The in situ characterization revealed that the Fe3+ species in alginate reduce to Fe2+ around 180 °C, at the same time as the alginate starts to restructure and decompose. The formation of FeO crystallites led to the loss of carboxyl and hydroxyl groups, and new CC and CO bonds in the alginate residues. At temperatures exceeding 630 °C, a complete reduction to α-Fe took place, where the carbon in the support acts as a reductant, with the observed release of CO and CO2. The resulting carbon support is formed by the deterioration of the alginate, and the most critical events that aid the formation of a catalyst with desirable properties took place at temperatures up to 400 °C. The results provide valuable knowledge to the rational design of metal-alginate-based materials with tailored structures and properties for various applications.Catalysts were synthesized by performing pyrolysis at temperatures between 400 and 700 °C, all of which resulted in appreciable material characteristics for heterogeneous catalysts. With increasing pyrolysis temperature, the particle size increased, as well as the reductive nature of the treatment, and a larger surface area but smaller pores were formed. All the catalysts exhibited great performance in high-temperature FTS, but higher catalytic activity was observed for catalysts synthesized at milder temperatures due to restricted alginate mass-loss. This resulted in lower Fe loading, which limited particle growth and had a beneficial effect on the catalytic surface area and activity. A greater extent of carburization during FTS was observed for the sample pyrolyzed at 700 °C, but this did not enhance the catalytic activity nor selectivity to a significant extent. Joakim Tafjord: Conceptualization, Methodology, Validation, Software, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization. Samuel K. Regli: Methodology, Software, Formal analysis, Investigation, Writing – review & editing. Achim Iulian Dugulan: Methodology, Formal analysis, Investigation, Data curation, Writing – review & editing. Magnus Rønning: Investigation, Resources, Supervision, Writing – review & editing. Erling Rytter: Conceptualization, Writing – review & editing. Anders Holmen: Conceptualization, Writing – review & editing. Rune Myrstad: Resources, Writing – review & editing. Jia Yang: Conceptualization, Investigation, Resources, Writing – original draft, Writing – review & editing, Supervision, Project administration, Funding acquisition.The manuscript was written through the contributions of all authors. All authors have approved the final version of the manuscript.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.The authors would like to acknowledge support from the Research Council of Norway through the Norwegian Center for Transmission Electron Microscopy, NORTEM (197405/F50); the iCSI (industrial Catalysis Science and Innovation) Centre for Research-based Innovation (Contract no. 237922); and the Swiss-Norwegian Beamlines at ESRF (Grant no. 296087). Funding from the Norwegian University of Science and Technology (NTNU) is also acknowledged. The assistance of Ljubisa Gavrilovic and the staff at the Swiss-Norwegian Beamlines at ESRF during the synchrotron beam-time is also acknowledged.Supplementary data associated with this article can be found in the online version at doi:10.1016/j.apcata.2022.118834. Supplementary material

Transition metals supported on carbons play an important role in catalysis and energy storage. By pyrolysis of metal alginate, highly active catalysts for the Fischer-Tropsch synthesis (FTS) can be produced. However, the evolution of the carbon (alginate) and transition metal (Fe3+) during pyrolysis remains largely unknown and was herein corroborated with several advanced in situ techniques. Initially, Fe3+ was reduced to Fe2+, while bound to alginate. FeO nucleated above 300 °C, destabilizing the alginate functional groups. Increasing temperatures improved carbonization of the carbon support, which facilitated reduction of FeO to α-Fe at 630 °C. Catalysts were produced by pyrolysis between 400 and 700 °C, where the highest FTS activity (612 µmolCO gFe −1 s−1) was achieved for the sample pyrolyzed at low temperature. Lower metal loading, due to less decomposition of alginate, moderated sintering and yielded larger catalytic surface areas. The results provide valuable knowledge for rational design of metal-alginate-based materials.

The typical strategies for NOx removal in automotive exhaust include Three-way Catalytic Conversion (TWC), NOx Storage Reduction (NSR) and Selective Catalytic Reduction (SCR) technologies [1,2]. A three-way catalyst requires to be operated under near stoichiometric conditions, while NSR and SCR catalysts are functioning for NOx removal during lean-burn operation. In the SCR technique, the reducing agent must be injected to react with NOx over the SCR catalyst, which demands enough space and complicated control system [3]. As for lean-burn gasoline and light-duty diesel vehicles, use of SCR system is limited due to the confined space. However, an NSR catalyst is very suitable to effectively remove NOx in the exhaust of vehicles equipped with lean-burn gasoline or light-duty diesel engines [3,4]. Zhang [5] et al. reported that addition of Mn into the traditional Pt/Ba/Al2O3 catalyst can significantly improve NOx removal efficiency, owing to the increased activity for NO oxidation and the superior NOx storage efficiency.From the perspective of environmental catalysis, a high N2-yield is the expected result for NOx removal. The PtRh bimetallic NSR catalysts have been widely employed to ensure the effective NOx conversion towards N2. The benefit of the bimetallic Pt/Rh/BaO/γ-Al2O3 catalyst with a Rh/Pt weight ratio of 0.5 is to largely accelerate the release of stored NOx and significantly promote the NOx reduction activity, but the simultaneous presence of Pt and Rh lightly suppresses the NO oxidation and NOx storage under lean-burn conditions [6]. The bimetallic separated Pt/Al2O3 + Rh/BaCO3 catalyst with a Rh/Pt weight ratio of 1 shows a better NSR performance, in comparison with that of the bimetallic Rh/Pt/BaCO3/Al2O3 catalyst [7]. Castoldi group [8] found that the Pt/Rh-Ba/Al2O3 catalyst with a Rh/Pt weight ratio of 0.5 shows higher selectivity of N2O and NH3 by-products than Pt-Ba/Al2O3 in the rich-phase during isothermal lean-rich cycles, resulting in a poorer N2-selectivity. Addition of ceria (CeO2) can improve the low-temperature NOx storage and reduction performance [9].However, it is still unclear how the Rh/Pt weight ratio impacts on the overall NOx removal efficiency and nitrogen yield, which limits the rapid and rational design of a highly efficient NSR catalysts. In the present work, a series of Pt/Rh bimetallic NSR catalysts with different Rh/Pt weight ratio was prepared by mechanical mixing of Pt/Ba/Mn/Al2O3 and Rh/CeO2 powders to investigate the impacts of the Rh/Pt weight ratio on the overall NOx removal efficiency and nitrogen yield.The Mn/Al2O3 powder with the target MnO2 loading of 10 wt% was prepared by the incipient wetness impregnation of γ-Al2O3 with an aqueous solution of Mn(CH3COO)2, followed by drying at 120 °C overnight and calcination at 550 °C in static air for 3 h. The Ba/Mn/Al2O3 powder was prepared by the same procedure as the Mn/Al2O3 powder, using Ba(CH3COO)2 aqueous solution and the prepared Mn/Al2O3. The target BaO loading was 15 wt% in the Ba/Mn/Al2O3 support. The prepared Ba/Mn/Al2O3 powder was further calcined at 850 °C in static air for 4 h to yield the final Ba/Mn/Al2O3 support.The Pt/Ba/Mn/Al2O3-x catalysts with different Pt loadings were prepared by the incipient wetness impregnation of the Ba/Mn/Al2O3 support with an aqueous solution of hydroxylamine Platinum(II), followed by drying at 120 °C overnight and calcination at 590 °C in static air for 2 h. The target Pt loadings of the Pt/Ba/Mn/Al2O3–1, Pt/Ba/Mn/Al2O3–2, Pt/Ba/Mn/Al2O3–3, Pt/Ba/Mn/Al2O3–4 and Pt/Ba/Mn/Al2O3–5 were 1.07, 1.20, 1.27, 1.33 and 1.20 wt%, respectively. The Pt/Al2O3 and Pt/Ba/Al2O3 catalysts with a target Pt loading of 1.2 wt% were also prepared by the same method to investigate the impacts of Mn and Ba on Pt dispersion. The Rh/CeO2-y catalysts were prepared by the same procedure as the Pt/Ba/Mn/Al2O3-x catalysts, using Rhodium(III) nitrate aqueous solution and CeO2 powder. The target Rh loadings of the Rh/CeO2–1, Rh/CeO2–2, and Rh/CeO2–3 were 2.34, 1.19 and 0.60 wt%, respectively.To achieve the model NSR catalyst with a target metal loading of 1.2 wt%, 5 g of the Rh/CeO2-y catalyst was mechanically mixed with 45 g of the Pt/Ba/Mn/Al2O3-x (y = x) catalyst to obtain Rh0.2Ce-Pt0.8BMA, Rh0.1Ce-Pt0.9BMA and Rh0.05Ce-Pt0.95BMA samples with Rh/Pt ratios of 0.25, 0.11 and 0.05, respectively. Similarly, 5 g of pure CeO2 powder was mechanically mixed with 45 g of the Pt/Ba/Mn/Al2O3–4 catalyst to obtain CePt1.00BMA sample. The Pt/Ba/Mn/Al2O3–5 catalyst was marked as Pt1.00BMA.Powder X-ray diffraction (XRD) patterns were recorded on a Philips X'pert Pro diffractometer in the 2θ range of 5–90° with an increment step of 0.02°, using a Ni filtered Cu Kα radiation (λ = 0.15418 nm) source. The X-ray tube was operated at 40 kV and 30 mA.The actual Pt loading was determined by inductively coupled plasma-optical emission spectroscopy (ICP-OES), and the content of metal oxides was measured by an X-ray fluorescence (XRF) spectrometry. Pt dispersion of the Pt/Al2O3, Pt/Ba/Al2O3 and Pt1.00BMA catalysts was measured by the CO-pulse method following the procedure reported in our previous work [10]. The specific surface area and total pore volume of the catalysts were determined by nitrogen adsorption at −196 °C using a Quantachrome NOVA2000e analyzer. The specific surface area was estimated using the Brunauer-Emmett-Teller (BET) equation and the total pore volume was estimated from the single nitrogen adsorption amount at the P/P 0 of ~0.98. High-angle annular dark field (HAADF) scanning transmission electron microscopy (STEM) images and EDS maps were collected on a Tecnai G2 TF30 microscope operating at 300 kV.Temperature-programmed reduction (H2-TPR) experiments were carried out on a Quantachrome CHEMBET 3000 automated chemisorption instrument equipped with a thermal conductivity detector (TCD) to monitor H2 uptake. After pretreatment at 350 °C for 30 min in Ar gas stream, H2-TPR experiments were performed in 10 vol% H2/Ar gas mixture with a total flow rate of 80 mL/min from 50 to 360 °C at a heating rate of 10 °C/min.The catalytic activity of the various NSR solids (Section 2.1) was evaluated using a continuous flow fixed-bed microreactor. The catalyst was pre-treated at 500 °C in 5% O2/N2 gas stream for 40 min. Then, the catalyst was cooled or heated to the reaction temperature of interest. The activity experiment was performed in lean/rich (45 s/15 s) cycling gas streams with a gas hourly space velocity (GHSV) of 120,000 mL h−1 g−1. Twenty cycles of lean-to-rich switching gas stream were performed at every temperature, and the last five cycles were used for estimation of catalyst activity. The inlet composition of the lean gas stream was 400 ppm NO, 5% O2, 5% H2O, 5% CO2 and ~ 85% N2. The inlet composition of the rich gas stream was 3500 ppm CO, 1000 ppm C3H6, 5% H2O, 5% CO2 and ~ 89.5% N2. The outlet gas concentrations were analyzed using an online MultiGas FT-IR Analyzer (2030DBG2EZKS13T). The NOx removal efficiency, X NOx (%), was estimated by the following Eq. (1) [11]: (1) X NO x % = 100 × ∫ 0 300 NO x in − NO x out d t ∫ 0 300 NO x in d t The various product yields were calculated according to the following Eqs. (2)–(5): (2) Y NH 3 % = ∫ 0 300 NH 3 out d t ∫ 0 300 NO in d t (3) Y N 2 O % = 2 ∫ 0 300 N 2 O out d t ∫ 0 300 NO in d t (4) Y NO 2 % = ∫ 0 300 NO 2 out d t ∫ 0 300 NO in d t (5) Y N 2 % = ∫ 0 300 NO in − NO out − NO 2 out − 2 N 2 O out − NH 3 out d t ∫ 0 300 NO in d t where [NOx] in , [NH3] out , [N2O] out and [NO2] out present the transient concentrations of the inlet NOx and outlet NH3, N2O and NO2, respectively, during the lean-to-rich cycling. [NOx] out is the outlet NOx concentration. The N2 yield is estimated indirectly through a material balance based on the other measured nitrogen-containing compounds, as presented in Eq. (5). The lean NOx storage efficiency was calculated according to the following Eq. (6) [12]: (6) ŋ NO x % = ∫ 0 45 NO in − NO out − NO 2 out d t ∫ 0 45 NO in d t In Eqs. (1)–(5) and (6), the time duration of each cycle was 300 s and 45 s, respectively.Before the activity test, blank calibration under lean/rich (45 s/15 s) cycling was conducted at every reaction temperature in order to achieve the transient concentrations of the inlet NO for the estimation of ∫0 300[NO] in dt .The specific surface area, total pore volume, and chemical composition of each catalyst are summarized in Table 1 . The values of specific surface area of Pt1.00MBA, CePt1.00MBA, Rh0.05Ce-Pt0.95MBA, Rh0.1Ce-Pt0.9MBA and Rh0.2Ce-Pt0.8MBA samples are 93.5, 97.5, 109.9, 110.7 and 110.3 m2/g, respectively. The corresponding values of total pore volume are 0.27, 0.39, 0.32, 0.32 and 0.28 cm3·g−1, respectively. The actual.Rh/Pt weight ratios of Rh0.05Ce-Pt0.95MBA, Rh0.1Ce-Pt0.9MBA and Rh0.2Ce-Pt0.8MBA samples are 0.06, 0.12 and 0.22, respectively, according to the results of actual Pt and Rh loadings listed in Table 1. The actual contents of BaO, MnO2 and Al2O3 in the Ce-containing samples are lower compared to that of Pt1.00MBA sample, which can be attributed to the dilution induced by CeO2 or Rh/CeO2 addition.Powder X-ray diffractograms of the catalysts are shown in Fig. 1 . All XRD patterns exhibit the characteristic reflections of alumina components at 2θ ≈ 37.2, 45.9 and 66.7° [13]. The characteristic peaks of Rh species were not detected in the Rh-containing samples because the concentration of Rh was lower than the detection limit of powder XRD technique [14]. The clear Pt reflections were also not observed in the X-ray diffractograms, although the calcination temperature was higher than that of the Pt/Al2O3 catalyst reported in our previous work [15]. The estimated Pt dispersion of the Pt/Al2O3, Pt/Ba/Al2O3 and Pt1.00BMA catalysts was 12.4, 25.3 and 35.4%, respectively. These results indicate that addition of Mn and Ba is advantageous for improving Pt dispersion.The XRD peaks centered at 2θ ≈ 23.9, 24.3, 34.6, 42.0, 44.9 and 47.0° can be assigned to the characteristic reflections of the orthorhombic BaCO3 phase, thereby confirming the decomposition of Ba(O2CCH3)2 into crystalline BaCO3 during catalyst calcination [16]. The estimated primary crystallite sizes of BaCO3 particles as summarized in Table 1, suggesting the commensurate dispersion of the crystalline BaCO3 phase in all catalysts. The characteristic XRD peaks appeared at 2θ ≈ 25.8, 31.4 and 41.2° for the Pt1.00MBA sample illustrates the formation of BaMnO3 phase due to the reaction between MnO2 and BaO during calcination [13]. By contrast, the Ce-containing samples show slightly weaker intensities of the peaks assigned to BaMnO3, BaCO3 and Al2O3 components, which can be ascribed to the reduced relative concentrations. From XRD patterns of the Ce-containing samples, additional diffraction peaks at 2θ ≈ 28.6, 33.1, 47.5, 56.3, 76.7, 79.2 and 88.6° can be clearly observed, which are closely related to the characteristic reflections of CeO2 crystallites [10]. The estimated primary crystallite sizes of CeO2 particles in the Ce-containing samples are summarized in Table 1, showing that by increasing the Rh loading conduces to the growth suppression of CeO2 crystallite size.The chemical elements distribution was clearly confirmed by HAADF-STEM images, and EDS maps are shown in Fig. 2 . In the Pt1.00MBA catalyst, the entire overlapping of the EDS map of Ba element with that of Mn i indicates, (indirectly) the formation of BaMnO3. Mn-rich phase was also observed, which illustrates the presence of MnO x species, although it is not detected by the powder XRD technique. From the EDS map of Pt element in Pt1.00MBA, it was confirmed that Pt location on Mn and Ba species occurs to a greater extent than Pt location on Al2O3. Rh distribution on the surface of CeO2 is observed from the EDS maps of Rh and Ce elements in the Rh0.2Ce-Pt0.8MBA catalyst. Fig. 3 shows H2-TPR profiles of the catalysts investigated. The H2-TPR profile of Pt1.00MBA presents a slight negative peak at 175 °C and three H2 consumption peaks at 292, 315 and 333 °C. Observations of the negative peak can be ascribed to the temperature-driven desorption of the very weakly chemisorbed hydrogen. The H2 consumption peak at 292 °C corresponds to the reduction of MnO2 to Mn2O3, whereas the peaks at 315 and 333 °C should be respectively attributed to the reduction of Mn4+ to Mn2+ in the BaMnO3 and the stepwise reduction of MnO x species [5]. Addition of CeO2 into the Pt1.00MBA sample leads to additional H2 consumption at 147 and 208 °C, corresponding to surface oxygen species in the CeO2 crystallites [10,15]. Compared with the CePt1.00MBA catalyst, the additional H2 consumption peaks appeared at 130, 116 and 104 °C in the H2-TPR profiles of the Rh0.05Ce-Pt0.95MBA, Rh0.1Ce-Pt0.9MBA and Rh0.2Ce-Pt0.8MBA samples, respectively, can be assigned to the reduction of active surface oxygen species in the vicinity of Rh sites [17].As shown in Fig. 4 , all the catalysts show a low NOx removal efficiency at 200 °C, which is much lower than the NOx storage efficiency, because the stored NOx is released and reduced under the rich phase [18,19]. By increasing the temperature to 350 °C leads to a significantly enhanced NOx storage and removal, which agrees well with the literature [9]. The Rh-containing samples showed the smaller gaps between NOx storage and overall removal efficiency compared to the Rh-free samples, below 350 °C, implying that the introduction of Rh sites promotes the reduction of the stored NOx. The decrease of NOx storage and overall removal efficiency above 400 °C is a result of the shift to an equilibrium-limited regime [20]. The results displayed in Fig. 4 indicate that by increasing the Rh/Pt weight ratio enhances NOx reduction under the rich phase in the kinetic regime but shows a negative effect in the equilibrium-limited regime.N2 yield directly reflects the ability to eliminate the hazardous N-containing gaseous pollutants. In Fig. 5 , no observation of the clear advantage of Rh-containing catalysts in the N2 yield below 250 °C is seen, and this is the result of the larger NH3 yield than in the Rh-free catalyst, due to the superior low-temperature activity of the water-gas shift reaction over the Rh/CeO2 catalyst [9,21]. In the region of 250–400 °C, by increasing the Rh/Pt weight ratio a clear advantage in N2 yield can be ibserved, corresponding to less N2O and NH3 production. Ammonia adsorbed species start to decompose into N2 and H2 at ~250 °C over the Rh-containing catalyst, which results in more N2 production [8]. N2 yield is improved with increasing Rh/Pt weight ratio due to the higher reactivity of Rh than Pt in the NH3 decomposition reaction [8]. With further increasing the reaction temperature, N2 yield declined although only tiny amounts of by-products were formed. Such observation is mainly ascribed to the thermodynamic limitation of NO conversion at 400 °C [22].Mn species existed in the form of MnO x and BaMnO3 phases in the model NSR catalysts investigated, whereas the identified Ba species included BaCO3 and BaMnO3. By increasing the Rh/Pt weight ratio enhanced NOx reduction under the rich phase and improved overall N2 yield in the kinetic regime of NO conversion but shows a negative effect in the equilibrium-limited regime above 400 °C. An NSR catalyst should be operated below 500 °C, and the Rh/Pt weight ratio should be selected based on the actual operating temperature that is dependent on the installation position. However, the NSR catalysts must keep the Rh/Pt weight ratio as low as possible, unless the overall operating temperature is below 350 °C, because rhodium is much more expensive than platinum.We declare that we have no financial and personal relationships with other people or organizations that can inappropriately influence our work, there is no professional or other personal interest of any nature or kind in any product, service and/or company that could be construed as influencing the position presented in the manuscript entitled “Rh/CeO2+Pt/Ba/Mn/Al2O3 model NSR catalysts: Effect of Rh/Pt weight ratio”.This research was supported by the National Natural Science Foundation of China (21862010), the Provincial Applied Fundamental Research Program of Yunnan (202101AT070237), the Major Science and Technology Programs of Yunnan (2019ZE001-2, 202002AB080001-1), and the National Engineering Laboratory for Mobile Source Emission Control Technology (NELMS2019C01).

A series of model NOx storage-reduction (NSR) catalysts with different Rh/Pt weight ratios were prepared independently by wet impregnation and then mechanically mixed to investigate the effect of the Rh/Pt weight ratio on the overall NOx removal efficiency and nitrogen yield. XRD, EDS and H2-TPR studies indicated the coexistence of BaCO3, MnO x and BaMnO3 phases in the catalysts. Increasing Rh/Pt weight ratio enhanced NOx reduction under the rich phase, and improved overall N2-yield in the kinetic regime of NO conversion, but shows a negative effect in the equilibrium-limited regime above 400 °C.

In the chemical industry, the catalytic dehydrogenation reactions are commonly used, and in the preparation of pharmaceuticals and fine chemicals, such reactions involving the oxygenated compounds are particularly important (Singh and Vannice, 2001). The main procedure for cyclohexanone production is the catalytic dehydrogenation of the cyclohexanol, which in the synthesis of pharmaceuticals and fine chemicals is an important intermediate. For instance, producing of the caprolactam and adipic acid, the main raw materials in manufacture of the polyamide fiber in nylon textiles such as nylon-6 and nylon-6.6, respectively. Most of the uses of the caprolactam depend on the type and amount of impurities which it contains. In this case, higher purity requirements must be increasingly met by the raw materials (Simón et al., 2012a,b).With catalysts based on the copper, cyclohexanol dehydrogenation processes are carried out as when they are carefully reduced, they present a highly dispersed copper phase. In this manner, they usually operate under mild conditions. Due to the copper sintering, these copper catalysts are not used at high temperature (Simón et al., 2012a,b). Recently, the field of alloy catalysts owing to their enhanced catalytic performance has been attracting a lot of interest compared to the individual components.Because of the unique biological, electronic, optical, magnetic, and specifically catalytic properties, the metallic nanoparticles (NPs) constructed from more than one metal have aroused great interest (Singh and Xu, 2013). In this case, owing to the interplay between electronic and lattice effects of the neighboring metals, multimetallic NP-based catalysts often show superior catalytic activities to their monometallic counterparts (Ataee-Esfahani et al., 2010). However, so far the controlled synthesis of the NPs consisting of multiple (n ≥ 3) metal components has remained relatively unexplored and scientists have focused mainly on bimetallic systems (Wang and Li, 2011).In order to enhance their properties, metals such as Rh (Mendes and Schmal, 1997), Co, Zn, Fe, Cr, Pd, Ni (Nagaraja et al., 2011) added to the copper catalysts. For catalytic activity of tri metallic catalysts, the shape and size distribution, surface segregation and crystalline structure, bulk and surface compositions are crucial factors (Ranga Rao et al., 2012).On the other hand, as a low cost support for heterogeneous catalysts one can use a suitable support and type of interaction in industry with support material graphene to replace current metal oxide based catalyst supports, in order to maximize the catalytic activity of a catalyst (He et al., 2013). In this case, due to their high external surfaces, it is expected to be efficient, excellent high electrical conductivity and thermal/chemical stability which are leading to increase the selectivity and rate of the reactions, preventing poisoning and inhibition of sintering of the active metal surface by coke deposition (Julkapli and Bagheri, 2014).This research focused on the formulation of the catalyst supported by nitrogen doped graphene (N-rGO) formulation of two different copper based supported catalysts (Cu, and tri metallic alloy CuNiRu), characterization by BET, XPS, TPR-H2, TPD-NH3 and XRD. For dehydrogenation of the cyclohexanol, testing the catalytic activity had been carried out for the two different catalysts to evaluate the role of the promoters (Ni and Ru), the effect of these two different catalysts on the activity and selectivity at different operating conditions (T = 200, 225, 250, 260 and 270 °C, P = 1 atm, liquid flow rate of reactant = (0.1 ml/min), gas flow rate of the carrier(N2 gas) = 25 ml/min, and time of the reaction ∼8h.For unpromoted and promoted catalysts designated as CuNiRu/N-rGO and Cu/N-rGO, the dehydrogenation of cyclohexanol carried out. By Modified Staudenmaier’s method, the Graphite oxide is prepared as reported by (Ambrosi et al., 2012); however, N-rGO was prepared as described in (Ning et al., 2013). Using incipient wetness impregnation of the Cu-precursor into the N-rGO support, the Cu/N-rGO catalyst was synthesised. The mass of the impregnated precursor was estimated to be equivalent to 1 wt% of the Cu. The N-rGO support was dried overnight at 110 °C, and the wet slurry containing the Cu-precursor (Cu(NO3)2·3H2O, R&M chemicals) calcined and subsequently reduced for 3 h to obtain as-synthesised 1 wt%Cu/99 wt%N-rGO catalyst in a flow of N2/H2 (10% (v/v)) at 275 °C. In order to synthesise CuNiRu/N-rGO catalyst, the same preparation method of Cu/N-rGO catalyst was followed. The Cu-Ni-Ru-precursors (Ni(NO3)2·6H2O, M. = 290.79 g/mole), (Cu(NO3)2·3H2O, M. = 241.60 g/mole), and (Ru(NO)(NO3)3, M. = 317.09 g/mole) (R&M chemicals). The mass of the impregnated precursor was estimated to be equivalent to 0.5 wt%Cu, 0.25 wt%Ni and 0.25 wt%Ru.Temperature programmed desorption using NH3 (TPD-NH3) was utilized to measure the acidity with temperature programmed reduction (TPR) studies of the catalysts were performed on a Thermos Finnigan TPDRO1100 series with 5% H2-Ar as reducing and carrier gas. For specific surface area, by adsorption-desorption isotherm using Brunauer-Emmett-Teller (BET method), all the catalysts were characterized using a Micromeritics Pulse Chemisorb 2700 instrument. Before measurements, the samples were oven dried at 393 K for 12 h and flushed in-situ with He gas for 2 h. As X-ray source operated at 25.6 W (beam diameter of 100 µm), the surface analysis of selected catalyst is carried out using the XPS (Ulvac-PHI, ULVAC-PHI Quantera II, INC.), with monochromatic Al-Kα (hv = 1486.6 eV). In this manner, the wide scan analysis was performed by using a pass energy of 280 eV with 1 eV per step. While, the narrow scan was performed using a pass energy of 112 eV with 0.1 eV per step. On a RIGAKU miniflex II X-ray diffractometer capable of measuring powdered diffraction pattern from 3 to 145° in 2 θ scanning range, the XRD patterns of the catalysts in reduced forms recorded. The X-ray source is Cu Kα with wavelength (λ) of 0.154 nm radiation. In this case, the XRD has been set up with the latest version of PDXL, RIGAKU full function powder-diffraction analysis software.In a stainless steel fixed bed reactor Cyclohexanol dehydrogenation was carried out using a nitrogen gas cylinder which serves as the carrier gas and a liquid micro pump for feeding of the cyclohexanol into the reactor. Fig. 1 shows the schematic of the experimental set up. In this case, by external electric heater, the reactor was heated and insulated with glass wool. Using a K-type thermocouple, the temperature of the catalytic bed was monitored. Reaction was normally conducted under the following standard conditions: 200, 225, 250, 260 and 270 °C temperatures, 0.1 g catalyst weight, atmospheric pressure, 0.1 ml/min pure cyclohexanol feed flowrate, ∼8 h reaction time.Here, the reaction sequence was as follows: with the appropriate amount of catalyst, the reactor has been loaded. In order to run for 8 h under the above mentioned conditions, the reaction was allowed. In this case, by GC (Shimadzu 17A, FID, CP-Sil 24 CB 30 m 0.25 mm 0.25 um), the Liquid samples were analyzed. The detected liquid products were as phenol and cyclohexanone. In case, the Gas analysis was performed in (Shimadzu 17A, TCD, HP-MoLesieve, 19095P-M50 30 m * 0.530 mm * 50 Mm) and the gaseous product was mainly hydrogen.1: Liquid feed tank, 2: Liquid pump, 3: Valves, 4: Preheater, 5: Fixed bed reactor, 6: Thermocouple, 7: External electric heater, 8. Catalyst bed, 9: Rotameters, 10: Cooler, Gas-liquid separator, 11: Liquid sample and 12: Gas sample (H2 gas).The heterogeneous catalysts are generally porous in nature. This characteristic plays an important role, in their catalysis application. The catalytic activity is closely linked to the available surface area for adsorption. The BET specific surface area, pore size and pore volume distributions of the CuNiRu/N-rGO and Cu/N-rGO samples analysed by the N2 adsorption/desorption isotherms. For all of the classified samples, the N2 adsorption-desorption isotherm (Balbuenat and Gubbins, 1993), in which two branches are almost parallel over a wide range of the P/Po due to the IUPAC classification. For Cu/N-rGO and CuNiRu/N-rGO samples, the N2 adsorption-desorption isotherm displaying the type-III isotherm (see Fig. 2 (a and b)). Indeed, the samples have macroporous (pore size > 50 nm) structures. The Cu/N-rGO and CuNiRu/N-rGO samples have a hysteresis of the type H3 according to the IUPAC classification. Attributed to the formation of the slit-shaped pores, these samples have plate-like structures. Compare to that of the Cu/N-rGO catalyst, the CuNiRu/N-rGO catalyst recorded a decreasing about 90% in pores diameters and an increasing ∼75% in the BET surface area (Table 1 ). Benefiting from the unique structure uniform distribution of the optimal size, the results imply that the Ni and Ru promoters provide a large surface area of the catalyst. In this way, with adding the different promoters such as Zn, Zr and Al, the improvement of the copper catalyst supported reduced graphene oxide (Fan and Wu, 2016). In this research, they found that by adding the promoters to the copper catalyst and the pore diameters deceased, the surface area of the catalyst would be enhanced.The NH3-TPD analysis is used to measure the samples acidity. The quantitative estimation of the acidic site of the samples is summarized due to the desorbed amount of the ammonia (Table 2 ). Based on the peak temperature, in three different regions as 200–400 °C, 400–600 °C and above 600 °C, the results illustrate that the acidic sites are distributed. In this manner, the bimetallic Cu-Ni catalysts supported on the γ-Al2O3 was prepared by Pudi et al., (2015) and they stated that to the ammonia desorption from the weak acidic sites, the first region is attributed while the second region refers to the medium strength acidic sites. In case, from the strong acidic sites, the third region represents the ammonia desorption.The desorption peaks at a range of medium acid sites illustrated by the CuNiRu/N-rGO and Cu/N-rGO catalysts. In this manner, the temperature at which the peaks desorption occurs related to the acidic strength. Table 2 shows that both of the CuNiRu/N-rGO and Cu/N-rGO and catalysts correspond to the medium acidic sites on their surfaces respectively with the NH3 total amount desorbed of 4.64 × 103 and 2.57 × 104 µmol/g. In this way, suggested by the NH3-TPD results, the addition of the promoters can change the total acidity and acidic strength of the catalyst surface. Ji et al. (2007) has been found to provide the suitable monovalent copper active sites based on the study of the dehydrogenation of the cyclohexanol over the Cu-ZnO/SiO2 catalysts, the ZnO help as a promoter. The total acidity of the trimetallic catalyst CuNiRu/N-rGO by 2.11 × 104 µmol/g was higher than the Cu/N-rGO. However, the acidic strength is weaker than the Cu/N-rGO due to the rule of the promoter’s Ni and Ru as encouraging increasing the activity of the catalyst and providing the suitable active sites as well.The XPS is used more widely than the others to analyse the surface composition and oxidation states of the industrial catalysts. The photoelectron lines of the wide and narrow scan spectrum evidently present the Cu according to the XPS results as illustrated in Fig. 3 . The metals loading in terms of the weight percent of the Cu to Ni and Ru and the loading of the CuNiRu alloy nanoparticles on the N-rGO support are close to the ratio of 50%:25%:25%. In case, it was found that due to the low content, there is no absorption peak of the Ni and Ru. Here, one can find the atomic percent (at%) of each of the elements for all the tested samples (see Table 3 ). From the atomic percentage results for the catalysts, the presence of the C, Cu, N and O in the sheets of the Cu/N-rGO and CuNiRu/N-rGO are proved. Fig. 3 shows the high resolution spectrum of the Cu element. The peak at 933.2 eV is assigned to the Cu2p3/2 as per previous studies, which is attributed to either of the Cu and/or Cu2O (Jia et al., 2015; Durando et al., 2008). Indeed, it is difficult to distinguish between these two species based on the Cu2p3/2 binding energies, as they are much close. Attributing to the binding energies of the Cu and/or Cu2O, the CuNiRu/N-rGO catalyst illustrate the peak at 933.1 eV as well as the Cu/N-rGO. However, the peak shifted from the observed Cu2p3/2 of the CuNiRu/N-rGO catalyst. In this case, by the shift of the Cu2p3/2 binding energy, a strong electronic interaction between the Cu and Ni with the Ru elements in the metallic alloy are indicated.The H2-TPR usually is used to determine the reduction behavior of the catalysts. One can categorize the reduction behaviors of the CuNiRu/N-rGO and Cu/N-rGO catalysts into two stages of the reduction behavior (Fig. 4 (a and b)). Due to the reduction behavior of the monovalent copper active sites (Cu2O) to the metallic copper active sites (Cuo), based on the results reported by Bridier et al. (2010), the first stage of the H2-TPR profile recorded at lower temperature. In the case of the Cu/N-rGO and CuNiRu/N-rGO catalysts, as described in Schlapbach et al. (2001), the second peak possibly refer to the adsorbed H2 on the C surface of the N-rGO. The CuNiRu/N-rGO catalyst comparing the Cu/N-rGO catalyst indicates decreasing about 18 °C in the reduction temperature. Due to the strong interaction between the Cu2O species and the promoters’ Ni and Ru causes highly dispersion of the copper active sites on the N-rGO support, the reduction temperature of CuNiRu/N-rGO catalyst could be shifted. In this case, the higher dispersion of the Cu2O species responsible for their ease of reduction indicated by the shift clearly. In this way, in the study of the promoted copper catalysts, the results are in good agreement with the obtained results by Lin et al. (2010).The crystallinity and phase determination of the samples were analyzed using the X-ray diffraction. Fig. 5 (a and b) shows the XRD diffraction patterns of the Cu/N-rGO and CuNiRu/N-rGO where the XRD pattern of the Cu/N-rGO catalyst illustrates there are five diffraction peaks centered at 2θ = 26.6°, 37.7, 43.28°, 50.68° and 61.7° (Fig. 5(a)). In this case, the 2θ = 37.7°, 43.28° and 61.7° might be indexed to the (111), (200) and (220) facets of the cubic phased Cu2O (ICDD card number 00–001-1241) due to the surface oxidation of the Cu NPs, the first short and weak peak at 2θ = 26.6° can be attributed to the (002) planes of the N-rGO. However, the strongest peak can be attributed in the Cu/N-rGO pattern to the (111) facet of the Cu2O. These finding are in good agreement with the results reported by Zhang et al. (2016) within providing the Cu2O/rGO in one step of the reduced GO and loading the Cu2O. The presence of the other phases of the copper interestingly indexed to the (200) confirming of the formation of the metal Cu and associated to the existence of the Cu crystallites detected at 2θ = 50.6°. In this case, the crystallite size of the Cu2O at 2θ = 37.7° of the (111) planes from Scherrer equation was estimated as 3.8 nm. In case, the presence of the Cu2O and the metal Cu phases on the N-rGO support revealed by the XRD results. Fig. 5(b) shows the XRD patterns of the CuNiRu/N-rGO catalyst. In this case, due to the presence of the amorphous carbon in N-rGO, the weak and broad peak at 2θ = 26.7° relates to the (002) planes. Attributed to the copper oxide (Cu2O) for the (111), (200) and (220) planes, the diffraction peaks centered at 2θ = 37.80°, 43.71° and 61.9° respectively. By the existence of the broad and weak diffraction peak at 2θ = 50.6°, the presence of the metal Cu could be confirmed which can be indexed to the (200) planes. The peak at 2θ = 37.80° attributed to Cu2O, NiO and RuO2 respectively for (111), (111) and (101) planes. Compare to the Cu/N-rGO, this peak illustrates a slightly shift in angles to the higher value. The shift in the diffraction angle is due to the formation of the CuNiRu alloy as indicated by Bai et al. (2015) in their pioneering research on preparation of the hollow PdCu alloy supported on the N-rGO. Moreover, the crystal size of the CuNiRu/N-rGO catalyst estimated to be 1.13 nm which is smaller than those of the Cu/N-rGO about 2.67 nm. As an evidence for the existence of the CuNiRu/N-rGO catalyst in a strong interaction alloy one can consider the mentioned results.The obtained results from the XRD diffraction support the H2-TPR. As it is observed, the reduction behavior of the catalysts samples encourages one step reduction process due to the reduction of the Cu2O (monovalent Cu+) to the Cuo (metallic Cu). For the CuNiRu/N-rGO and Cu/N-rGO catalysts, from the XRD diffraction results the CuNiRu/N-rGO recorded the highly presence of the Cu2O active sites and the smallest crystal sizes as well. Figs. 6–8 illustrates the effect of the reaction temperature on conversion of the cyclohexanol over the CuNiRu/N-rGO and Cu/N-rGO catalysts and the cyclohexanone yields and selectivity. Here, the process temperature varies between 200 and 270 °C. In general, it is obvious that the cyclohexanol conversions the cyclohexanone yields (Figs. 6 and 7) and increases when the reaction temperature increases, while the selectivity of the cyclohexanone decreases at the same operating conditions (see Fig. 8).This is consistent with the activation energy trends for temperature dependent reactions, whereby the catalytic performance increases with respect to the temperature increasing in terms of the conversion of the reactants and the yield of the products (Smith, 2008). Compare to the catalyst Cu/N-rGO, the catalyst CuNiRu/N-rGO has an increase of 3.7% of the cyclohexanol conversion, 8.1% of the cyclohexanone yield, 8.8% of the cyclohexanone selectivity, at 270 °C and 60 min (Figs. 6–8).As shown in Fig. 8, with rise of the temperature (200–270 °C), the cyclohexanone selectivity on both catalysts decreases. This can be due to the formation of the other by products e.g. phenol and cyclohexene via the side reactions occurring at the elevated temperatures such as the aromatization of the cyclohexanol to phenol and dehydration of the cyclohexanol to cyclohexene (Ji et al. (2007).Not only for the reaction of the dehydrogenation of alcohol to the ketone the sites of the metallic copper are active sites but also for the reaction of aromatization of the cyclohexanol to the phenol. Therefore, the metallic copper active sites are not selective. As indicated from the NH3-TPD results, for this particular reaction, the CuNiRu/N-rGO is more suitable as within the actual dehydrogenation of the cyclohexanol to the cyclohexanone prefers the lower acidity active sites (Chang and Abu Saleque, 1993).The stability of the Cu/N-rGO and CuNiRu/N-rGO catalysts was evaluated based on the times on stream (TOS) ∼8 h (530 min). In this manner, the three different reaction stages can be identified as per Figs. 6–8, for the Cu/N-rGO as follows: • For TOS = 0–60 min and reaction temperature 200–270 °C, the activation stage (Stage 1) shows increasing about 23.3% of the conversion of the cyclohexanol, 7% of selectivity of the cyclohexanone and 14% of yield of the cyclohexanone. The results’ increasing is due to the activity of the fresh Cu/N-rGO catalyst. • For TOS = 60–110 min Steady state stage (Stage 2) shows that the results are almost constant. • The deactivation stage (Stage 3) shows a decline in the results for TOS = 110–530 min and reaction temperature 200–270 °C. The reduction in the results in terms of the reductive values between temperatures 200–270 °C as: the cyclohexanol conversion was about 10%, the selectivity of the cyclohexanone was 7%, yield of the cyclohexanone was 14%. The results suggest that the deactivation affects the reactions. For TOS = 0–60 min and reaction temperature 200–270 °C, the activation stage (Stage 1) shows increasing about 23.3% of the conversion of the cyclohexanol, 7% of selectivity of the cyclohexanone and 14% of yield of the cyclohexanone. The results’ increasing is due to the activity of the fresh Cu/N-rGO catalyst.For TOS = 60–110 min Steady state stage (Stage 2) shows that the results are almost constant.The deactivation stage (Stage 3) shows a decline in the results for TOS = 110–530 min and reaction temperature 200–270 °C. The reduction in the results in terms of the reductive values between temperatures 200–270 °C as: the cyclohexanol conversion was about 10%, the selectivity of the cyclohexanone was 7%, yield of the cyclohexanone was 14%. The results suggest that the deactivation affects the reactions.In this manner, one can identify the three different reaction stages for the CuNiRu/N-rGO as (see Figs. 6–8): • The activation stage (Stage 1) shows increasing in conversion of the cyclohexanol for TOS = 0–60 min and reaction temperature 200–270 °C and selectivity and yield of the cyclohexanone. In this case, by the activity of the fresh CuNiRu/N-rGO catalyst the increasing in the results is caused. • The steady state stage (Stage 2) shows that the results are almost constant for TOS = 60–380 min. • The deactivation stage (Stage 3) shows a decline in the results for TOS = 380–530 min and reaction temperature 200–270 °C. In terms of reductive values of the temperatures 200–270 °C, in the results the reduction in the cyclohexanol conversion, yield and selectivity of cyclohexanone. Thus, by the deactivation based on the obtained results, the reactions slightly affected. The activation stage (Stage 1) shows increasing in conversion of the cyclohexanol for TOS = 0–60 min and reaction temperature 200–270 °C and selectivity and yield of the cyclohexanone. In this case, by the activity of the fresh CuNiRu/N-rGO catalyst the increasing in the results is caused.The steady state stage (Stage 2) shows that the results are almost constant for TOS = 60–380 min.The deactivation stage (Stage 3) shows a decline in the results for TOS = 380–530 min and reaction temperature 200–270 °C. In terms of reductive values of the temperatures 200–270 °C, in the results the reduction in the cyclohexanol conversion, yield and selectivity of cyclohexanone. Thus, by the deactivation based on the obtained results, the reactions slightly affected.The Cu/N-rGO catalyst observed the decreasing of 67.1% of the cyclohexanol conversion, 46.4% of the cyclohexanone yield and 51.4% of the cyclohexanone selectivity. All the results support the idea of the fast deactivation of the Cu/N-rGO vigorously occurred after 110 min from the reaction started.While, the CuNiRu/N-rGO catalyst has illustrated better stability at high TOS up to 380 min, and then a slightly reduced in the results was observed up to the end of the reaction. Compared to that of the Cu/N-rGO in terms of increasing the reaction conversion the excellent CuNiRu/N-rGO performance and its stability at different TOS of the reactions could be attributed to the promotional effect of Ru and Ni in the formation of CuNiRu/N-rGO catalyst.To investigate the catalytic performance of the supported catalysts in the dehydrogenation of cyclohexanol to cyclohexanone and to analyse the properties of the synthesised catalysts using TPD-NH3, BET, XPS, TPR-H2, TGA and XRD techniques, this research was revolved to formulate two types of the supported catalysts namely supported copper (Cu/N-rGO) and supported tri metals alloy (CuNiRu/N-rGO) in paper forms. The briefly conclusions from the major findings of this study are: • The promoters (Ni and Ru) were added to the Cu/N-rGO catalyst and raised supervisor behaviors in terms of the provided suitable and selective active sites for catalytic dehydrogenation of the cyclohexanol to the cyclohexanone reaction with the smallest crystals size, higher thermal stability and larger surface area. • The reaction results in terms of the highest cyclohexanol conversion, cyclohexanone yield and selectivity and hydrogen productivity in case of CuNiRu/N-rGO catalyst detected much improvement with in a reduction in phenol yield and selectivity as well. • Due to the copper sintering, coke deposition and reduced active sites, the fastly deactivated in the Cu/N-rGO which might be. Thus, the CuNiRu/N-rGO has a much longer operational life. This is based on the performance evaluation of the catalysts. The promoters (Ni and Ru) were added to the Cu/N-rGO catalyst and raised supervisor behaviors in terms of the provided suitable and selective active sites for catalytic dehydrogenation of the cyclohexanol to the cyclohexanone reaction with the smallest crystals size, higher thermal stability and larger surface area.The reaction results in terms of the highest cyclohexanol conversion, cyclohexanone yield and selectivity and hydrogen productivity in case of CuNiRu/N-rGO catalyst detected much improvement with in a reduction in phenol yield and selectivity as well.Due to the copper sintering, coke deposition and reduced active sites, the fastly deactivated in the Cu/N-rGO which might be. Thus, the CuNiRu/N-rGO has a much longer operational life. This is based on the performance evaluation of the catalysts.This work was fully sponsored by the Ministry of Higher Education, Malaysia under the Fundamental Research Grant Scheme - (03-02-1522FR). All analyses, otherwise specified, were conducted at the Material Characterization Laboratory, Universiti Putra Malaysia.

In different hydrocarbons reactions, copper based catalysts have industrial importance especially in the synthesis of the cyclohexanone from dehydrogenation of the cyclohexanol. At operating conditions, one of the significant problems in the industrial process is fast deactivation of the copper based catalysts. The present work focuses on the formulation of two types of the supported catalysts namely supported tri metals alloy (CuNiRu/N-rGO) in paper forms and supported copper (Cu/N-rGO), analysing the properties of the synthesised catalyst support (N-rGO) by Brunauer-Emmett-Teller (BET), X-ray photoelectron spectroscopy (XPS), Temperature-programmed desorption (TPD-NH3), Temperature Programmed Reduction (TPR-H2) and X-ray diffraction (XRD) as well as to investigate the catalytic performance of the two supported catalysts in the dehydrogenation of cyclohexanol to the cyclohexanone. All experiments on the catalytic performance were conducted at moderate temperatures (200–270 °C), 1 atm, 0.1 ml/min cyclohexanol flow rate and ∼8 h time on stream (TOS). The performances of the catalysts were evaluated in the gas phase dehydrogenation of cyclohexanol to the cyclohexanone. The conversion of the cyclohexanol using CuNiRu/N-rGO is 4% higher compare to use of the Cu/N-rGO. The selectivity for cyclohexanone in case of the Cu/N-rGO catalyst is about 83.88%, whilst, the CuNiRu/N-rGO illustrated approximately 6% better performance. The yield of the cyclohexanone using the Cu/N-rGO is about 78%, while by adding the Ni and Ru as promoters with the improvement of the Cu/N-rGO the yield of cyclohexanone was improved by 8%. The duration of the steady state period significantly improved by using CuNiRu/N-rGO which was proposed up to 7 times. This research shows that the CuNiRu/N-rGO catalyst provides the suitable and selective active sites for the dehydrogenation of cyclohexanol to the cyclohexanone reaction.

Data will be made available on request.An increasing demand for efficient electrochemical energy storage and conversion system in modern society has stimulated the development of novel rechargeable batteries that can realize the portable electronic revolution, in which highly sophisticated portable devices such as drones are widely and intensively utilized [1–5]. Rather than using conventional lithium-ion batteries that catch fire and explode, the emerging metal–air batteries in particular have been gaining increasing attention as they have the potential to contribute to the development of sustainable electrochemical energy and storage systems due to their high energy density and enhanced environmental friendliness [6,7]. For example, rechargeable zinc–air batteries (ZABs) have the potential to provide four times more energy (1086 Wh kg−1) in a sustainable and environmentally friendly manner than the most advanced lithium-ion batteries [8–14]. Additionally, the ZABs utilize a two-electrode system consisting of a catalytic air cathode and a nonprecious zinc anode, which significantly reduces the anthropological cost of these batteries [7,15–17]. However, practical challenges in zinc–air batteries, such as low cycle life, poor reversibility, and low energy conversion efficiency [18–20], persist because of higher energy barriers, including the slow kinetics in the electrocatalytic oxygen reduction reaction (ORR) and oxygen evolution reaction (OER), and poor electrochemical durability in alkaline electrolytes [21,22].For the improved bifunctional activity of electrocatalysts in ZABs, precious metals (e.g., Pt-based for ORR and Ir–Ru-based for OER) [23–25] and their corresponding oxides (IrO2 and RuO2) have typically been studied as the catalytic air cathode that uses molecular oxygen as fuel for the electrocatalytic reaction [26,27]. Alternatively, nonprecious metals (i.e., Fe, Co, and Ni), transition-metal composites (oxides, nitrides, and phosphides), and metal-supported carbon–based hybridization materials have also been investigated [28–32]. However, the long-term operation of these single metals and metal oxide–based composite catalysts is characterized by an inherent low conductivity and rapid deactivation of metal sites, resulting in an unbalanced symmetry of OER and ORR [33–35].To address the aforementioned issues, intensive research has been conducted on the development of heterogeneous electrocatalysts containing both OER and ORR elements [15,36]. Accordingly, multielement random alloy (MRA) catalysts, in which more than one metal elements are randomly mixed together to produce high-tech metal materials with modified crystal and electronic structure, have been considered for the desirable bifunctionally in active catalysts owing to their high conductivity, surface area, and selective active sites [37–45]. In particular, the electrocatalytic features of MRA can be further extended when it change the chemical composition (i.e., mixing ratio of each component), which give a rise to the resultant crystal structure to determine a stability of active metal particles and a charge distribution with internal resistance. Unfortunately, however, predicting the desirable crystal structure and relevant electrocatalytic properties of MRA depending on the metal composition have not been investigated yet due to the thermodynamic complex of each component. In parallel, sophisticated control of metal composition and crystal structure of MRA during the fabrication are also required to ensure their electrochemical catalytic properties for the efficient rechargeable zinc-air batteries. In addition, the electrocatalytic properties of MRA can be enhanced further by coupling with a metal oxide-based electrocatalyst. Because, the integrating MRA into metal oxide system and building a new multi-component dual-phase (MRA/metal oxide) electrocatalyst with different heterointerface can induce a synergistic effect between various metals as well as expose more active sites and efficient charge transfer/redistribution, resulting in a high symmetry of OER and ORR, providing superior electrocatalytic performance [46,47]. Unfortunately, the electrical properties and durability of dual-phase electrocatalysts continue to lag behind commercialization due to a lack of knowledge of their composition–structure–transport relationship.Herein, we report a novel dual-phase electrocatalyst comprised of AgNi random alloy and CoNb2O6 nanocube as the bifunctional multicomponent system derived using a sequential hydrothermal synthesis. By employing virtual crystal approximation (VCA), the optimal composition design and crystal structure of AgNi can be determined to have a specific atomic ratio of 6:4 (Ag:Ni) and a hexagonal closed-packed (hcp) structure, resulting in the highest electrical conductivity (σ ∼2 × 107 Scm−1) and ionized potential (∼−5.4 eV). Based on this information, we have successfully fabricated Ag0.6Ni0.4 electrocatalysts that are distributed on top of CoNb2O6 nanocubic electrocatalysts via a sequential hydrothermal process, allowing sophisticated control of chemical composition. The resultant dual-phase CoNb2O6 @Ag0.6Ni0.4 offers a high surface-to-volume ratio, exposed active sites and defect-enriched surface, thereby enhancing the OER/ORR bifunctional activity and charge transports. The CoNb2O6 @Ag0.6Ni0.4 catalyst exhibits outstanding electrochemical activity (Ej = 10 (OER) − E1/2 (ORR) = 0.49 V) and excellent ORR and OER cycle durability. Furthermore, the dual-phase CoNb2O6 @Ag0.6Ni0.4 catalysts are directly applied as an air cathode for zinc–air batteries, providing a stable discharge/charge voltage gap of 0.81 V over 587 h at a current density of 10 mAcm−2, also delivers excellent peak power density (178.9 mW cm−2 at 213 mA cm−2) and specific capacity (806.8 mA h g−1). From a practical perspective, we also designed pouch-type zinc–air batteries using CoNb2O6 @Ag0.6Ni0.4 catalysts as the air cathode, which exhibit an excellent rate capability, peak power density (135.6 mW cm−2 at 150 mA cm−2) and long-term stability for more than 158.6 h at a current density of 10 mA cm−2.Through density functional theory (DFT), we initially examined the electronic structural and electric transport properties of Ag1−xNix random alloy as a function of Ag content for hcp and fcc structures. Ag1−xNix random alloy (0 ≤x ≤ 1) was thoroughly optimized using the VCA (R1) provided in the Vienna ab initio simulation package (VASP) (R2) for the hcp and fcc structures. Our DFT calculations employed the frozen-core projector augmented wave method (R3) to describe the core-valence interaction using the generalized gradient of Perdew, Burke, and Ernzerhof (R4) for the exchange-correlation functional with a cut-off energy of 450 eV for plane waves, a set of 500 k-points for the irreducible Brillouin zone, self-consistent-field convergence threshold of 10 − 5 eV, and atomic force of 0.1 meV/Å. The electric transport properties of Ag1−xNix random alloy (0 ≤x ≤ 1) for the hcp and fcc structures were simulated at 400 K using the BoltzTrap code (R5) with dense k-mesh, specifically a set of 10,000 k-points for the irreducible Brillouin zone.The multielement random alloy based CoNb2O6 @Ag0.6Ni0.4 heterogeneous electrocatalyst was prepared using a sequential two-step process.Initially, highly crystalline CoNb2O6 nanocubes were synthesized using hydrothermal methods. In a typical synthesis of CoNb2O6 nanocubes on an fluorine-doped tin oxide FTO substrate, cobalt nitrate hexahydrate [Co(NO3)2.6H2O, 0.01 mol] and niobium ethoxide [Nb (OCH2CH3)5, 0.01 mol] were mixed in 50 mL of aqueous citric acid (0.02 mol). The reaction solution was homogeneously mixed for 40 min at 30 °C and magnetically stirred. The solution was agitated for 2 h after the addition of ethylene glycol (EG, 18 mL), polyvinyl alcohol (2.6 mmol), and hexamethylenetetramines (HMT, 0.002 mol). The resultant mixed solution was transferred to a Teflon-lined autoclave (100 mL) containing a CoNb seed layer-coated FTO plate (see the supporting Information for details on FTO cleaning and CoNb seed layer coating) and heated at 150 °C for 15 h. The obtained sample was rinsed with deionized water and ethanol before being dried at 30 °C for 1 h. The obtained sample was calcined in a muffle furnace at 500 °C for 2 h to produce a cube-shaped CoNb2O6 nanostructure.In accordance with standard protocol, nickel nitrate hexahydrate [Ni(NO3)2·6 H2O, 8 mmol], polyvinylpyrrolidone (PVP, 20 mg), ethylene glycol (EG, 12 mmol), and hexadecyltrimethylammonium bromide (CTAB, 2 mmol) were dissolved in 50 mL citric acid (8.4 g) to form a clear solution. Then, sodium borohydride (NaBH4, 5 mL) and silver nitrate (Ag(NO3)·H2O 18 mmol) were added to the solution, which was stirred in a round-bottom flask equipped with a reflux condenser at 50 °C for 30 min. The reaction mixture was transferred to a Teflon-lined autoclave and heated at 120 °C for 6 h. The as-obtained precipitate was collected and washed with ethanol and water, and the dry product was calcined at 600 °C for 2 h in a 10% H2/N2 flow at a heating rate of 2 °C/min to produce an Ag0.6Ni0.4 random alloy. Ag0.2Ni0.8, Ag0.4Ni0.6, and Ag0.8Ni0.2 random alloy nanoparticles were also prepared in the similar manner. For comparison, Ag and Ni metal particles were also prepared in the same way, with exclusion of Ni(NO3)·H2O or Ag(NO3)·H2O, respectively. After heat treatment, Ni, Ag and Ag0.6Ni0.4 samples were used for the textural characterization and electrochemical measurement (see the details in Figs. S1 and S2 of the Supplementary Material). Elemental stoichiometric ratio of the prepared AgNi random alloy samples were examined with ICP-OES (ICP-OES, Perkin Elmer Optima 8300, see the Table S1).The prepared Ag0.6Ni0.4 random alloy atoms were embedded in the CoNb2O6 nanocubes using a simple hydrothermal method. To embed the CoNb2O6 nanocubes, 30 mg of Ag0.6Ni0.4 random alloy powder was dispersed into a solution comprising a mixture of 20 mL acetyl-acetone (nanoparticle-capping agent) and 5 mL α-terpinol (binding agent). The resulting mixture was sonicated for 30 min at room temperature. After diluting the colloidal mixture with a 1:1 ratio of ethanol and water, it was transferred to a 100 mL Teflon-lined stainless steel autoclave containing a CoNb2O6 nanocubes-coated FTO plate and heated at 100 °C for 1 h. The collected sample was dried and then kept at 400 °C for 1 h to obtain a final CoNb2O6 @Ag0.6Ni0.4 catalyst. Bulk composition of CoNb2O6 and CoNb2O6 @Ag0.6Ni0.4 samples were examined with ICP-OES analysis (see the Table S2) CoNb2O6 @Ag0.2Ni0.8, CoNb2O6 @Ag0.4Ni0.6, and CoNb2O6 @Ag0.8Ni0.2 samples were also prepared in the similar manner. For comparison, CoNb2O6 @Ag and CoNb2O6 @Ni catalysts were also prepared in the same way, with exclusion of Ni(NO3)·H2O or Ag(NO3)·H2O, respectively.The physicochemical properties of the prepared catalyst were studied via X-ray diffraction (XRD), scanning electron microscopy (SEM), Energy-dispersive X-ray spectroscopy (EDX), high-resolution transmission electron microscopy (HR-TEM), N2 adsorption/desorption, X-ray photoelectron spectra (XPS), X-ray absorption near-edge structure (XANES), extended X-ray absorption fine structure (EXAFS), electron paramagnetic resonance (EPR), and grazing incident wide-angle X-ray scattering (GIWAX) (see Supplementary Material for detailed instrumentation and experimental protocols).Linear-sweep voltammetry (LSV) and cyclic voltammetry (CV) data were collected using potentiostat/galvanostat/ZRA based on a conventional three-electrode set-up (Gamry reference 300). Electrochemical impedance spectra were conducted using an AUTOLAB/PGSTAT 128 N analyzer at a frequency range of 100 kHz–0.10 mHz in a 1 M KOH solution. The oxygen (O2) evolution performance was estimated in a closed cell through in situ gas chromatography (YL instrument 6500 GC system) to analyze the headspace. ORR activity of catalysts was evaluated using a rotating disk electrode (RDE, Biologic Science Instruments) connected to a Dy 2300 potentiostat in an O2-saturated 0.1 M KOH solution (see the Supplementary Material for detailed experimental protocols).A two-electrode cell with a CoNb2O6 @AgNi catalyst–dispersed gas diffusion layer (GDL) as the air cathode and a polished Zn plate as the anode was used to test the performance of the rechargeable Zn–air battery. Catalyst ink was spray-coated onto a GDL surface to prepare the air cathode for the Zn–air battery. In accordance with the standard technique, 10 mg catalysts, 8 mL deionized water, 2 mL isopropanol, and 0.05 mL Nafion were mixed and sonicated in an ultrasonic bath to produce a homogenous ink. Subsequently, 10 µL of this ink was coated onto the GDL surface and dried to obtain a mass loading of ∼25 µg/cm2. The Zn–air battery performance of catalysts was examined in a 6 M KOH solution containing 0.2 M zinc acetate (ZnC4H6O4) electrolyte.The rechargeable Zn–air battery performance of the catalysts was evaluated using a two-electrode cell in a 6 M KOH solution containing a 0.2 M zinc acetate (ZnC4H6O4) electrolyte. The discharge polarization curve and peak power density of catalysts were measured using a Biologic potentiostat (Biologic, VMP-3) device. The discharge/charge cycling performance of catalysts was measured using the Wonatech cycler system (Wonatech, WBCS3000). Additional details of the Zn–air batteries measurement and the module of Zn–air battery pouch cell fabrication have been provided in the Supplementary Material.A typical crystal structure of neat Ag and Ni is hcp and fcc, respectively ( Fig. 1 a). To understand the impact of metal composition on their crystal structure, electrical properties, and stability, we consider Ag1−xNix random alloy (0 ≤x ≤ 1) at the first stage using VCA (see more details in Supplementary Material and experimental section) [48,49]. Interestingly, we found that the calculated relative energy of Ag1−xNix random alloys are in between hcp and fcc structure and prefer to have the hcp structure with the lowest energy minima when approaching the portion of Ni at x ∼ 0.4 ( Table S3 ). The calculated work function of Ag1−xNix random alloy gradually increased after adding the Ag component and approached the maximum value of ∼− 5.4 eV at x ∼ 0.4 ( Fig. 1 b). This indicated that the increasing portion of Ni upon Ag can promote a structural phase of transition from fcc to hcp structure with the lowest energy minima, which is in a good agreement with the earlier publications [50,51]. We anticipated that the change crystal structure of Ag1−xNix random alloy influences their electrical properties. In fact, the electrical conductivity σ of Ag1−xNix random alloy (0 ≤x ≤ 1) increases gradually and then increases drastically when x ∼ 0.4 with increasing Ag content, especially electrical conductivity σ of pure Ag is significantly greater than that of pure Ni ( Fig. 1 c).The significant change in the electrical conductivity of Ag1−xNix random alloy as a function of composition ratio prompted the hypothesis that the magnetic moment of Ag1−xNix random alloy can be modulated. The calculated magnetic moment of Ag1−xNix random alloy (0 ≤x ≤ 1) for the hcp and fcc structures are provided in Fig. 1 d. The magnetic moment disappeared near x ∼ 0.4, indicating that the electronic structure of Ni at the Fermi level should be derived from the s/p-state rather than the d-state. Hence, we evaluated the electronic structures of pure Ag, pure Ni, and Ag0.6Ni0.4, which revealed that the Fermi region of the pure Ni system is dominated by an incompletely filled d-state; however, the d-orbitals of pure Ag and Ag0.6Ni0.4 are completely filled ( Fig. S3 ). This implied that the Fermi region of pure Ag and Ag0.6Ni0.4 are contributed from s/p states, where the effective mass m* is substantially decreased. Therefore, it is plausible to conclude that the rapid increase in electrical conductivity (σ) of Ag1−xNix random alloy (0 ≤x ≤ 1) at x ∼ 0.4 is due to the removal of magnetism associated with the electronic structure near the Fermi level, which is mostly contributed by the s/p-state rather than the d-state.The facile design and formation route for multielement random alloy-based CoNb2O6 @Ag0.6Ni0.4 heterogeneous electrocatalyst is illustrated in Fig. 2a. The corresponding distinct grazing incident wide angle X-ray scattering (GIWAXS) patterns and summary of lattice parameters of the Ag, Ni, Ag0.6Ni0.4, and CoNb2O6 @Ag0.6Ni0.4 samples are shown in Figs. S4a-4d and Table S4. The observed diffraction peaks of Ag metal particles for q (Å−1) = 1.5019, 2.3787, 2.6351, and 3.0474 are indexed to the (100), (001), (120), and (200) crystal planes of the hcp structure, respectively. Additionally, the diffraction peaks of Ni metal for q (Å−1) = 3.0082, 3.5939 and 4.9896 are associated with the reflection planes (111), (200), and (220), respectively, corresponding to the fcc strucutre. Interestingly, we found a minor displacement of the diffraction angles upon Ag0.6Ni0.4 nanoparticles as compared to those of Ag and Ni metal particles. This indicated the formation of Ag0.6Ni0.4 random alloy, conforming to hexagonal crystal structure with lattice parameters a = 3.3357 Å, b = 3.3357 Å, c = 3.4151 Å, α = 90°, β = 90° and γ = 60°, which is in good agreement with the results of VCA above. Moreover, a structural analysis based on the diffraction peaks of CoNb2O6 revealed that their crystal structure correspond to an orthorhombic unit cell with lattice parameters a = 5.7219 Å, b = 14.149 Å, c = 5.0489 Å, α = 90°, β = 90° and γ = 90°. After Ag0.6Ni0.4 random alloy was deposited on top of CoNb2O6 nanostructures, the GIWAXS patterns for CoNb2O6 @Ag0.6Ni0.4 nanostructures indicated that they remained their own crystal structure. The X-ray diffraction patterns of CoNb2O6, Ag0.6Ni0.4, and CoNb2O6 @ Ag0.6Ni0.4 are displayed in Fig. S5, which corresponds to the standard PDF cards (ICDD –PDF-032–0304, JCPDS-04–0783 and JCPDS-04–0850). Interestingly, the diffraction peak of AgNi random alloy was found on the CoNb2O6 @AgNi XRD result, proving that a dual-phase CoNb2O6 @AgNi was successfully synthesized.The morphological characteristics of the prepared samples were analyzed through SEM and TEM. The low- and high-magnification SEM images of CoNb2O6 exhibit the typical hierarchical nanocube morphology ( Fig. 2 b, c and Fig. S6a , b). Subsequently, the atomic percentage and purity of Co, Nb, and O (14.40/27.26/58.34) were confirmed through corresponding energy dispersive X-ray spectra and elemental mapping ( Fig. S7 ). Fig. 2 d shows the spherical-like structure of Ag0.6Ni0.4 random alloy particles with porous nature. The EDX spectra with elemental mapping showed that the atomic percentages of Ag and Ni were 58.43% and 41.57%, respectively, which was very close to the stoichiometric ratio of Ag0.6Ni0.4 ( Fig. S8 ).After the incorporation of Ag0.6Ni0.4 random alloy nanoparticles on CoNb2O6 nanocubes, the morphology of the CoNb2O6 nanocubes was found to be highly preserved, as depicted by the SEM image shown in Fig. 2 e. Additionally, the SEM-EDX elemental mapping of CoNb2O6 @Ag0.6Ni0.4 demonstrates that Ag0.6Ni0.4 random alloy particles are uniformly distributed over the CoNb2O6 nanocubes ( Fig. S9 ). SEM images and EDX elemental mapping of CoNb2O6 @Ni and CoNb2O6 @Ag exhibited a similar trend to that of CoNb2O6 @Ag0.6Ni0.4, as shown in Figs. S10, S12 and Figs. S11, S13 , respectively.To investigate the crystal structure and lattice spacing of CoNb2O6 @Ag0.6Ni0.4, HR-TEM measurement was performed on the Ag0.6Ni0.4 sample. The Ag0.6Ni0.4 nanoparticles exhibited typical spherical shapes, and their lattice spacing of 0.204 and 0.236 nm was ascribed to the (111) and (111) plans (ICDD-00–004–0850, 00–004–0783) of metallic Ni and Ag of Ag0.6Ni0.4 random alloy, respectively ( Fig. 2 f, g and h). The corresponding TEM-EDS mapping validated the formation of the Ag0.6Ni0.4 random alloy shown in Fig. 2 i, j and k. The low and high-magnification images of CoNb2O6 showed a nanocube structure, which is consistent with SEM images ( Fig. 3 a and Figs. S14a, b ). The well-resolved lattice fringe distances of 0.172 nm observed in the HR-TEM image of Fig. 3 b correspond to the (062) crystal plane of CoNb2O6 (ICDD-00–032–0304). Furthermore, the elemental mapping image of CoNb2O6 shows that the nanocubes consist of only Co, Nb, and O ( Fig. 3 c). The TEM image of CoNb2O6 @Ag0.6Ni0.4 shows a homogeneous dispersion of Ag0.6Ni0.4 random alloy on the surface of the CoNb2O6 nanocubes, maintaining the original nanocube shape without structural degradation (Fig. 3 d, e and Figs. S14c-S14e ). The high-resolution HR-TEM image of CoNb2O6 @Ag0.6Ni0.4 ( Fig. 3 f, g, h, i and Figs. S15 ) reveals a distinct interface between CoNb2O6 and Ag0.6Ni0.4, confirming the coexistence of CoNb2O6 and Ag0.6Ni0.4 phase. The lattice spacing of 0.219 nm corresponds to the (231) crystal plane of CoNb2O6 (ICDD-00–032–0304) while the observed 0.236 and 0.204 nm lattice distances relate to the metallic Ag (111) and Ni (111) crystal planes of Ag0.6Ni0.4 random alloy, respectively.Selected area electron diffraction (SAED) patterns of CoNb2O6 @Ag0.6Ni0.4 ( Fig. 3 g insets) display polycrystalline rings corresponding to the (130), (131) and (220), (311) crystal plans of CoNb2O6 (ICDD-00–032–0304) and Ag0.6Ni0.4 (ICDD-00–004–0850, 00–004–0783) random alloys, respectively. Additionally, EDX elemental mapping of CoNb2O6 @Ag0.6Ni0.4 ( Fig. 3 j) reveals the homogeneous dispersion of Ag0.6Ni0.4 random alloy particles on the surface of CoNb2O6 nanocubes, demonstrating the successful synthesis of CoNb2O6 @Ag0.6Ni0.4 via the proposed sequential hydrothermal method.Through XPS, the elemental composition and surface valence state of the Ag0.6Ni0.4 random alloy incorporation effect were determined. The survey XPS spectrum of the CoNb2O6 @Ag0.6Ni0.4 catalyst implies the existence of Co, Nb, Ag, Ni, and O elements ( Fig. S16 ). Negative shifts were observed in the core level Co 2p and Nb 3d peaks of CoNb2O6 @Ag0.6Ni0.4 relative to CoNb2O6 (see the details in supplementary material Figs. S17a and S17b). In contrast, the Ag 3d peaks of CoNb2O6 @Ag0.6Ni0.4 have higher binding energies than those of Ag0.6Ni0.4 random alloy (see the details in Fig. S17c and Fig. S17d).The peaks at 853.3 and 870.9 eV and 855.2 and 872.8 eV [52–54] in the high-resolution Ni 2p region of Ag0.6Ni0.4 random alloy were attributed to spin-orbit doublets Ni0 and Ni2+ (Ni(OH)2), respectively ( Fig. S17e ). Ni0 peak at 853.7 and 871.4 eV and Ni (OH)2 peak at 855.5 and 873.1 eV were shifted to higher binding energies (∼0.4–0.5 eV) following the integration of the Ag0.6Ni0.4 random alloy on top of CoNb2O6 ( Fig. 4 a). The intensity of Ni0 peak decreased more compared to those of pristine Ag0.6Ni0.4, indicating changes in Ni oxidation states. Notably, the new peak at 854.1 eV [55,56] corresponds to NiO while the peak at 857.3 eV indicates the abundance of the Ni3+ state on the surface of CoNb2O6 @Ag0.6Ni0.4. Hence, it was plausible to conclude that the formation of Ni3+ species can play crucial roles in oxygen electrolysis [57,58].Three significant peaks belong to lattice oxygen (529.5 eV for M–O), hydroxyl species (531.1 eV for OH), and surface-adsorbed H2O (532.4 eV) in the O 1 s spectra of CoNb2O6 ( Fig. 4 b). However, the O 1 s spectra of CoNb2O6/Ag0.6Ni0.4 shows a new peak at 530.4 eV [59,60], which should be attributed to highly oxidative oxygen species (O2 2−/O−) and is associated with surface oxygen vacancies [61]. The relative ratio of peak area for four oxygen species is summarized in Table S5. A relative concentration of 33.23% O2 2−/O− species on the CoNb2O6 @Ag0.6Ni0.4 surface was reported to contribute to the superior ORR and OER activities [62].To evaluate the OER electrocatalytic activity, the synthesized catalyst and commercial RuO2 were used as working electrodes in a three-electrode test system containing a 1 M KOH electrolyte. As displayed in Fig. 5 a, the I–R corrected LSV polarization curve of the CoNb2O6 @Ag0.6Ni0.4 catalyst exhibits an overpotential of 110 mV for 10 mAcm−2, which is much less than those of CoNb2O6 (330 mV), CoNb2O6 @Ag (250 mV), CoNb2O6 @Ni (220 mV), and RuO2 (300 mV), and outperforms the recently reported metal catalysts at comparable conditions (Table S6). The OER performance of CoNb2O6 with different AgNi random alloy compositions is also depicted in Supplementary Material (Fig. S18 , Table S7 ). In Particular,CoNb2O6 @Ag0.6Ni0.4 required a relatively low overpotential (η) of 400 mV to achieve a current density of 100 mA cm−2, compared to CoNb2O6 @Ni (520 mV), CoNb2O6 @Ag (620 mV), and CoNb2O6 (700 mV). Accordingly, it was hypothesized that the outstanding activity of CoNb2O6 @Ag0.6Ni0.4 resulted from a strong electronic interaction between CoNb2O6 and Ag0.6Ni0.4 that modifies the local electronic structure of CoNb2O6 @Ag0.6Ni0.4, and it can be confirmed that Ag0.6Ni0.4 activates more reactive sites of OER. Furthermore, CoNb2O6/Ag0.6Ni0.4 exhibited a small Tafel slope value of 40 mV dec−1, which is lower than CoNb2O6 @Ni (50 mVdec−1), CoNb2O6 @Ag (54 mVdec−1), and CoNb2O6 (100 mV dec−1) counterparts and RuO2 (89 mV dec−1), indicating its fast OER reaction kinetics [63–65] ( Fig. 5 b). Reasonably, the Tafel slope of CoNb2O6 @Ag0.6Ni0.4 decreases due to the introduction of the Ag0.6Ni0.4 random alloy to provide a good connectivity of the CoNb2O6 interface, which enhances charge and mass transfer during OER.The excellent intrinsic activity of the CoNb2O6 @Ag0.6Ni0.4 catalyst is also evaluated by their larger turnover frequency (TOF) (0.2300 s−1 at an overpotential of 320 mV), which is significantly higher than those of CoNb2O6 @Ag (0.0118 s−1), CoNb2O6 @Ni (0.0166 s−1), and CoNb2O6 (0.0062 s−1) under the same overpotential condition, indicating more favorable reactive sites on the CoNb2O6 @Ag0.6Ni0.4 for OER reaction (details of the calculations are provided in the Supporting Information). Moreover, to gain a better understanding of the synergistic OER activity of CoNb2O6 and Ag0.6Ni0.4 random alloy, the electrochemical active surface area (ECSA) of the catalysts was determined using the electrochemical double layer capacitance (Cdl) derived from the CV curves in a non-Faradic potential region ( Figs. S19a-S19d ). As shown in Fig. S19e in the Supplementary Material, CoNb2O6 @Ag0.6Ni0.4 exhibited a larger Cdl of 4.89 mF cm−2 than CoNb2O6 @Ag (2.10 mF cm−2), CoNb2O6 @Ni (2.60 mF cm−2), and CoNb2O6 (1.80 mF cm−2), indicating the highest ECSA for CoNb2O6 @Ag0.6Ni0.4 after incorporation of Ag0.6Ni0.4 (see calculation details in the Supplementary Material). Moreover, the calculated roughness factor (RF) of the catalysts has a trend similar to that of the ECSA following the order of CoNb2O6 @Ag0.6Ni0.4 > CoNb2O6 @Ag > CoNb2O6 @Ni > CoNb2O6 ( Table S8 ). CoNb2O6 @Ag0.6Ni0.4 with a higher RF value has been shown to have a more favorable exposed active surface for oxygen electrolysis [66].The stability test of CoNb2O6 @Ag0.6Ni0.4 was performed using continuous chronoamperometric responses, which revealed that the initial OER current density was almost maintained for 264.8 h at 220 mV, which is significantly better than the CoNb2O6 @Ni (200.8 h at 280 mV), CoNb2O6 @Ag (165.5 h at 330 mV) and CoNb2O6 catalyst (a loss of 0.80% was observed after the continuous chronoamperometric operation of the CoNb2O6 catalyst for 149 h at 380 mV ( Fig. 5c, Fig. S20 )). Furthermore, there is low deterioration (0.01 V) and a positive shift of the OER polarization for CoNb2O6 @Ag0.6Ni0.4 after 264.8 h of continuous stability testing ( Fig. 4 a, red dotted line), indicating the outstanding structural stability of CoNb2O6 @Ag0.6Ni0.4. After performing the continuous OER activity, the structure of CoNb2O6 @Ag0.6Ni0.4 electrode was reexamined by the XPS, SEM and TEM under the same conditions as the initial measurement. The high resolution XPS spectra of the Co 2p, Nb 3d, and Ni 2p level spectrum in CoNb2O6 @Ag0.6Ni0.4 are shown in Fig. S21. The energy level XPS spectra of Co 2p and Nb 3d show slight negative peak shifts and the spectra of Ni exhibits a positive shift after the OER electrolysis. This suggests that constructing dual phase CoNb2O6 @Ag0.6Ni0.4 can effectively maintained the electronic structure of the Co, Nb core and optimize the adsorption of reaction intermediates, hence promoting electrocatalytic stability. SEM and TEM images of the CoNb2O6 @Ag0.6Ni0.4 electrode during and after a stability test are shown in Figs. S22, S23. The overall shape and structure of the catalyst was maintained and still retained original nano cube structure of CoNb2O6 @Ag0.6Ni0.4, indicating the structural stability of the catalyst surface.The corresponding ORR activity of the catalysts was evaluated using a RDE in 0.1 M KOH. LSV and cyclic voltammetry (CV) results obtained in N2-saturated and O2-saturated 0.1 M KOH for various catalysts are shown in Supplementary Material Fig. S24, S25. Polarization curves reveal that CoNb2O6 @Ag0.6Ni0.4 exhibits superior ORR activity compared to all other prepared catalysts (CoNb2O6, CoNb2O6 @Ni, CoNb2O6 @Ag, and commercial Pt/C) in terms of a more positive onset potential of 1.12 V, higher half-wave potential (E1/2) of 0.85 V, and a higher limiting current density of -5.60 mAcm-2 ( Fig. 5 d and Table S9 ). Notably, the ORR activity of the CoNb2O6 @Ag0.6Ni0.4 catalyst outperforms the nonprecious metal catalyst reported in the literature ( Table S6 ). Additional ORR polarization curves of CoNb2O6 with various AgNi random alloy compositions are also shown in the Supplementary Material (Fig. S26 , Table S7 ). The superior ORR performance of CoNb2O6 @Ag0.6Ni0.4 led us to believe that the random alloy composition of integrated CoNb2O6 and Ag0.6Ni0.4 can provide synergistic effects for the ORR electrolysis.Additionally, we found that the measured Tafel slope of CoNb2O6 @Ag0.6Ni0.4 is 50 mVdec−1, which is less than that of CoNb2O6 @Ag (69 mVdec−1), CoNb2O6 @Ni (73 mVdec−1), CoNb2O6 (75 mVdec−1), and Pt/C (64 mVdec−1), indicating the fast ORR kinetics on the CoNb2O6 @Ag0.6Ni0.4 ( Fig. 5 e). The electron transfer number, n, of the CoNb2O6 and CoNb2O6 @Ag0.6Ni0.4 catalysts was also determined from the Koutecky–Levich (K–L) plots under various potentials (0.3–0.7 V), which were obtained from ORR polarization curves at various rotation speeds (400–1600 rpm, Figs. S27a and S27b). The K–L plots of CoNb2O6 and CoNb2O6 @Ag0.6Ni0.4 at different potentials exhibit excellent linearity and near parallel fitting, revealing typical first-order reaction kinetics (Figs. S27c and S27d ). Furthermore, the calculated electron transfer number (n per O2) for the CoNb2O6 and CoNb2O6 @Ag0.6Ni0.4 catalysts was 3.9 and 4, respectively, indicating that the ORR process followed the four-electron transfer pathway closely [67]. The ORR stability of the CoNb2O6, CoNb2O6 @Ag, CoNb2O6 @Ni and CoNb2O6 @Ag0.6Ni0.4 catalysts was further evaluated through chronoamperometry testing in an O2saturated 0.1 M KOH solution at 0.7 V ( Fig. 5 f, Fig. S28 ). The CoNb2O6 @Ag0.6Ni0.4 catalyst retains 98% of the initial ORR current density after 24.5 h of continuous chronoamperometric performance, indicating that the CoNb2O6 @Ag0.6Ni0.4 catalyst has a well ORR-stability. In comparison to the CoNb2O6 @Ag0.6Ni0.4 electrocatalyst, CoNb2O6 @Ag (96% after 20.1 h), CoNb2O6 @Ni (96% after 18.6 h) and pristine CoNb2O6 (95% after 13 h) exhibit lower stability, which further confirming that the constructed dual-phase sample of CoNb2O6 @Ag0.6Ni0.4 has the best ORR electrocatalytic performance. Additionally the nearly identical ORR curves before and after durability test provide further evidence of the excellent stability of CoNb2O6 @Ag0.6Ni0.4 (depicted using the red dotted line in Fig. 5 d).The plotted histogram compared the OER overpotential at 10 mA cm−2 and the ORR half-wave potential of prepared catalysts, revealing the significantly enhanced ORR and OER performance of CoNb2O6 @Ag0.6Ni0.4, which is superior to CoNb2O6 and RuO2 and surpasses that of Pt/C ( Fig. 5 g). Due to the excellent oxygen electrolysis, the overall OER and ORR activities of catalysts were further analyzed ( Fig. 5 h). The potential difference ΔE (ΔE = Ej = 10 − E1/2) between the OER potential at 10 mAcm−2 and the ORR half-wave potential (E1/2) can be used to estimate the bifunctional activity of a catalyst. In general lower the ΔE value of an electrode, the greater its bifunctionality [68]. As shown in Fig. 5 h, CoNb2O6 @Ag0.6Ni0.4 exhibits a small ΔE value of 0.49 V, which is smaller than those of CoNb2O6 (0.99 V) and the metal-based bifunctional catalysts as reported in Tables S6.Additionally, in situ gas chromatographic analysis was also conducted to confirm the amount of oxygen evolution. The Faradaic efficiency of CoNb2O6 and CoNb2O6 @Ag0.6Ni0.4 catalysts at 10 mA cm−2 with oxygen evolution time was measured. As shown in Fig. 5 i and Fig. S29, the amount of O2 volume detected with increasing operation duration is comparable to the volume calculated at constant current density, demonstrating a nearly 100% Faradaic efficiency during the OER. Furthermore, CoNb2O6 @Ag0.6Ni0.4 exhibited excellent methanol tolerance without any change in the ORR current when 0.5 mL (3 M) methanol was added to the O2-saturated electrolyte at 520 s ( Fig. S30 ).Overall, the electrocatalytic efficiency of CoNb2O6 @Ag0.6Ni0.4 for OER and ORR has improved significantly, likely due to the efficient charge transport and redistribution at the interface between CoNb2O6 and Ag0.6Ni0.4. To gain a deeper understanding of the catalytic interaction at the CoNb2O6 @Ag0.6Ni0.4 interface, we conducted DFT calculations to calculate the electrical conductivity of CoNb2O6 based on its charge carrier density (ne) and oxidation numbers ( Fig. 6 a). Interestingly, we found that the conductivity of CoNb2O6 could be further enhanced through oxidation or reduction compared to neutral CoNb2O6, with a conductivity of approximately 3 × 103 S/m. We also observed that the conductivity of CoNb2O6 gradually increased as its oxidation number increased up to + 3, but decreased when its oxidation number decreased down to + 3. These results can be explained by changes in the density of states (DOS). When electrons are removed from CoNb2O6 (i.e., cation), the Fermi level moves closer to the broad valence band maximum (VBM), referred to as 1, which leads to a decrease in effective mass (m*) and an increase in conductivity (Fig. 6 b). Conversely, when electrons are added to CoNb2O6 (i.e., anion), the Fermi level moves closer to the sharp conduction band maximum (CBM), referred to as 2, which leads to an increase in m* and a decrease in conductivity.Then, we compared the relative energy of CoNb2O6 and Ag0.6Ni0.4 using DFT calculations, considering their oxidation levels (Fig. 6 c). The relative energy of CoNb2O6 was found to be slightly lower than that of Ag0.6Ni0.4 when an electron was removed from CoNb2O6. Conversely, the relative energy of Ag0.6Ni0.4 was lower than that of CoNb2O6 after an electron was added to CoNb2O6. These results imply that it is possible to secure a favorable oxidation state for efficient charge transport.The local electronic properties, a coordination environment and bond distance of Ag0.6Ni0.4, CoNb2O6, and CoNb2O6 @Ag0.6Ni0.4 catalysts were analyzed through XANES. As shown in Fig. 7 a, the normalized Co L 3pre-edge XANES for CoNb2O6 @Ag0.6Ni0.4 showed a slightly lower energy compared to pristine CoNb2O6, indicating a shift in local coordination due to the presence of Ag0.6Ni0.4 random alloy. The Ni L-edge XANES spectra for CoNb2O6 @Ag0.6Ni0.4 showed apparent changes from those of Ag0.6Ni0.4 random alloy ( Fig. 7 b). The fine multiple splitting of the Ni L3 and Ni L2 peaks was attributed to the crystal field effects from the electronic structure. These characteristics, which result from the dipole transition from the 2p to the empty 3d state, were sensitive to changes in the oxidation state and local geometry of Ni [69].Ni and Co coordination conditions in CoNb2O6 @Ag0.6Ni0.4 were further investigated using K-edge XANES spectra for Ni and Co. As shown in Fig. 7 c, the shape of the pre-edge (1 s to 3d transition) and absorption edge (multiple scattering) of the normalized Ni K-edge spectra obtained from CoNb2O6 @Ag0.6Ni0.4 differs from that of the Ag0.6Ni0.4 random alloy, indicating that there are probably local lattice strain changes around the targeted Ni atom. The Ni K-edge spectra of CoNb2O6 @Ag0.6Ni0.4 exhibited an increase in the white line and a shift of the absorption edge toward high photon energy compared to those of Ag0.6Ni0.4 random alloy, indicating that the Ni has more empty d‐orbital states and less electron density. Thus, the results matched the oxidation valence of Ni (Ni2+ is oxidized to Ni3+/Ni4+) [70]. Comparing the Co K-edge XANES spectra ( Fig. 6 d) of CoNb2O6 @Ag0.6Ni0.4 to those of CoNb2O6, the white line is significantly reduced and the absorption edge shifts to slightly negative values, confirming the higher d‐orbital occupancy due to the surface charge polarization caused by the electron transfer [71]. Fig. 6 e shows the Fourier transform (FT) of EXAFS R-space Ni K-edge spectra (k2-weighted) for CoNb2O6 @Ag0.6Ni0.4. The dominant peak in the first coordination shell at 1.6 Å corresponds to the scattering path of Ni–O, whereas the second coordination shell peak at 3.5 Å originated from the scattering path of Ni–Ni [72,73]. Comparatively, the decreased peak intensities of the Ni–O and Ni–Ni coordination shells are associated with lower coordination numbers and defects in the structure, reflecting changes to the electronic structure [74,75]. Besides, the FT curve of Co K-edge R-space spectrum for CoNb2O6 @Ag0.6Ni0.4 exhibits the lowest EXAFS when compared to CoNb2O6, demonstrating its more disordered octahedral coordination around the Co atom [76] due to theAg0.6Ni0.4 random alloy incorporation ( Fig. S31 ). Additionally, the Co–O and Co–Co bond lengths of CoNb2O6 @ Ag0.6Ni0.4 are 1.4 and 2.6, [77] which is shorter than those of CoNb2O6 at 1.49 and 2.66. The contraction of Co–O and Co–Co bond length reflects the enhanced interaction between Ag0.6Ni0.4 random alloy and CoNb2O6, which shifted the CoNb2O6 @Ag0.6Ni0.4 peak toward the lower energy side, as shown in Fig. S31.As shown in Fig. 7 f, the CoNb2O6, CoNb2O6 @Ag0.6Ni0.4, and Ag0.6Ni0.4 random alloy samples exhibited significant changes near the O K-edge, indicating alterations in the local chemical and electronic structure around O atoms [78]. Notably, the main absorption edge peak below 535 eV corresponds to metal 3d band electronic transitions while the peaks at 536–543 eV relate to metal 4sp band electronic transitions [79]. Additionally, the normalized O K-edge intensity of CoNb2O6 @ Ag0.6Ni0.4 is lower than that of CoNb2O6, and the adsorption pre-edge shifts to higher energy coupled with a new peak at 533.4 eV, indicating characteristics of oxygen vacancies [80]. From the XANES- and EXAFS data, we concluded that the incorporation of Ag0.6Ni0.4 random alloy into CoNb2O6 surface had been significantly affecting the Ni atoms and oxidized to high-valence states (Ni3+), accompanying surface generate oxygen vacancies, which synergistically contribute to high ORR and OER catalytic activity of CoNb2O6 @Ag0.6Ni0.4.In addition, the K-edges of the Co and Ni of CoNb2O6 @Ag0.6Ni0.4 were found to shift compared to the bare CoNb2O6 and Ag0.6Ni0.4. These shifts can be attributed to the formation of the interface between CoNb2O6 and Ag0.6Ni0.4 and promising interfacial charge transfer from Ag0.6Ni0.4 to CoNb2O6. This drastically affected the local configuration of Ag0.6Ni0.4 and CoNb2O6 structural sites. These combined electronic and chemical properties of CoNb2O6 @Ag0.6Ni0.4 alter the ORR/OER properties of the catalyst surface, generating an architecture with high conductivity and potential for active site exposure. Additionally, they offer an intuitive means of revealing bifunctional activity by monitoring the strong electronic bonding between Ag0.6Ni0.4 random alloy and CoNb2O6. This allows them to maintain structural stability even when electrochemical conditions are extremely severe.To further validate the existence of oxygen vacancies, EPR spectra of the CoNb2O6 @Ag0.6Ni0.4 catalyst were acquired. As shown in Fig. 7 g, the CoNb2O6 @Ag0.6Ni0.4 catalyst exhibits a strong EPR signal at g = 2.001, which arises from the unpaired electrons trapped in the CoNb2O6 @Ag0.6Ni0.4 catalyst, thus proving the existence of oxygen vacancies induced by Ag0.6Ni0.4 random alloy incorporation. Thus, the CoNb2O6 @Ag0.6Ni0.4 catalyst is expected to minimize the charge transfer resistance during the electrolysis. Furthermore, Fig. 7 h displays the Nyquist plots derived from electrochemical impedance spectroscopy fitting for the CoNb2O6, CoNb2O6 @Ag, CoNb2O6 @Ni, and CoNb2O6 @Ag0.6Ni0.4 electrodes. The fitting parameters are estimated and listed in Table S10 of the supporting Information. The results revealed a substantially lower charge transfer resistance (Rct) compared to CoNb2O6 from the fitting result, indicating that the significantly improved interfacial charge transfers can promote the reaction kinetics of the oxygen electrolysis.Additionally, the N2 sorption isotherms (type IV isotherm with H3 hysterics loop) of the CoNb2O6 @Ag0.6Ni0.4 catalyst indicate the existence of mesopores ( Fig. 5 i). Moreover, the catalyst exhibits a high Brunauer–Emmett–Teller (BET) surface area of 198 m2 g−1 compared with that of CoNb2O6 electrode (93 m2 g−1), which indicates that the Ag0.6Ni0.4 random alloy incorporation positively enlarges the intrinsic surface area of CoNb2O6 @Ag0.6Ni0.4. Fig. 7 i inset depicts the Barrett–Joyner–Halenda pore size distribution of CoNb2O6 @Ag0.6Ni0.4 catalyst, with the majority of pores falling into a mesopores size range of 5–50 nm; these mesoporous and high specific surface area of CoNb2O6 @Ag0.6Ni0.4 significantly allow more active site, facilitating accessible mass transfer and collection efficiencies during electrolysis [81,82]. The estimated pore size, BET surface area, and pore volume of the catalysts are summarized in Table S11.In addition to their excellent ORR and OER bifunctional electrocatalytic activity, rechargeable Zn–air batteries were evaluated to demonstrate the charge–discharge performance of CoNb2O6 and CoNb2O6 @Ag0.6Ni0.4 catalysts. Fig. 8 a illustrates a schematic diagram of a customized two-electrode Zn–air battery system. The Zn–air cell driven by optimized CoNb2O6 @Ag0.6Ni0.4 displays an open circuit potential (OCV) of 1.425 V ( Fig. S32 ), which is higher than the OCV of the Zn–air cell driven by pristine CoNb2O6 (1.41 V). This finding demonstrates the lower internal resistance of CoNb2O6 @Ag0.6Ni0.4. The assembled zinc–air batteries polarization (I–V) curves and corresponding power density (P–V) plots are shown in Fig. 8 b. Based on the discharge curve, the peak power density of the CoNb2O6 @Ag0.6Ni0.4-air cathode battery (178.9 mW cm−2 at a current density of 213 mA cm−2) is higher than that of the Pt+C/RuO2 (131.8 mW cm−2 at a current density of 173 mA cm−2) and CoNb2O6-based Zn–air battery (107 mW cm−2 at a current density of 155 mA cm−2), indicating the high catalytic activity of CoNb2O6 @Ag0.6Ni0.4 even in practical Zn–air battery conditions. The specific capacity of air cathode at various current densities is shown in Fig. 8 c. At a discharge current density of 10 mA cm−2, the CoNb2O6 @Ag0.6Ni0.4 air-cathode–based battery demonstrates an excellent specific capacity of 806.8 mA h g−1, compared to 606.3 mA h g−1, 576.6 mA h g−1 of Pt+C/RuO2 and CoNb2O6. Furthermore, when the current density reached 20 and 50 mA cm−2, the CoNb2O6 @Ag0.6Ni0.4 cathode was still capable of discharging capacities of 788.2 and 726.4 mA h g−1, respectively, confirming the practical capability of the as-designed catalysts. Fig. 8 d displayed the Galvanostatic discharge–charge polarization cycles for Zn–air batteries based on CoNb2O6 @Ag0.6Ni0.4, RuO/PtO and CoNb2O6 air cathodes. The cycle tests were measured at room temperature at a current density of 10 mAcm−2. The Zn–air batteries with CoNb2O6 @Ag0.6Ni0.4 cathodes were able to operate a long life cycle at a voltage gap of 0.81 V without voltage loss for 587 h; however, the discharge voltage of Zn–air batteries with Pt+C/RuO2 and CoNb2O6 air cathodes decreased within 156 h and 286 h, respectively.Owing to the excellent power density, specific capacity, and cycling stability of CoNb2O6 @Ag0.6Ni0.4, the portable Zn–air pouch cell was constructed with CoNb2O6 @Ag0.6Ni0.4 as the air cathode (see the details in Supplementary Material). Images of the CoNb2O6 @Ag0.6Ni0.4 air cathode–based rechargeable Zn–air battery pouch cell are shown in Fig. 8 e and Fig. S33. The fabricated Zn–air pouch cell using CoNb2O6 @Ag0.6Ni0.4 catalyst as the air cathode ( Fig. 8 f) demonstrates an OCV of 1.41 V and a specific capacity of 716.6 mA h g−1 (at 30 mA cm−2 (Fig. S34 )) based on the mass of consumed Zn (93.4% of the theoretical capacity). The galvanostatic discharge measurements ( Fig. 8 g) of the CoNb2O6 @Ag0.6Ni0.4-catalyzed Zn–air pouch cell show a small voltage drop between 5 and 50 mA cm−2. Additionally, when the current density is lowered to 10 mA cm−2, the discharge can be reversed, demonstrating that the discharge voltage rate performance and their reversibility are excellent. The peak power density of the pouch cell was measured to be 135.6 mW cm−2 at 150 mA cm−2 ( Fig. 8 h). As shown in Fig. 8 i, the cycling stability of a CoNb2O6 @Ag0.6Ni0.4 air-cathode–based battery was studied at a current density of 10 mA cm−2 and a charging and discharging duration of 10 min each. The Zn–air pouch cell with CoNb2O6 @Ag0.6Ni0.4 air cathode has a small charge–discharge voltage gap of 0.86 V. After 158.6 h of cycling, CoNb2O6 @Ag0.6Ni0.4-based batteries exhibited a minor voltage drop of 0.03 V, demonstrating their excellent stability. Following cycling performance, three fabricated pouch cells integrated in series with ultrasonic welding (OCV is 4.23 V, a single battery is 1.41 V) lit an LED house for several hours ( Fig. 8 j), demonstrating the application potential of the prepared CoNb2O6 @Ag0.6Ni0.4 air cathode.In conclusion, we provided an efficient sequential hydrothermal method for the fabrication of highly dispersive mesoporous Ag0.6Ni0.4 random alloy nanoparticles on CoNb2O6 nanocubes. The random alloy based CoNb2O6 @Ag0.6Ni0.4 heterogeneous catalyst demonstrated outstanding ORR and OER activities and remarkable stability in alkaline environments. Detailed simulation and characterization data, including VCA, XPS, XNEAS, XRFA, and advanced electrochemical experiments, revealed that the improved electrochemical performance of the CoNb2O6 @Ag0.6Ni0.4 catalyst is likely attributable to the incorporation of Ag0.6Ni0.4 random alloy. This modified the electrical and chemical properties of CoNb2O6 and created CoNb2O6 @Ag0.6Ni0.4 with high conductivity and the possibility for active site exposure. The formation of defect-enriched surface, Ni3+ active intermediates, an abundance of highly oxidative oxygen species, and a mesoporous structure resulted in additional catalytic support, which improved overall electrochemical performance. The strong electronic bonding and structural advantages of CoNb2O6 @Ag0.6Ni0.4 facilitated charge transfer and ensured structural stability even under extreme electrochemical conditions. The CoNb2O6 @Ag0.6Ni0.4 air cathode delivered excellent specific capacity (806.8 mA h g−1 at 10 mAcm−2), power densities (178.9 mW cm−2 at 213 mA cm−2), and stable cycling life in Zn–air battery applications (587 h at 10 mA cm−2). Exceptionally, the designed pouch-type zinc–air batteries possessed a peak power density of 135.6 mW cm−2 at 150 mA cm−2, excellent rate capability, and a stable discharge/charge cycle life of over 158.6 h at 10 mA cm−2. The proposed fabrication of random alloy dispersed CoNb2O6 @Ag0.6Ni0.4 catalysts facilitated the construction of various metal alloy catalyst systems for numerous sustainable energy conversion technologies. Chandran Balamurugan: Project conception and organization, Investigation, Sample and device fabrication, Formal analysis, Interpretation of data, Visualization, Writing - original draft, Writing-review. Changhoon Lee: DFT calculation & Interpretation, Visualization. Kyusang Cho: Helped in device fabrication & characterization, Validation. Jehan Kim: GIWAXS measurements & Validation. Byoungwook Park: Graphical support. Woochul Kim: Helped in materials characterization. Namsoo Lim: Helped in materials characterization. Hyeonghun Kim: Helped in materials characterization. Yusin Pak: Equipment provision and discussions. Keun Hwa Chae: Materials characterization, Validation. Ji Hoon Shim: Equipment provision and discussions. Sooncheol Kwon: Project conception and organization, supervision, Interpretation of data, Writing & editing of manuscript.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.This research was supported by the Young Researchers Program of the NRF funded by the Ministry of Science, ICT & Future Planning (NRF-2021R1A2C4001904). This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (NRF-2022R1I1A1A01072238). This work was supported by the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT (NRF-2022M3H4A1A04074153; NRF-2020M3H4A2084417; NRF-2022M3C1A3091988). This research at MPK/POSTECH was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (NRF-2021R1F1A1063478).Supplementary data associated with this article can be found in the online version at doi:10.1016/j.apcatb.2023.122631. Supplementary material .

The integration of bifunctionally active sites of multielement random alloy catalysts with other metal oxide electrocatalysts is a promising strategy for efficient electrochemical reactions. In this study, a novel combination of virtual crystal approximation and hydrothermal synthesis was used to investigate the composition-dependent structure and electrical property in a Ag1−xNix catalyst. The combination showed that a hexagonal closed-packed structure of Ag1−xNix with a compositional ratio of 6:4 (Ag:Ni) had electrical conductivity of ∼2 × 107 S∙cm−1 and an ionization potential of − 5.4 eV. Furthermore, the bifunctional oxygen electrocatalytic efficiencies of Ag0.6Ni0.4 were improved by forming a heterointerface with the CoNb2O6 electrocatalyst, resulting in a discharge-charge voltage gap of 0.81 V over 587 h, peak power density of 178.9 mW∙cm−2, and specific capacity of 806.8 mA∙h∙g−1 in a zinc–air battery. This approach was applied to pouch-type zinc–air batteries, resulting in long-term stability of over 158.6 h at 10 mA∙cm−2.

Hydrogen (H2) economy is an envisioned future in which the main source of energy of a nation is derived from hydrogen storage and application in the short, medium, and long term [1]. H2 produced from green chemistry based on water electrolysis – mostly derived from the use of wind and solar energy sources [1] – is required to be stored and distributed in a proper fashion [2] until usage.Electrolyzers normally use platinum-group metals (PGM) - which are noble metals, as eletrocatalysts for the electrolysis of water [3] and for maintaining the green concept involving the application of H2 as a low-carbon energy source. H2 is used in fuel cell systems [4] which also normally employ noble metals as electrocatalysts for the generation of clean energy [3]. The reactions involved in water electrolysis and fuel cell systems are generally hydrogen and oxygen evolutions (HER and OER), hydrogen oxidation (HOR), and oxygen reduction (ORR) reactions [3].Generally, Pt-based electrocatalysts are the benchmark electrocatalysts for ORR and HER processes, while RuO2 and IrO2 are regarded the ideal electrocatalysts for OER processes [5]. The major disadvantages regarding the use of these noble metals/oxides lie in their high costs and scarcity. In view of that, there has been an increasingly growing interest in the search for cheap Earth-abundant elements which are equally efficient for application as electrocatalysts in ORR, HER and OER processes in place of the aforementioned noble metals/oxides [5,6].In the search for cheap Earth-abundant elements, Bezerra and Maia [5] and Martini and Maia [6] found crystalline NiCo2O4 and CoMoSe/GNR (Co(OH)2‒CoMoO4‒MoSe2/GNR) as extremely stable and highly efficient when applied as electrocatalysts for OER.Other catalysts reported to have improved OER electrocatalytic responses include the following: MoOx formed on the surface of N-doped MoS2 (MoOx@N-doped MoS2− x ) [7]; one-dimensional CoS2−MoS2 nano-flakes decorated MoO2 sub-micro-wires (CoMoOS) [8]; MoS2 quantum dots (MSQDs) [9]; metallic octahedral type molybdenum disulfide (1T MoS2) [10]; Fe-doped MoS2 nanosheets (Fe-MoS2 NSs) [11]; phosphorus incorporated cobalt molybdenum sulfide on carbon cloth (P-CoMoS/CC) [12]; MoSe2 nanosheet/MoO2 nanobelt/carbon nanotube membrane (MoSe2 NS/MoO2 NB/CNT-M) [13]; one-dimensional MoO2–Co2Mo3O8@C nanorods [14]; porous NiMoO4− x /MoO2 hybrids [15]; CeO2 shells on the surfaces of ZIF-67-derived porous N-doped Co3O4@Z67-NT (Co3O4@Z67-NT@CeO2, T = temperature) [16]; ultrathin Co3O4 nanomeshes (Co-UNMs) [17]; hierarchically structured Co3O4/NiCo2O4/Ni foam (CO/NCO/NF) composite [18]; defect-activated Co3O4 (DA-Co3O4) [19]; porous Co3O4/CoMoO4 nanocages [20]; Co3O4/MoS2 heterostructure [21]; crystal lattice distorted ultrathin cobalt hydroxide (CLD-u-Co(OH)2) nanosheets [22]; mixed NiO/NiCo2O4 nanocrystals [23]; and porous NiO/NiCo2O4 heterostructure [24].The central idea behind the development of the present work (the use of Co/Mo-based catalysts) was derived from our recently published work which reported the combination of graphene nanoribbons (GNRs), cobalt salt, and commercial MoSe2 for the synthesis of OER electrocatalysts; the combined application of these materials resulted in the construction of a suitable electrocatalyst containing Co and Mo oxides, some MoSe2, and GNR (CoMoSe/GNR electrocatalyst), which was highly efficient for OER [6]. The application of non-commercial MoS2 nanosheets and MoSe2 nanoribbons, in place of GNR, as supporting materials for the Co/Mo-based electrocatalysts (synthesized and studied in the present work) was found to be very useful. Clearly, the efficient performance of the Co/Mo-based electrocatalysts in OER helped confirm the suitability of the non-commercial MoS2 nanosheets and MoSe2 nanoribbons as supporting materials for the OER electrocatalysts proposed in this study.Taking these considerations into account, in the present work, the main idea was to use Ni and/or Co salts and urea in combination with MoSe2 and MoS2 to produce nanocomposites which are effectively capable of electrocatalyzing OER in alkaline solution through the application of hydrothermal and calcination methods. The physical characterizations of the nanocomposites revealed the following: i) CoMoO4 nanoparticle oxides are supported on MoSe2 nanoribbons for the NiCoMoSe nanocomposite; ii) NiCo2O4 (and CoMoO4) nanoparticle oxides are supported on MoSe2 nanoribbons for the NiCoMo nanocomposite; iii) Co2Mo3O8 (and CoMoO4) nanoparticle oxides are supported on MoSe2 nanoribbons for the CoMo nanocomposite; iv) CoMoO4, Co2Mo3O8, and Co3O4 nanoparticle oxides are supported on MoS2 nanosheets and MoSe2 nanoribbons for the CoMoSe nanocomposite. The factors that mainly contributed to the improvement of OER electrocatalysis for the four catalysts mentioned above can be summarized as follows: 1) N atoms, from the urea used in the synthesis, bonded to relatively high amount of Ni and/or Co (and Mo); 2) the electrons released (3 to 2 electrons) from the oxidation of Co from the 2+ to 3+ state (one electron released), Ni from the 2+ to 3+ state (one electron released), and Mo from the 4+ to 6+ state (two electron released) identified by the presence of Ni2+ and Ni3+, and/or Co2+ and Co3+, and Mo4+ and Mo6+ in the nanocomposites; and 3) the presence of MoS2 nanosheets and MoSe2 nanoribbons as supporting material for the metal oxides in the nanocomposites contributed to a reduction in both the ECSA values and the charge transfer resistance.The key advantages of the chosen method lie in the fact that it is a much simpler and straightforward technique which involves the production of efficient and relatively cheaper non-commercial supporting materials for OER catalysts compared to the tedious technique involving the production of GNRs. The proposed technique was used for the production of non-commercial MoS2 nanosheets and MoSe2 nanoribbons with markedly low charge transfer resistance (Rct), which were used as supporting materials in a relatively simple synthesis for the development of highly efficient and innovative low-cost catalysts (nanocomposites) for OER. Carbon paper was employed as electrode substrate due to its low cost, good conductivity, and relative stability in OER.All reagents used in this work were of analytical purity and were not subjected to any previous treatment before use. CoCl2•6H2O, MoSe2 (this compound was derived from two different sources; as such, it will be referred to as MoSe2–1), IrO2, and RuO2 were purchased from Sigma Aldrich. NiCl2•6H2O, KOH, and HNO3 were acquired from Vetec, and urea and H2SO4 were obtained from Neon. MoSe2 (MoSe2–2) and MoS2 were synthesized from stoichiometric mixtures of the elements, in evacuated and sealed ampoules (see below).All electrochemical measurements were performed in a three‒electrode system. A graphite plate was used as counter electrode, a reversible hydrogen electrode was used as reference electrode, and a carbon paper (CP) HCP030N (sheet of 1.0 cm2) or Au-disk/Pt-ring RRDE (0.196 and 0.11 cm2 geometric areas, respectively) with a collection efficiency of N = 0.26, according to the manufacturer's information, was employed as working electrode.Synthesis of MoS2: Molybdenum metal was annealed at 1000 °C in H2 stream prior to the synthesis of the materials. Molybdenum powder (3.258 g, 0.03395 mol) and crystalline sulfur (2.178 g, 0.06791 mol) were placed in a quartz ampoule, evacuated under dynamic vacuum, and sealed. The ampoule was heated in a muffle furnace up to 650 °C for 7 h, kept at this temperature for 72 h, and cooled thereafter in the furnace. The MoS2 product obtained was a black homogeneous powder.Synthesis of MoSe2–2: Molybdenum powder (1.936 g, 0.02018 mol) and selenium powder (3.183 g, 0.04031 mol) were placed in a quartz ampoule, evacuated under dynamic vacuum, and sealed. The ampoule was heated in a muffle furnace up to 850 °C for 9 h, kept at this temperature for 100 h, and cooled thereafter in the furnace. The MoSe2 product obtained consisted of a black homogeneous powder with metallic luster.To prepare CoMoSeS, 16 mg of MoSe2–1, 16 mg of MoS2, 100 mg of CoCl2•6H2O and 500 mg of urea were weighed and mixed together, and 30 mL of ultrapure water was added to the mixture. The mixture was subjected to ultrasonication for 20 min. After that, the mixture was transferred into a Teflon‒lined stainless‒steel autoclave and subjected to heating in a muffle furnace at a temperature of 180 °C for 24 h (Scheme 1 ). After it was cooled to room temperature, the mixture was washed with ultrapure water by centrifugation several times and dried at 40 °C in an oven for 12 hTo prepare CoMo, an amount of CoMoSeS was placed inside a quartz boat, taken to a muffle furnace and heated at a temperature of 600 °C for 3 h (Scheme 1). CoMo/AL was produced by adding 30 mL of a solution of 0.5 M HNO3:0.5 M H2SO4 to CoMo and keeping the mixture under magnetic stirring and heating (50 °C) for 8 h (Scheme 1). Thereafter, the mixture was washed with ultrapure water by centrifugation until neutral pH was obtained. The mixture was then dried at 40 °C for 12 h in an oven.NiCoMoSe was prepared by hydrothermal method. Specifically, 16 mg of MoSe2–2, 100 mg of NiCl2•6H2O, 100 mg of CoCl2•6H2O, 500 mg of urea and 30 mL of deionized water were mixed and kept under ultrasonic bath for 20 min. The dispersion was transferred into a Teflon‒lined stainless‒steel autoclave and heated to 180 °C for 24 h (Scheme 1). After cooling to room temperature naturally, the product was washed with ultrapure water several times by centrifugation, and finally dried in an oven at 40 °C for 12 hTo produce NiCoMo, a fraction of the prepared NiCoMoSe was calcined in a muffle furnace at 600 ºC for 3 h (Scheme 1).The sequence of syntheses described above (Scheme 1) is similar to the syntheses performed by Bezerra, Martini, and Maia (2020,2021) [5,6].The CP sheet was cleaned by sonication in deionized water several times. The Au disk electrode was polished with alumina (1 and 0.05 μm meshes) until a mirror-like surface was obtained. After that, the electrode was sonicated in ultrapure water, acetone, and then again in ultrapure water for 5 min in each solvent. The electrode was then subjected to 10 voltammetry cycles using a scan rate of 50 mV s − 1 and potential range of 0.05 to 1.70 V in N2 saturated 0.5 M H2SO4. During this analysis, the solution was changed when needed [6].The modified working electrodes were prepared by dripping an aqueous solution of the catalysts (1 mg mL−1 concentration) on the surface of the electrodes; this produced a uniform thin film with a 150 µg cm‒2 loading. After drying at room temperature, the modified electrodes were immersed in deionized water prior to being immersed in the electrolyte (1 M KOH aqueous solution), which was saturated with N2 (5.0 purity) or O2 (4.0 purity) - both gasses were acquired from White Martins.The procedure employed by Trotochaud et al. [25] was used for the purification of Fe in order to test the interference of Fe from the KOH electrolyte (see details in the Supporting Information) in the OER responses.The morphology and distribution of the nanocomposites and nanoparticles were analyzed by TEM, STEM, and EDX using JEOL JEM 2200F, FEI TECNAI G² F20 HRTEM, and JEOL JEM 2100 plus microscopes with electron beam at 200 kV. Non-electrochemically (i.e., as synthesized material) and electrochemically stabilized (denoted by es; i.e., material obtained after long-term electrochemical stability test) CoMoSeS, CoMo, NiCoMoSe, and NiCoMo nanocomposites were diluted in ultrapure water and applied by dripping on a 400-mesh copper grid from Ted Pella with ultrathin carbon films supported on lacey carbon film. The microstructure of the nanocomposites and nanoparticles was visually characterized by SEM using a JEOL JSM-6380LV scanning electron microscope operating at 20 kV [5]. Field-emission scanning electron microscopy (FESEM) imaging of the synthesized nanomaterials was performed using Gemini scanning electron microscope (Germany) operating at an accelerating voltage of 9-7 kV.X-ray diffraction (XRD) analysis was carried out in order to determine the crystalline structure of the samples using a Rigaku X-ray diffractometer (model: ULTIMA IV, Rigaku, Japan). The XRD equipment employed operated at a scanning rate of 3° min−1 in 2θ ranging from 5 to 100°, with CuKα X-ray radiation (λ = 1.54056 Å). The crystallite size, Dhkl, was calculated using the Scherrer equation Dhkl = K λ / (Bhkl cos θ), where K is the crystallite-shape factor (0.94), λ is the wavelength of the X-rays, Bhkl is the width of the diffraction peak in radians, and θ is the Bragg angle [26,27].The composition of the samples and the chemical states of the elements were also investigated by X-ray photoelectron spectroscopy (XPS) using Omicron spectrometer assembled with hemispherical analyzer (SPHERA), a 400 Al Kα (1486.7 eV) X-ray source (DAR), a Thermo-Scientific ESCALAB Xi+ spectrometer with a monochromatic Al Kα X-ray source (1486.6 eV), and a spherical energy analyzer operating in the CAE (constant analyzer energy) mode using the electromagnetic lens mode. The CAEs employed for the survey spectra and high-resolution spectra were 100 eV and 50 eV, respectively. In the course of the analysis, the chamber was evacuated at 1 × 10−8 mbar and the spectra were deconvoluted using a Voigt-type function with Gaussian (70%) and Lorentzian (30%) combinations [5].The thermogravimetric analyses (TG) of MoS2, MoSe2–1, MoSe2–2, CoMo, CoMoSeS, NiCoMoSe, and NiCoMo nanocomposites were performed using a Shimadzu TGA-50 thermogravimetric analyzer under a synthetic air gas (5.0 FID) with flow rate of 50 mL min−1, heating rate of 10 °C min−1, and in alumina crucibles.An AFP2 WaveDriver 20 bipotentiostat‒galvanostat (Pine Research Instrumentation) was used to perform the cyclic voltammetry (CV), linear sweep voltammetry (LSV), and chronoamperometry (CA) experiments.A PGSTAT128N potentiostat-galvanostat (Autolab) equipped with FRA2.X module was used to perform the electrochemical impedance spectroscopy (EIS) experiments, running at open-circuit potential (OCP) in the frequency range of 10 mHz-100 kHz, with potential perturbation of 10 mV (rms).The LSV and CA curves normalized by geometric area or mass of catalysts or electrochemical active surface area (ECSA)[5,6,28–31] were corrected by manual 100% iR drop compensation “Ru (= uncompensated resistance) - imparted iR drop” [5,6,32], verified through EIS high frequency intercept (the average Ru values in 1 M KOH was 4.5 Ω), based on the works of Maia et al. (2020, 2021) and Anantharaj et al. (2018) [5,6,32], using the following equation: (1) iR drop free E OER = E RHE − E iR = E RHE − − ( I mea × R u ) where ERHE and Imea are measured potentials and currents, respectively.The ECSA values were calculated using the double-layer capacitance (Cdl) values, which were obtained through CV analysis. The CVs were obtained in a 0.1 V potential window without faradaic current response and centered on OCP. The CV analyses performed in different potential scan rates were initiated from more positive potential to negative potential and the potential was held steady for 10 s at each potential extremity. Thus, the Cdl was determined by the following relation [5,6,32]: (2) C dl = ( ( Δ I 2 ) = ( I a − I c 2 ) ) / ν where Ia and Ic are the anodic and cathodic currents at OCP, respectively, and ν is the potential scan rate. The ECSA value was obtained by dividing the Cdl by the specific capacitance (Cs) values, considering the Cs value as 0.040 mF cm−2 in KOH 1 M [33]. Figs. 1 and S1 present the scanning electron microscopy (SEM), transmission electron microscopy (TEM), high-resolution transmission electron microscopy (HR-TEM), electron diffraction pattern, annular bright-field scanning transmission electron microscopy (ABF-STEM), and energy-dispersive X-ray (EDX) mapping images of MoSe2–1, MoSe2–2, MoS2, NiCoMoSe, NiCoMo, CoMoSeS, and CoMo nanocomposites and nanoparticles.As can be observed in Figure S1, the MoSe2–1, MoSe2–2 and MoS2 nanocomposites exhibited sheet-shaped particles. However, Fig. 1A and the inset of the figure show that the NiCoMoSe nanocomposite was constituted mainly by nanoribbons with extension of up to approximately 5 µm. The TEM images confirmed the presence of nanoribbons with width of 23 nm on average and length ranging from 0.1 to 1 µm (Figs. 1B-C). The HR-TEM images (Figs. 1D-E) showed the presence of dark spots on the nanoribbons, with lattice fringe of 0.22 nm which was assigned to (003) plane of CoMoO4 [34] (Fig. 1E). The inset of Fig. 1E exhibited a crystalline electron diffraction pattern for the NiCoMoSe nanocomposite. The EDX elemental mapping analysis of the NiCoMoSe nanocomposites (Fig. 1F) showed the presence of all the starting hydrothermal synthesis metals as well as Se and O under a seemingly uniform distribution with minor presence of Mo and smaller amount of Se (see Table S1).Although the surface morphology of NiCoMo (Figs. 1G and inset) was found to be similar to that of NiCoMoSe in terms of the presence of ribbons, the nanoribbons in the NiCoMo nanocomposite were found to be covered with small flake-like nanoparticles, which may be linked to the fact that the sample has been subjected to high temperature thermal treatment. The TEM images in Figs. 1H-I showed a noticeable alteration in the structure of the NiCoMo nanocomposite after calcination at 600 ºC compared to the NiCoMoSe nanocomposite (Figs. 1B-C). The nanoribbons, which previously exhibited defined edges, showed a wrinkled, rounded shape with width of 24 nm on average and length ranging from 0.1 to 1 µm (Figs. 1H-I); the diameter of the rounded shape was about 20 nm on average. The nanoparticles can be found to have been well-distributed on the smooth surface of the nanoribbons in the NiCoMo nanocomposite. The HR-TEM images in Figs. 1J-K enabled us to identify the dark rounded nanoparticles distributed over the nanoribbon-shaped structures of the NiCoMo nanocomposite, with lattice fringe of 0.29 nm which was assigned to (220) plane of NiCo2O4 [35,36] (Fig. 1K). The inset of Fig. 1K exhibited a mixed crystalline electron diffraction pattern for the NiCoMo nanocomposite. The elements in the NiCoMo nanocomposite (Fig. 1L) can be found to be uniformly distributed with minor presence of Ni and smaller amount of Se (see Table S1).As can be seen in Fig. 1M, the SEM images of CoMoSeS showed the presence of structures composed of nanoribbons and seemingly rounded nanoparticles with size of approximately 0.1 − 0.2 µm. The images in Figs. 1NO showed the presence of nanoribbons with width of 49 nm and length of around 1 µm as well as rounded nanoparticles with a diameter of 53 nm, both supported on nanosheets with length of 600 nm and width of 200 nm on average. The HR-TEM image displayed lattice fringe of 1.07 nm (Figure S2) for the nanosheets (Figs. 1P and S2) which was assigned to (002) plane of MoS2 [37] and the rounded nanoparticles exhibited lattice fringe of 0.49 nm which was associated with (002) plane of Co2Mo3O8 [14,38] (Fig. 1Q). The inset of Fig. 1Q exhibited a mixed crystalline electron diffraction pattern for the CoMoSeS nanocomposite. The elements on the CoMoSeS nanocomposite can be found to be uniformly distributed with smaller amount of Se, though one can find some agglomeration of Mo and S in some regions of the mapping image (highlighted in Fig. 1R) (see Table S1).As can be observed in Fig. 1S, the SEM images of CoMo showed the presence of structures composed of nanoribbons and seemingly rounded nanoparticles with size ranging from approximately 0.2 − 0.5 µm. The TEM images in Figs. 1T-U showed the presence of nanoribbons with width of 90 nm and length of approximately 557 nm, as well as rounded nanoparticles with diameter of 56 nm on average. The increase in width of the nanoribbons in the CoMo nanocomposites can be attributed to the wrinkling of the nanosheets present in the CoMoSeS nanocomposites prior to calcination at 600 ºC; similarly, the increase in length of the nanoribbons in the CoMo nanocomposites can be attributed to the wrinkling of the nanoribbons present in the CoMoSeS nanocomposites prior to calcination at 600 ºC. The HR-TEM images exhibited lattice fringes of 0.26 nm, 0.15 nm, 0.49 nm, and 0.18 nm which were related to the nanoparticles and corresponded to (−222) and (−424) planes of CoMoO4 [39], (002) plane of Co2Mo3O8 [14,38], and (331) plane of Co3O4 [40,41], respectively (Figs. 1V-W). The junction of different crystal planes revealed different angles; these included the following: crystal planes of CoMoO4 (−222) and Co2Mo3O8 (002) with junction angle of 147.8°; crystal planes of Co3O4 (331) and CoMoO4 (−222) with junction angle of 140.7°; crystal planes of CoMoO4 (−424) and Co3O4 (331) with junction angle of 145.5°. The following angle was recorded for the junction of the same crystal planes: crystal planes of Co3O4 (331) and Co3O4 (331) with junction angle of 103.6° (Figs. 1V-W). The inset of Fig. 1V exhibited a mixed crystalline electron diffraction pattern for the CoMo nanocomposite. The elements in the CoMo nanocomposite can be found to be uniformly distributed with minor presence of Se (Fig. 1X) (see Table S1). Fig. 2 shows the TEM, HR-TEM, electron diffraction pattern, ABF-STEM, and EDX mapping images of NiCoMo-es, CoMoSeS-es, and CoMo-es nanocomposites and nanoparticles.The TEM images of the NiCoMo-es (Fig. 2A and inset) and NiCoMo nanocomposites (Figs. 1H-I) exhibited similar structures (Figs. 1H-I). The nanoribbons of the NiCoMo-es nanocomposite were characterized by a wrinkled, close rounded shape with width of 24 nm on average and length of 0.2 µm (Fig. 2A); the diameter of the rounded shape of the nanoribbons was 22 nm on average. The nanoparticles were found to be well distributed on the smooth surface of the nanoribbons of the NiCoMo-es nanocomposite. The HR-TEM images (Figs. 2B-C) enabled us to identify the dark rounded nanoparticles distributed over the nanoribbon-shaped structures of the NiCoMo-es nanocomposite, with lattice fringes of 0.19, 0.22, 0.26, and 0.39 nm which were assigned to (421), (003), (−222), and (021) planes of CoMoO4 [34,39] (Figs. 2B-C). The junction of the different crystal planes showed the existence of a 115° angle between the CoMoO4 (003) and CoMoO4 (021) planes (Fig. 2B). In addition, a great defect was observed between at least two different crystal planes of CoMoO4 (see the red dashed line in Fig. 2C). The inset of Fig. 2C exhibited a mixed crystalline electron diffraction pattern for the NiCoMo-es nanocomposite. The elements in the NiCoMo-es nanocomposite (Fig. 2D) were also found to be uniformly distributed with the presence of smaller amount of Mo. The identification of Ni (Fig. 2D) in the NiCoMo-es nanocomposite indicated the presence of Ni nanoparticles in the nanocomposite even though we were unable to identify lattice fringes for NiCoO4 nanoparticles as observed for the NiCoMo nanocomposite (Fig. 1K). Also, the fact that Se was not identified in the NiCoMo–es (Fig. 2D) points to the corrosion of this element after the OER long term stability test.The TEM images of CoMoSeS-es (Figs. 2E and inset) showed nanoribbons with width of 7 nm and length of around 59 nm as well as rounded nanoparticles with a diameter of 100 nm on average; both the nanoribbons and nanoparticles were supported on nanosheets (Figs. 2E and inset). The TEM images of the CoMoSeS-es nanocomposite were found to be similar to the TEM images of the CoMoSeS nanocomposite (Figs. 1NO) even under different dimensions; this may be attributed to the fact that different regions of the nanocomposite samples were selected for analysis. The HRTEM images exhibited lattice fringes of 0.23 nm which were related to the nanosheets and corresponded to (103) plane of MoS2 [42,43] (Figs. 2F-G), while the nanoparticles exhibited lattice fringes of 0.26 and 0.49 nm which were associated with (−222) plane of CoMoO4 [39] and (002) plane of Co2Mo3O8 [14,38], respectively (Figs. 2F-G). The junction of the different crystal planes revealed the presence of a 90° angle between the MoS2 (103) and CoMoO4 (−222) planes (Fig. 2F), a 124° angle between the CoMoO4 (−222) and CoMoO4 (−222) planes (Fig. 2F), and a 93° angle between the MoS2 (103) and MoS2 (103) planes (Fig. 2G). In addition, the lattice fringes observed were assigned to MoS2 (nanosheets) and Co2Mo3O8 (002) planes in the CoMoSeS nanocomposite (Figs. 1P-Q). The inset of Fig. 2G exhibited a mixed crystalline electron diffraction pattern for the CoMoSeS-es nanocomposite. There was a uniform distribution of elements in the CoMoSeS-es nanocomposite with minor presence of Se and smaller amount of S (Fig. 2H); the distribution of elements in the CoMoSeS-es nanocomposite was found to be very similar to that observed in the CoMoSeS nanocomposite (Fig. 1R).The TEM images of CoMo-es (Figs. 2I and inset) showed the presence of nanoribbons with width of 49 nm and length of approximately 239 nm, as well as rounded nanoparticles with diameter of 80 nm on average. The TEM images recorded for the CoMo-es nanocomposite were found to be similar to those of the CoMo nanocomposite (Figs. 1T-U) even under different dimensions; this may be linked to the fact that different regions of the nanocomposite samples were selected for analysis. The HRTEM images exhibited lattice fringes of 0.26, 0.49, and 0.18 and 0.23 nm for the nanoparticles which were assigned to (−222) plane of CoMoO4 [39], (002) plane of Co2Mo3O8 [14,38], and (331) and (220) planes of Co3O4 [40,41,44], respectively (Figs. 2J-K). The junction of the different crystal planes revealed the presence of different angles between the nanocomposites: 130° angle between Co2Mo3O8 (002) and Co3O4 (220); 83.4° angle between Co3O4 (220) and CoMoO4 (−222); 53.2° angle between CoMoO4 (−222) and Co3O4 (331); and an angle of 88° between the same crystal planes of Co3O4 (331) and Co3O4 (331) (Figs. 2J-K). Although similar lattice fringe distances were identified for the different crystals, these lattice fringes resulted in different angles (this was probably related to the fact that a different region of the nanocomposite sample was selected for analysis) between the junction of different crystal planes for the CoMo nanocomposite in comparison with the CoMo-es nanocomposite; furthermore, different angles were also observed for the junction of similar crystal planes for the CoMo nanocomposite in comparison with the CoMo-es nanocomposite (compare Figs. 1V-W with 2J-K). The inset of Fig. 2K exhibited a mixed crystalline electron diffraction pattern for the CoMo-es nanocomposite. The elements in the CoMo-es nanocomposite were found to be uniformly distributed with minor presence of Se and smaller amount of S (Fig. 2L), as observed in the CoMo nanocomposite (Fig. 1X).The nanocomposites crystalline structures were investigated by XRD analysis (Figure S3), and the results obtained showed that all the samples were characterized by a mixed composition. The XRD diffraction pattern for the NiCoMoSe nanocomposite (Figure S3) exhibited peaks at 13.6º, 37.8º, and 47.3º which corresponded to the (002), (103), and (105) planes of MoSe2 (JCPDS card 029–0914), respectively. The peak at 17.3º was assigned to the (002) plane of Co2Mo3O8 (JCPDS card 034–0511). The peaks at 26.7º; 33.8º, 39.7º, and 59.9º were associated with the (002), (−222), (003) and (−352) planes of CoMoO4 (JCPDS card 021–0868) and the peak at 62.6º corresponded to the (220) plane of NiO (JCPDS card 047–1049). Based on the diffraction peaks of the nanocomposite, the crystallite size of CoMoO4 was 25.1 nm on average and the lattice fringes assigned to the (003) plane of CoMoO4 was identified in Fig. 1E.Compared to the NiCoMoSe nanocomposite, the NiCoMo nanocomposite exhibited a different diffraction pattern (Figure S3), which resulted from the modification of the crystallinity of the material after calcination at 600 ºC (Scheme 1). The peaks at 18.8º, 31.1º, 36.9º, 44.8º, 59.2º, and 65.0º corresponded to the (111), (220), (311), (400), (551), and (440) planes of NiCo2O4 (JCPDS card 020–0781). The diffraction peaks at 26.4º, 40.2º, 43.3º, and 62.6º were linked to the (002) and (003) planes of CoMoO4 (JCPDS card 021–0868), and the (200) and (220) planes of NiO, respectively (JCPDS card 047–1049). Based on the diffraction peaks of the nanocomposite, the crystallite size of NiCo2O4 was 17.4 nm on average and the lattice fringes assigned to the (220) plane of NiCo2O4 was identified in Fig. 1K.The XRD pattern of the CoMoSeS nanocomposite (Figure S3) indicated the presence of (002), (103), (006) and (105) planes of MoS2 at the following peaks: 14.4°, 39.5°, and 44.3° and 49.9°, respectively (JCPDS card 006–0097). The peak at 17.2° was correlated to the (002) plane of Co2Mo3O8 (JCPDS card 034–0511). The peaks at 26.5°, 33.7°, 47.1°, and 60.4° were linked to the (002), (−222), (241), and (−424) planes of CoMoO4 (JCPDS card 021–0868). Furthermore, the peak at 32.6° was correlated to (100) plane of Co(OH)2 (JCPDF card 074–1057). Based on the diffraction peaks of the nanocomposite, the crystallite size of MoS2 was 54.5 nm on average and the lattice fringes assigned to (002) planes of MoS2 and Co2Mo3O8 were identified in Figures S2 and 1Q, respectively.The change in crystalline structure between the NiCoMoSe and NiCoMo nanocomposites was evidently clear, with the disappearance of the peaks attributed to MoSe2, Co2Mo3O4, and CoSe2 and the presence of the peaks related to NiCo2O4 phase in the NiCoMo nanocomposite. This change was also confirmed by the results obtained from TG analysis (Figure S4) once the NiCoMo nanocomposite was found to present negligible mass loss while the NiCoMoSe nanocomposite exhibited a mass loss starting effectively at 230 °C, reaching a mass loss close to 26% at 600 °C, and finally recording 31% of mass loss at 900 °C (Figure S4).The second mass loss for the NiCoMoSe nanocomposite recorded between 531 and 629 °C coincided approximately with the first mass loss for MoSe2–2 (551–632 °C). MoSe2–2 presented additional (second) mass loss effectively after 776 °C and finally recorded close to 100% of mass loss at 900 °C (Figure S4).The MoSe2–1 (commercial sample) sample presented its first mass loss between 381–463 °C; this was around 170 °C lower compared to the first mass loss observed for MoSe2–2. The difference in temperature was associated with the different structures of the two MoSe2 materials identified through the SEM images (Figure S1). However, the additional (second) mass loss observed for MoSe2–1 was exactly equal to that of MoSe2–2, and the former recorded a mass loss of approximately 100% at 900 °C (Figure S4).The MoS2 sample exhibited its first mass loss between 518–629 °C, followed by an additional (second) mass loss effectively after 767 °C, and finally recording a mass loss of approximately 80% at 900 °C (Figure S4).The first mass losses observed for MoSe2–1 and MoSe2–2 were related to the burning of Se, which was gasified into SeO2 [6,45,46] (Figure S4 and Table S1). The pyrolysis of MoSe2 followed the reaction 2 MoSe2 + 7 O2→ 2 MoO3 + 4 SeO2; as the temperature reached the sublimation temperature of SeO2, SeO2 underwent volatilization [6,46] (see Figure S4). MoO3 sublimed [6,47] effectively after 776 °C in both samples (see Figure S4).The first mass loss observed in MoS2 was related to the burning of S, which was gasified into SO2 [48] (Figure S4 and Table S1). The pyrolysis of MoS2 followed the reaction 2 MoS2 + 7 O2→ 2 MoO3 + 4 SO2; as the temperature reached the sublimation temperature of SO2, SO2 underwent volatilization [48] (see Figure S4). MoO3 sublimed [6,47] effectively in the same temperature in both the MoSe2–1 and MoSe2–2 samples (see Figure S4).For the NiCoMoSe nanocomposite, the first mass loss was related to the partial burning of Se, which was gasified into SeO2 [6,45,46] (Figure S4 and Table S1) and the second mass loss was related to the partial sublimation of MoO3 [6,47]. The occurrence of gasification at much lower temperatures can be attributed to the presence of additional metals (Ni and Co) in the structure of the NiCoMoSe nanocomposite, which facilitated the burning of the nanocomposite in comparison with the pure sample of MoSe2–1 and MoSe2–2 which did not contain these metals.For the NiCoMoSe nanocomposite, the wt.% for Co, Ni, Se, and Mo was 29.8%, 27.4%, 5.7%, and 1.8%, respectively (Table S1). For the NiCoMo nanocomposite, the wt.% recorded for Co, Ni, Se, and Mo was 33.5%, 32.0%, 0.1%, and 6.0%, respectively (Table S1). The burning (Scheme 1) of the NiCoMoSe nanocomposite enriched the Co, Ni, and Mo compositions in the NiCoMo nanocomposite (Table S1). In addition, these composition responses (or changes in composition) (Table S1) were found to be in relative consonance with the EDX mapping images of the NiCoMoSe and NiCoMo nanocomposites (Figs. 1F and 1L).For the CoMoSeS nanocomposite, the first mass loss was related to the partial burning of Se, which was gasified into SeO2 [6,45,46] (Figure S4 and Table S1). The gain in mass observed between 534–624 °C can be attributed to the production of MoO3 and SO2 which remain in the structure of the CoMoSeS nanocomposite as a result of the pyrolysis of MoS2 - based on the reaction 2 MoS2 + 7 O2→ 2 MoO3 + 4 SO2 [48]; MoO3 and SO2 are released/gasified at temperatures greater than 624 up to 748 °C, reaching close to 27% of mass loss at 900 °C (see Figure S4). The CoMo nanocomposite presented negligible mass loss (see Figures S4). For the CoMoSeS nanocomposite, the wt.% recorded for S, Co, Se, and Mo was 4.1%, 49.3%, 3.8%, and 6.0%, respectively (Table S1). For the CoMo nanocomposite, the wt.% recorded for S, Co, Se, and Mo was 0.4%, 45.5%, 0.4%, and 13.8%, respectively (Table S1). The burning (Scheme 1) of the CoMoSeS nanocomposite enriched the Mo composition in the CoMo nanocomposite (Table S1). In addition, the composition responses (Table S1) were found to be in line with the EDX mapping images of the CoMoSeS and CoMo nanocomposites (Figs. 1R and 1X). Figs. 3 and S5–6 present the XPS survey spectra as well as the high-resolution XPS (HR-XPS) spectra for CoMoSeS, CoMo, NiCoMoSe, and NiCoMo, and for the CoMoSeS-es, CoMo-es, and NiCoMo-es nanocomposites.Looking at the survey spectra in Figure S5, one can observe the presence of the peaks (on average) at 54, 64, 106, 231, 284, 397, 531, 643, 714, 780, 855, and 927 eV related to Se 3d, Co 3p, Co 3 s, Mo 3d, C 1 s, N 1 s, O 1 s, Ni LMN, Co LMN, Co 2p, Ni 2p, and Co 2 s respectively, for the NiCoMoSe and NiCoMo, and NiCoMo-es nanocomposites. For the NiCoMo-es nanocomposite, the Se 3d peak was unable to be identified. For the CoMoSeS nanocomposite survey spectrum (Figure S5), the peaks corresponding to Se 3d, Ni LMN, and Ni 2p were unable to be identified, and apart from the existing peaks in the survey spectrum, there is an additional peak around 163 eV, which corresponds to S 2p. For the CoMoSeS-es nanocomposite, the peaks related to S 2p, Mo 3d, and N 1 s were unable to be identified. The XPS survey spectra of the CoMo nanocomposites exhibited peaks related to Co 3p, Co 3 s, Mo 3d, C 1 s, N 1 s, O 1 s, Co LMM, Co 2p, and Co 2 s at approximately 63, 104, 232, 284, 400, 532, 716, 782, and 930 eV, respectively. The C 1 s peak at 284 eV present in the samples was associated with the CP used as supporting material for the samples.For the NiCoMoSe and NiCoMo nanocomposites, the atomic contents recorded for Co and Ni were approximately 10% each (atomic content ratio close to 1:1; see Table S2). The atomic content of N was about 12% (derived from the urea used in the synthesis, Scheme 1), while the atomic content of Se was about 0.9% (see Table S2). A comparison between the NiCoMo nanocomposite and the NiCoMoSe nanocomposite showed that the atomic content of Mo in the former was 8.4 times higher than that observed in the latter and the atomic content of O in the former was 7% lower than that observed in the latter (see Table S2); the lower O atomic content observed in the NiCoMo nanocomposite can be attributed to the fact that this nanocomposite was obtained from the burning of the NiCoMoSe nanocomposite at 600 °C (Scheme 1 and Figure S4).With regard to the NiCoMo-es nanocomposite, the atomic contents recorded for Co and Ni were significantly lower - around half and a quarter, respectively, compared to the Co and Ni atomic contents recorded in the NiCoMo nanocomposite (see Table S2); this implies strong oxidation and loss (corrosion) of Co and Ni from the surface of the NiCoMo-es nanocomposite during OER. The occurrence of strong oxidation is corroborated by the increase in atomic content of O observed in the NiCoMo-es nanocomposite compared to the NiCoMo nanocomposite (see Table S2). In addition, it is clear that the following factors point to the oxidation/corrosion of the NiCoMo nanocomposite when subjected to electrochemical stabilization: i) the non-detection of Se, ii) the decrease of 18% in Mo atomic content; and iii) the decrease of approximately 30% in atomic content of N (see Table S2).With regard to the CoMoSeS nanocomposite, the atomic content recorded for Co was 20%. The atomic content of N was 5.7% (derived from the urea used in the synthesis, Scheme 1), while the atomic contents recorded for S, Mo, and O were 3.5%, 1.3%, and 69.3%, respectively (see Table S2). The atomic contents of O and Co in the CoMoSeS nanocomposite were close to the combined atomic contents of O, Co and Ni in the NiCoMoSe nanocomposite (see Table S2). A comparison of the CoMoSeS and NiCoMoSe nanocomposites showed that the atomic content of Mo in the former was 2.3 times higher than that of the latter, and the atomic content of S in the former (CoMoSeS) was 4.2 times higher than the atomic content of Se in the latter (NiCoMoSe) (see Table S2).For the CoMoSeS-es nanocomposite, the atomic content recorded for Co was 18%; this value was very close to that observed in the CoMoSeS nanocomposite (see Table S2). However, considering that there was a significant increase in the atomic content of O (∼82%) and S, Mo, and N were unable to be detected, this shows that the CoMoSeS nanocomposite underwent corrosion when it was electrochemically stabilized (see Table S2).With regard to the CoMo nanocomposite, the atomic content of Co was half the value recorded for the content of the element in the CoMo-es nanocomposite (the atomic content of Co was 25.6%, which was slightly higher than the Co atomic content recorded in the CoMoSeS nanocomposite; see Table S2). This can be attributed to the following: i) significant reduction of the atomic content of N (to around a quarter of the value recorded before electrochemical stabilization) in the CoMo nanocomposite after electrochemical stabilization (the atomic content of Mo is also reduced to around one third of the value), and ii) a 10% increase in the atomic content of O after electrochemical stabilization (see Table S2); in essence, this suggests the occurrence of oxidation/corrosion of N and Mo in the CoMo nanocomposite after electrochemical stabilization.The Ni 2p high resolution HR-XPS spectra for the NiCoMoSe and NiCoMo, and NiCoMo-es nanocomposites exhibited two pairs of peaks which were linked to Ni 2p3/2 and Ni 2p1/2 levels, with a content ratio of about 2.2:1 [5,49], and two other peaks which were related to satellites peaks [5,49] (Fig. 3). The deconvoluted peaks were attributed to the following: Ni2+ 2p3/2 and Ni2+ 2p1/2 at approximately 854.5 and 872.1 eV; Ni3+ 2p3/2 and Ni3+ 2p1/2 at approximately 856.2 and 874.2 eV; and their respective satellites peaks at 860.3, 878.0, 862.9, and 881 eV [5,49] (Table S3). The presence of Ni3+ and Ni2+ species on the nanocomposite surfaces was confirmed by the spin-orbital splitting observed at around 18 eV between the peaks, in addition to the presence of satellites peaks [5,49] (Fig. 3 and Table S3). With regard to the NiCoMoSe nanocomposite, the content percentages of Ni2+ and Ni3+ were approximately 25.1 and 24.4%, respectively (Table S3). For the NiCoMo nanocomposite, the content percentages of Ni2+ and Ni3+ were 19.9 and 31.3%, respectively, while the content percentages of Ni2+ and Ni3+ in the NiCoMo-es nanocomposite were 32.6 and 21.2%, respectively (Table S3). Interestingly, the content percentages of Ni2+ and Ni3+ appeared to have been inverted in terms of quantity after the NiCoMo nanocomposite was electrochemically stabilized (Table S3). The CoMoSeS and CoMo nanocomposites did not present HR-XPS signal for Ni.The Co 2p HR-XPS spectra exhibited two peaks corresponding to Co 2p3/2 and Co 2p1/2 levels with a ratio of about 2.0:1, in addition to the respective satellite peaks for the CoMoSeS, CoMo, NiCoMoSe, and NiCoMo, as well as for the CoMoSeS-es, CoMo-es, and NiCoMo-es nanocomposites (Fig. 3). The deconvoluted peaks were attributed to the following: Co3+2p3/2 and Co3+2p1/2 at approximately 780 and 794.1 eV; Co2+2p3/2 and Co2+2p1/2 at approximately 782 and 798 eV; and their respective satellites peaks at 782.9, 802, 787, and 804.6 eV [5,6,50] (Table S3). The presence of Co3+ and Co2+ species on the nanocomposites surfaces was further confirmed by the spin-orbital splitting occurring at around 18.8 eV between the peaks, in addition to the presence of the satellites peaks [5,6,50] (Fig. 3 and Table S3). The content percentages recorded for Co3+ and Co2+ in the CoMoSeS nanocomposite were 22.7 and 52.4% respectively. The content percentages of Co3+ and Co2+ recorded in the CoMo nanocomposite were 41.8 and 37.5% respectively. With regard to the NiCoMoSe nanocomposite, the content percentages recorded for Co3+ and Co2+ were 31.3 and 22% respectively. The content percentages recorded in the NiCoMo nanocomposite for Co3+ and Co2+ were 29.5 and 38.1%, respectively (Table S3). The content percentages recorded for Co3+ and Co2+ in the CoMoSeS-es nanocomposite were 22.8 and 51.5%, respectively. With regard to the CoMo-es nanocomposite, the content percentages of Co3+ and Co2+ recorded in the sample were 48 and 33.4%, respectively. The content percentages of Co3+ and Co2+recorded in the NiCoMo-es nanocomposite were 46.2 and 36.6%, respectively (Table S3).The Mo 3d HR-XPS spectra exhibited nearly 2–4 peaks mostly in the 3d5/2 and 3d3/2 regions for the CoMoSeS, CoMo, NiCoMoSe, and NiCoMo nanocomposites, as well as for the CoMo-es and NiCoMo-es nanocomposites (Fig. 3). The deconvolution of Mo 3d HR-XPS spectra resulted in four peaks which were attributed to Mo4+3d5/2 and 3d3/2 species at approximately 231 and 233 eV, respectively, and Mo6+ 3d5/2 and 3d3/2 species at approximately 234 and 235 eV [6,38,51], respectively (Fig. 3 and Table S3). The CoMoSeS nanocomposite exhibited an additional shoulder at 226 eV (Fig. 3) which was attributed to S 2 s [52,53]. The deconvoluted peaks corresponding to Mo4+ species at lower binding energies which were attributed to the NiCoMoSe nanocomposite were no long observed in the Mo 3d HR-XPS spectrum of the NiCoMo nanocomposites (Fig. 3); this shows that the molybdenum species on the surface of these nanocomposites were oxidized to the 6+ state – with content percentage of 100% (see Table S3) – after calcination at 600 °C (Scheme 1). The content percentages of Mo4+ and Mo6+ species recorded in the CoMoSeS nanocomposite were approximately 52 and 48%, respectively (Table S3). The content percentages of Mo4+ and Mo6+ species recorded in the CoMo nanocomposite were 60.5 and 39.5%, respectively (Table S3). The content percentages of Mo4+ and Mo6+ species recorded in the NiCoMoSe nanocomposite were 57.1 and 42.9%, respectively (Table S3). The content percentages of Mo4+ and Mo6+ species recorded in the CoMo-es nanocomposite were 51.2 and 48.8%, respectively (Table S3).The S 2p HR-XPS spectrum related to the CoMoSeS nanocomposite exhibited two peaks in the 2p3/2 and 2p1/2 regions [53,54] and a third peak in the sulfate region [52,54] (Fig. 3); this was the only nanocomposite that presented XPS signal for S. The S 2p HR-XPS spectrum was deconvoluted into the following peaks: S2− 2p3/2 and 2p1/2; S2 2− 2p3/2 and 2p1/2; and sulfate peaks at 162.8 and 164.2; 165.7 and 166.9; and 171.5 eV, respectively [10,52,54–56] (see Table S3). The content percentages recorded for S2− and S2 2− were found to be equal to 42.4% (Table S3).The Se 3d HR-XPS spectrum related to the NiCoMoSe and NiCoMo nanocomposites exhibited one peak and a shoulder (Fig. 3), which were deconvoluted into Se 3d5/2 and Se 3d3/2 levels [13,57] at 53.9 and 54.8 eV, respectively (see Table S3). The content percentages of Se 3d5/2 level recorded in the NiCoMoSe and NiCoMo nanocomposites were 78.4 and 65.1%, respectively (Table S3).The O 1 s HR-XPS spectra exhibited a broad peak and a shoulder for the CoMoSeS, CoMoSeS-es, CoMo, CoMo-es, NiCoMoSe, and NiCoMo nanocomposites, and two peaks for the NiCoMo-es nanocomposite (Figure S6); these were deconvoluted into four peaks centered around 530, 531, 532.5, and 534 eV (Table S3), which were assigned to metal oxides, metal hydroxides, oxygen atoms located at defect sites, and adsorbed water molecules, respectively [13,15,58]. The average content percentages recorded were as follows: M-O: 34%, OH: 43.6%, and O defect sites: 16.7% (Tables S3). These results were in total agreement with the results obtained from the XRD analysis (Figure S3).The N 1 s HR-XPS spectra exhibited a larger broad peak and a shoulder for the CoMoSeS nanocomposite and a broad peak for the CoMo, NiCoMoSe, and NiCoMo nanocomposites (Figure S6). The CoMo-es and NiCoMo-es nanocomposites (Figure S6) exhibited two broad peaks, which were deconvoluted into three or four peaks for different N species coexisting on the surfaces of the nanocomposites; these peaks were centered around the following: i) at 394.7, 399.3, and 402.7 eV which corresponded to N-Mo, NCo, and NH for the CoMoSeS nanocomposite; ii) at 396.5, 399.1, and 404.5 eV which corresponded to N-Mo, NCo, and NH for the CoMo nanocomposite; and iii) at 396, 397.7, 398.7, and 400.5 eV which corresponded to N-Mo, NNi, NCo, and NH, respectively, for the NiCoMoSe, NiCoMo and NiCoMo-es nanocomposites surfaces (Figure S6) [59–61]. The content percentages of NCo recorded in the CoMoSeS and CoMo, and CoMo-es nanocomposites were 57.6% on average, while the content percentages of NNi and NCo combined recorded in the NiCoMoSe and NiCoMo and NiCoMo-es nanocomposites were 84% on average (Tables S3); these results show that N derived from urea is bonded to the metals in the nanocomposites structures.The CV profiles obtained for the bare CP and modified CP electrodes are shown in Figs. 4 A and S7A.The CV profiles obtained for the bare and modified electrodes (Figs. 4A and S7A) exhibited mostly small capacitive current densities, as observed by Bezerra and Maia [5]. However, for the IrO2/CP electrode (inset of Fig. 4A), the responses of the current densities were found to be high and were similar to the result obtained by Souza et al. [62]. Similarly, the responses of the current densities recorded for the RuO2/CP electrode (Figure S7A) were found to be similar to the result obtained by Martini and Maia [6]. For the CoMoSeS/CP electrode (Fig. 4A), the values of the current densities obtained were found to be similar to those observed for the CoMoSe/GNR/CP electrode [6], though with different CV shapes. The increased current densities observed for the modified CP electrodes shown in Figs. 4A and S7A were attributed to the different active sites present in these electrodes.The LSV curves obtained for the bare CP and modified CP electrodes are shown in Figs. 4B and S7B. Considering the overpotential required to achieve a current density of 10 mA cm−2, which corresponds to a solar-to-fuel device operating at 10% efficiency illuminated under 1 sun [63,64], one can say that the best OER electrocatalysts obtained in the present study were IrO2, NiCoMo, CoMoSeS, CoMo, and NiCoMoSe nanocomposites; these catalysts exhibited overpotentials (η j at 10 mA cm‒2) of 280, 356, 375, 375, and 390, respectively, which are superior to the overpotential of the commercial RuO2 (Table S4). The low η value recorded for the IrO2 catalyst was expected since it is a benchmark catalyst for OER [5] and its low η is attributed to its high ECSA value (378.8 cm2, see Table S4); however, the stability of the catalyst was found to be very poor, as will be proved below. The low η value recorded for the NiCoMo nanocomposite was found to be close to the values reported in the literature (Table S5).The mass-specific current density [5,6,28–32,64] obtained at η j of 10 mA cm‒2 was 68.5 A g−1 (Fig. 4B); this value was close to the value reported by Bezerra and Maia [5]. At 10.0 A g−1, the NiCoMo/CP modified electrode recorded the lowest overpotential of 260 mV (Fig. 4B); this value was close to the values reported by Bezerra, Martini, and Maia [5,6], and was relatively close to the overpotential of the IrO2/CP modified electrode - which was 220 mV (Fig. 4B).The specific current densities of the NiCoMo/CP, CoMoSeS/CP, and CoMo/CP modified electrodes (current per ECSA [5,6,28–32,64]; see the values below for ECSA and in Table S4) at 10.0 mA cm−2 by ECSA iR free yielded η of approximately 320 mV (Fig. 4C). As the IrO2/CP modified electrode exhibited considerably high ECSA value (Table S4), its specific current densities disappeared in Fig. 4C.It is worth noting that the addition of Co and Ni to MoS2 and MoSe2 and the formation of the oxides (Figs. 1 and S3) contributed toward the improvement of the OER catalysis (compare Fig. 4B with Figure S7B). This outcome can also be linked to the structures formed in the materials (compare Fig. 1 with Figure S1), since the four samples with similar morphology – NiCoMo, NiCoMoSe, CoMoSeS, and CoMo – presented similar OER catalytic performance.An increase is observed in the ECSA value of the NiCoMo nanocomposite (Table S4) due to the following: i) the NiCoMo nanocomposite is constituted mainly by nanoribbons and nanoparticles, with the nanoparticles supported on the nanoribbons; ii) the presence of NiCo2O4 and NiO oxides in the NiCoMo nanocomposite; and iii) the calcination of the nanoribbons (MoSe2–2) at 600 ºC, which leads to the wrinkling of the nanoribbons, where they present junction (and defect) of different crystal planes for the CoMoO4 oxides. It is worth noting that the increase in the ECSA value of the NiCoMo nanocomposite enhances its catalytic performance in OER.The CoMoSeS and CoMo nanocomposites are constituted mainly by nanosheets, nanoribbons and nanoparticles, and the last two are supported on nanosheets. The hydrothermal heating of MoSe2–1 and MoS2 gives rise to nanoribbons and nanosheets, respectively. When they are calcined at 600 ºC, the nanoribbons (MoSe2–1) become wrinkled and the nanosheets (MoS2) are generally shrunken into nanoribbons. In addition, the CoMoSeS and CoMo nanocomposites exhibit junctions of different crystal planes for MoS2 and Co2Mo3O8, CoMoO4, and Co3O4 oxides (in addition to the presence of hydroxide Co(OH)2). All these factors lead to the enhancement of the OER activity of CoMoSeS and CoMo nanocomposites, thus enabling the catalysts to present the second best OER catalytic performance. In addition to the presence of MoSe2 and CoSe2, Co2Mo3O8, CoMoO4 and NiO oxides, the presence of dark spots (nanoparticles) on the surface of the nanoribbons in the NiCoMoSe nanocomposite did not lead to a significant improvement in the OER catalytic performance of the NiCoMoSe nanocomposite.The improved OER catalytic performance which is linked to the atomic contents recorded in the nanocomposite surfaces (Table S2) – caused by the different components used and the different steps of synthesis employed based on the application of the same element – leads us to the following analytical observations: i) the NiCoMo nanocomposite presents a slightly higher atomic content of Co in comparison with Ni (11.8 and 10.9%, respectively) – considered an ideal atomic content of Co (and Ni) in the nanocomposite surface – and this leads to a considerable improvement in the OER catalytic performance of the nanocomposite; ii) the CoMo nanocomposite presents Co atomic content of 12.7%, which yields a good OER activity for the nanocomposite; iii) the increase in the atomic content of Co to 20.3%, as was the case of the CoMoSeS nanocomposite, does not lead to additional improvements in the OER catalytic performance of the nanocomposite. Finally, the NiCoMoSe nanocomposite presents Co and Ni atomic contents of 10.1 and 11.2%, respectively, and this contributes toward the deterioration of the OER catalytic performance.The Tafel plots are shown in Figs. 4E-F, S7C-D, and S9 (S9 obtained from S8). The NiCoMoSe, CoMoSeS, CoMo, and NiCoMo nanocomposites presented the following Tafel slope values: 59 (closer to the value recorded for IrO2), 60, 63, and 83 mV dec−1, respectively; these values were found to be lower than the values recorded for CoMo/AL, RuO2, MoS2, and MoSe2–2 (120, 138, 175, and 178 mV dec−1, respectively) (Table S4). The Tafel slope values recorded for the samples in this study are close to the values reported in the literature (Table S5). The Tafel slope values obtained from the chronoamperometry data were quite close to the values obtained from the SLV experiments (Figure S9). It is generally accepted that the determining step of the reaction rate is closer to the final step of a series of reactions when a decrease is observed in the Tafel slope; essentially, this is a sign of a good OER electrocatalyst [64]. Thus, the combination of Co and Ni in MoS2 and MoSe2 was found to be convenient as it helped produce efficient OER electrocatalysts.The Tafel slope values obtained suggest the involvement of 3 to 2 electrons (60 and 90 mV dec−1, respectively) [5,6,32] in the OER mechanism involving the application of the following catalysts: NiCoMoSe, CoMoSeS, CoMo, and NiCoMo.The number of electrons released (3 to 2 electrons), which was based on the Tafel slope values, was derived from the following oxidation processes: Co from the 2+to 3+ state (one electron released), Ni from the 2+ to 3+ state (one electron released), and Mo from the 4+ to 6+ state (two electrons released), respectively, for the catalysts containing the respective metals in their compositions [5,6,29,31,32,65,66]. Also, the oxidation states recorded for the aforementioned metals were confirmed by the results obtained from the HR-XPS spectra analyses (Fig. 3).In addition, the mass surface percentage values (Table S2) of Co and/or Ni and N were found to be extremely high (Mo presented relatively lower mass surface percentage values), as confirmed by the XPS survey results (Figure S5); these results were also in agreement with the EDX mass percentage values (the only exception here was N; see Table S1).The electrocatalytic OER mechanism involving these electron numbers can be summarized as follows [5,6,29,31,32]:Initially, the oxides/hydroxides [5,6,32] are involved in the following way: (4) M 2 + − OH + O H − ( aq ) → M 3 + ( O ) − OH + H + ( aq ) + e − (5) M 3 + ( O ) − OH + O H − ( aq ) → M 2 + − OH + H + ( aq ) + O 2 ( g ) + e − where M is Co or Ni. M stands for Mo; note that the mechanism begins with M4+.One needs to consider the occurrence of OER mechanisms via i) electrochemical oxide (reactions 6–8) and ii) oxide (reactions 9–11) [5,6,29,32]: (6) S + O H − ( aq ) → S − OH + e − (7) S − OH + O H − ( aq ) → S − O + H 2 O + e − (8) 2 S − O → 2 S + O 2 ( g ) (9) S + O H − ( aq ) → S − OH + e − (10) 2 S − OH → S − O + S + H 2 O (11) 2 S − O → 2 S + O 2 ( g ) where S stands for surface active sites [5,29], which consist of Co, Ni, and Mo.To study the stability of the catalysts, chronoamperometry analysis was performed for 24 h using a potential that could produce a current density of 10 mA cm−2 (Figure S10). The current vrs. time responses for CP-modified CoMoSeS, CoMo and NiCoMo electrodes are shown in Figure S10. After every 4 hrs, the cell system was shaken to remove eventual O2 covering the catalyst surfaces; this was linked to the presence of pulse current densities (Figure S10).After the long-term stability test, the LSV responses were compared with the LSV initially obtained prior to the test (Fig. 4D). The CoMoSeS-es nanocomposite exhibited a decrease of 43% in the maximum current density (45 mA cm−2) and a negligible shift in overpotential at 10 mA cm−2. The CoMo-es nanocomposite exhibited a decrease of 26% in the maximum current density (60 mA cm−2) and an increase of only 10 mV in overpotential at 10 mA cm−2. The NiCoMo-es nanocomposite exhibited a decrease of only 5% in the maximum current density (33 mA cm−2) and a decrease of 12 mV in overpotential at 10 mA cm−2. These results point to the stability and reliability of the materials in terms of catalytic activity in OER, once they did not show any signals of deactivation, even after operating for a long period of time (after long-term stability test, Fig. 4D). The IrO2-es nanocomposite exhibited a significant shift in overpotential at 10 mA cm−2 – from 280 to 420 mV (Fig. 4D); this clearly points to the instability of IrO2 as a good catalyst for OER long-term stability test.While it is evident that there is a corrosion of nanoparticles, nanosheets and nanoribbons (compare the element atomic contents in Table S2 with the CoMoSeS, CoMo, and NiCoMo nanocomposites) in the CoMoSeS-es, CoMo-es, and NiCoMo-es nanocomposites, the difference in electrochemical stability between these nanocomposites can be associated with the atomic contents of the element recorded on the nanocomposites surfaces (Table S2), and this leads us to the following observations: i) the NiCoMo-es nanocomposite presents a high atomic content of Co in comparison with Ni (5.4 and 2.5%, respectively) – a good amount of Co (and Ni) atomic content on the nanocomposite surface helps ensure a better OER catalytic performance for this nanocomposite; ii) the CoMo-es nanocomposite exhibits Co atomic content of 25.6%, and this negatively affects the OER catalytic performance of the nanocomposite; iii) the Co atomic content of 18.0% in the CoMoSeS-es nanocomposite also negatively affects the OER catalytic performance of the nanocomposite (Table S2). With regard to the CoMoSeS-es nanocomposite, there are no traces of Mo, S, and N (Table S2) on the nanocomposite surface, and this contributes toward worsening the OER catalytic performance of the nanocomposite.The results obtained from the OER activity for the NiCoMo/Au electrode before and after the long-term stability test in O2-saturated purified 1 M KOH solution (Fe free) (Figure S11) presented responses quite similar to those shown in Fig. 4D obtained from the application of 1 M KOH solution containing Fe impurities. The conclusion that can be drawn here is that the Fe impurities present in the 1 M KOH solution are clearly not responsible [62] for the high electrocatalytic performance and stability of the NiCoMo catalyst.Figure S12 presents the ring current responses obtained for the bare Pt ring along with the HLS curves for the CoMoSeS/Au and NiCoMo/Au disk electrodes; the figure shows the occurrence of residual currents related to the Pt oxidation without any currents related to the oxidation of HO2 − which could be derived from OH-oxidation during OER. This implies that the OER effectively resulted in O2 production instead of the production of HO2 −[5,6].The integration (Figure S13B) of CV responses (Figure S13A) was used to quantify the surface concentration of active sites per cm2 (Γ) [5,6,31,32,66]. Taking into account the values related to the current densities and redox pairs, the CVs recorded for the CoMoSeS/Au and NiCoMo/Au disk electrodes were considerably different (Figure S13A). The CoMoSeS/Au disk electrode exhibited a redox pair at 1.1 V with high current densities; this redox pair was attributed to Co2+/3+ oxidation/reduction [6,66] and Co2+/3+ was the main active site present in the CoMoSeS electrocatalyst. The NiCoMo/Au disk electrode exhibited only shoulders with maximum small current densities at 0.94 and 0.58 V, respectively, for Ni2+/3+ oxidation and Ni3+/2+ reduction [5,65] (Figure S13A-B), where Ni2+/3+ was the main active site present in the NiCoMo electrocatalyst.The CV integration (Figure S13B) values divided initially by 0.05 V s − 1 and additionally divided by the electron charge (1.602 × 10−19C) [5,6,32] yielded Γ values of 1.915 × 1016 and 7.52 × 1013 atoms cm−2 for the CoMoSeS/Au and NiCoMo/Au disk electrodes, respectively. The Γ values obtained here were close to the Γ values reported by Bezerra, Martini, and Maia [5,6]. The Γ values and hydrodynamic SLV responses (Figure S13C) were used to calculate the relationship between the turnover frequency (TOF, see equation S1) and potential (Fig. 5 ).The TOF values obtained for the CoMoSeS/Au disk electrode were 0.42 and 13.37 s − 1 at 1.56 and 1.68 V, respectively, and the TOF values obtained for the NiCoMo/Au disk electrode were 1.06 and 331.26 s − 1 at 1.50 and 1.66 V, respectively (Fig. 5). These TOF values were found to be close to the values reported by Bezerra, Martini, and Maia [5,6].The faradaic efficiency (FE, see equation S2) relative to the potential (Figure S13F) was calculated using the current densities of the hydrodynamic linear potential scan starting from 1.0 V for the NiCoMo/Au modified disk electrode in N2‒saturated 1.0 M KOH simultaneously with the current densities of the bare Pt ring set at 0.4 V (Figure S13E) [5,6]. The O2 produced from the OER in the NiCoMo/Au modified disk electrode under purely diffusion control [5,6,29] was detected in the bare Pt ring electrode.However, as the O2 produced from the OER in the disk electrode increased at higher current densities, the O2 was found to accumulate in the disk-ring interspace [67] (Teflon interspace, Figure S13D), and despite using a Ti or Au thin wire close to the Teflon interspace surface to dislodge the O2 bubbles formed specifically on the interface between the disk and ring spacer [67] (Teflon interspace), we were not able to improve the ring current densities in the potential range of the high current densities in the disk (Figure S13E); this resulted in a narrow potential window where FE could be determined with some degree of reliability (Figure S13F).The FE values obtained for the NiCoMo/Au disk electrode were close to 100% at 1.55 (without Ti or Au wire), 1.57 (with Au wire), and 1.59 V (with Ti wire) (Figure S13F). The decrease observed in the FE at more positive potentials (Figure S13F) was attributed to both the huge production of O2 (including O2 bubbles [68]) and the bubbles nucleation in the Teflon spacer between the disk and the ring [69], which was caused by a sudden increase in gas concentration in the solution that flowed past the spacer [70], decreasing the current densities on the surface of the bare Pt ring [5,6] (Figures S13E).The double layer capacitance (CDL) was obtained from CV in a non-faradaic potential region at different scan rates [5,6,68] (Figure S14 and Eq. (2)). CDL divided by specific capacitance [33] yields the electrochemically active surface area (ECSA) [5,6,68] of the catalytic surface (Figure S15), as shown in Table S4. The calculated ECSA values obtained for the bare electrode and the modified MoS2, MoSe2–2, RuO2, and IrO2 electrodes were approximately 0.04, 0.04, 0.29, 10.8, and 378.8 cm2, respectively. Considering the geometric current densities of the electrode, the extremely high ECSA value obtained for the CP-modified IrO2 electrode is found to be responsible for the improvement in OER activity presented by the catalyst. When the specific current densities of the electrode are taken into account, one notices that the extremely high ECSA value obtained for the CP-modified IrO2 electrode is responsible for the worsening of the OER activity observed for the catalyst (Figs. 4B-C). The CoMoSeS, CoMo and CoMo/AL modified electrodes exhibited ECSA values of approximately 0.05 cm2; this result shows that there was no variation in the number of active sites after the calcination and acid leaching process during the synthesis. The NiCoMoSe and NiCoMo recorded ECSA values of 0.05 and 0.7 cm2, respectively; these relatively low ECSA values were responsible for the good OER activity obtained in these catalysts considering the geometric and specific current densities of the nanocomposites (Figs. 4B-C).With the exception of RuO2 and IrO2, all the samples recorded an increase in ECSA values after the OER and long-term stability (when applied) experiments; this justifies the inefficient performance of IrO2 after OER long-term stability test (Fig. 4D). The NiCoMo nanocomposite recorded a 1.74-fold increase in the ECSA value after OER in comparison to its initial ECSA value, and its ECSA value after the stability experiment was equal to that observed after the OER experiment. The CoMoSeS and CoMo nanocomposites exhibited very similar behavior, with both recording an 8.6-fold increase in the ECSA values after OER in comparison to their initial ECSA values; furthermore, both nanocomposites recorded about 1.25-fold increase in the ECSA values after the stability test compared to the ECSA values they recorded after OER. These increases in ECSA value justify the efficient performance of NiCoMo, CoMoSeS, and CoMo after the OER long-term stability test (Fig. 4D).Electrochemical impedance spectroscopy analyses were performed for the bare CP and the CP-modified electrodes at OCP in N2-saturated 1 M KOH before and after OER, and after long-term stability experiments; the results obtained are shown by Nyquist plots in Figure S16. There were very small variations in Ru values for the experiments conducted before and after OER experiments; the mean Ru values recorded for the CP-modified electrodes and the bare CP electrode were 4.5 Ω and 6.4 Ω, respectively (Figure S16). The average Rct values obtained for the experiments conducted before and after OER, and after long-term stability were as follows: 2 Ω for CoMoSeS/CP, CoMo/CP, NiCoMoSe/CP, MoS2/CP, and RuO2/CP modified electrodes; 8 Ω for the MoSe2–2/CP modified electrode; 40Ω for the NiCoMo/CP modified electrode; 230 Ω for the CoMo/AL/CP modified electrode (Figure S16). The IrO2/CP modified electrode presented Rct value of 2 Ω before OER; there was a significant increase in the Rct value after the OER and the long-term stability test; this points to the instability of the IrO2/CP modified electrode after the long-term stability test (Fig. 4E).The NiCoMo/CP modified electrode recorded a decrease in the Rct value before the OER experiment compared to the value it recorded after the OER experiment; this explains why it was chosen as the best OER electrocatalyst in the present study. As the CoMoSeS/CP and CoMo/SP modified electrodes presented very low Rct values, these electrodes were chosen as the second best OER electrocatalysts (Figure S16).MoSe2–2 presents a slightly higher Rct value in relation to MoS2; thus, when the oxide nanoparticles that compose the NiCoMo nanocomposite are supported on MoSe2–2 nanoribbons, the Rct value presented by this nanocomposite continues to be slightly higher. The NiCoMoSe nanocomposite presents a relatively smaller Rct value in comparison with the NiCoMo nanocomposite; this can be mainly attributed to the different oxides (less resistant to charge transfer; for example, NiCo2O4 was not identified in the NiCoMoSe nanocomposite) that constitute the oxide nanoparticles supported on MoSe2–2 nanoribbons.As MoS2 presents relatively lower resistance to charge transfer in comparison with MoSe2, the oxides nanoparticles supported on the nanosheets and/or nanoribbons (MoS2) tend to keep the Rct values rather low in the CoMoSeS and CoMo nanocomposites.The present study showed that the factors responsible for the best OER catalytic response for the NiCoMoSe, NiCoMo, CoMo and CoMoSeS nanocomposites included the following: (1) Firstly, the presence of different oxide nanoparticles supported on MoSe2 nanoribbons helped enhance the OER catalytic response. For CoMoSeS in particular, the presence of different oxide nanoparticles supported on MoS2 nanosheets and MoSe2 nanoribbons contributed toward an improvement in the OER catalytic activity. (2) The second factor that contributed toward the improvement in catalytic activity was related to the main oxides identified in the synthesis process; these oxides included CoMoO4, NiCo2O4 (and CoMoO4), Co2Mo3O8 (and CoMoO4), and CoMoO4, Co2Mo3O8, and Co3O4, for the NiCoMoSe, NiCoMo, CoMoSeS, and CoMo nanocomposites, respectively. Apart from that, the junction of the different oxide crystal planes generated additional suitable sites that contributed toward the improvement of OER electrocatalysis. (3) Thirdly, the N atoms, from the urea used during the synthesis, bonded to relatively high amount of Ni and/or Co (and Mo) on the surface of the nanocomposites and this helped enhance OER electrocatalysis. Similarly, the presence of Ni2+ and Ni3+, and/or Co2+ and Co3+, and Mo4+ and Mo6+ on the nanocomposite surfaces also contributed toward enhancing the OER electrocatalytic activity of the nanocomposites. (4) Finally, the electrons released (3 to 2 electrons) from the oxidation of the following: i) Co from the 2+ to 3+ state (one electron released); ii) Ni from the 2+ to 3+ state (one electron released); and iii) Mo from the 4+ to 6+ state (two electron released), contributed toward enhancing the OER electrocatalytic activity. Firstly, the presence of different oxide nanoparticles supported on MoSe2 nanoribbons helped enhance the OER catalytic response. For CoMoSeS in particular, the presence of different oxide nanoparticles supported on MoS2 nanosheets and MoSe2 nanoribbons contributed toward an improvement in the OER catalytic activity.The second factor that contributed toward the improvement in catalytic activity was related to the main oxides identified in the synthesis process; these oxides included CoMoO4, NiCo2O4 (and CoMoO4), Co2Mo3O8 (and CoMoO4), and CoMoO4, Co2Mo3O8, and Co3O4, for the NiCoMoSe, NiCoMo, CoMoSeS, and CoMo nanocomposites, respectively. Apart from that, the junction of the different oxide crystal planes generated additional suitable sites that contributed toward the improvement of OER electrocatalysis.Thirdly, the N atoms, from the urea used during the synthesis, bonded to relatively high amount of Ni and/or Co (and Mo) on the surface of the nanocomposites and this helped enhance OER electrocatalysis. Similarly, the presence of Ni2+ and Ni3+, and/or Co2+ and Co3+, and Mo4+ and Mo6+ on the nanocomposite surfaces also contributed toward enhancing the OER electrocatalytic activity of the nanocomposites.Finally, the electrons released (3 to 2 electrons) from the oxidation of the following: i) Co from the 2+ to 3+ state (one electron released); ii) Ni from the 2+ to 3+ state (one electron released); and iii) Mo from the 4+ to 6+ state (two electron released), contributed toward enhancing the OER electrocatalytic activity.The presence of MoS2 nanosheets and MoSe2 nanoribbons as supporting material for the metal oxides led to a decrease in both the charge transfer resistance and the ECSA values; this resulted in the improvement of the OER electrocatalytic activity of the electrocatalysts.The TOF and FE values obtained for the NiCoMo and CoMoSeS nanocomposites were high under low potentials. In addition, after the OER long-term stability test, the NiCoMo, CoMoSeS, and CoMo nanocomposites were found to be stable despite the fact that some elements of corrosion were detected; this shows that the electrocatalysts have high OER electrocatalytic activity.The Fe impurities present in the 1 M KOH solution were not found to be responsible for the high OER electrocatalytic performance and stability of the electrocatalysts.The authors are grateful to the LCE/DEMa/UFSCar for the support with TEM analyses and to LAMAS – Laboratório Multiusuário de Análise de Superfícies from UFRGS, for providing us with the XPS facilities. The authors also acknowledge the financial support provided by the following Brazilian funding agencies: CNPq (grants 303759/2014-3, 303351/2018-7, and 405742/2018-5), Fundect-MS (grant 59/300.184/2016), CAPES-PRINT (grant 88881.311799/2018-01), and CAPES – Finance Code 001. This study was partly funded by the Federal University of Mato Grosso do Sul – UFMS/MEC – Brazil. L.S.B. and B.K.M. are grateful to CAPES for the fellowship granted them during the course of this research.

It is well known that the benchmark electrocatalysts for OER in alkaline solution are RuO2 and IrO2; however, the high cost, scarcity, and instability of these metal oxides impede their ample use in OER processes, and this has fueled the search for cheap Earth-abundant elements which are equally efficient for application as electrocatalysts in OER. The present work reports the use of hydrothermal and calcination methods for the synthesis of nanocomposites made up of Ni and/or Co salts and urea in combination with MoSe2 and MoS2 and their application as efficient and stable electrocatalysts for OER in alkaline solution. The NiCoMoSe, NiCoMo, CoMoSeS, and CoMo nanocomposites constructed in the present study presented high OER electrocatalytic activity and stability mainly as a result of the following: the combination of N atoms bonded to Ni and/or Co (and Mo); the electrons released from the oxidation of Co from the 2+ to 3+state, Ni oxidation from the 2+ to 3+state, and Mo oxidation from the 4+ to 6+state; and the metal oxides (CoMoO4, NiCo2O4, Co2Mo3O8, and Co3O4) supported on MoS2 nanosheets and MoSe2 nanoribbons which contributed to a decrease in charge transfer resistance, apart from keeping the ECSA values relatively low.